UNIVERSITY OF MANCHESTER CEAS
245
1 UNIVERSITY OF MANCHESTER – CEAS A compositional breakage equation for first break roller milling of wheat A Thesis submitted to the University of Manchester for the degree of Doctor of Philosophy in the Faculty of Engineering and Physical Sciences SILVIA PATRICIA GALINDEZ NAJERA 2014 Supervisor: Prof. Grant Campbell Co-supervisor: Prof. Colin Webb Satake Centre for Grain Process Engineering School of Chemical Engineering and Analytical Science
Transcript of UNIVERSITY OF MANCHESTER CEAS
1
UNIVERSITY OF MANCHESTER – CEAS
A compositional breakage equation for
first break roller milling of wheat
A Thesis submitted to the University of Manchester for the degree of Doctor of Philosophy in the Faculty of Engineering and Physical Sciences
SILVIA PATRICIA GALINDEZ NAJERA
2014
Supervisor: Prof. Grant Campbell
Co-supervisor: Prof. Colin Webb
Satake Centre for Grain Process Engineering
School of Chemical Engineering and Analytical Science
Table of Contents
2
TABLE OF CONTENTS
TABLE OF CONTENTS .................................................................................................................... 2
LIST OF TABLES .............................................................................................................................. 7
LIST OF FIGURES ............................................................................................................................ 8
ABSTRACT ...................................................................................................................................... 11
ABBREVIATIONS .......................................................................................................................... 13
NOMENCLATURES ....................................................................................................................... 15
DECLARATION .............................................................................................................................. 17
COPYRIGHT STATEMENT ........................................................................................................... 17
DEDICATION .................................................................................................................................. 18
ACKNOWLEDGEMENTS .............................................................................................................. 19
CHAPTER 1 MODELLING OF WHEAT BREAKAGE ............................................................ 20
1.1 WHEAT, AN IMPORTANT CEREAL ................................................................................. 20
1.2 DEBRANNED WHEAT AND MODELLING OF WHEAT ROLLER MILLING.............. 21
1.3 SCOPE OF THE CURRENT WORK .................................................................................... 25
CHAPTER 2 UNDERSTANDING WHEAT, MILLING AND DEBRANNING ....................... 28
2.1 INTRODUCTION ...................................................................................................................... 28
2.2 WHEAT: A UNIQUE CEREAL GRAIN ................................................................................... 28
2.3 CLASSIFICATION OF WHEAT ............................................................................................... 31
2.4 THE STRUCTURE OF THE WHEAT KERNEL...................................................................... 32
2.4.1 The Endosperm................................................................................................................... 36
2.4.2 The Bran.............................................................................................................................. 37
2.4.3 The Germ............................................................................................................................ 39
2.5 CHEMICAL COMPOSITION OF WHEAT .............................................................................. 41
Table of contents
3
2.6 USES OF WHEAT ..................................................................................................................... 44
2.7 WHEAT MILLING .................................................................................................................... 45
2.8 DEBRANNING OF WHEAT – A NEW FRACTIONATION TECHNOLOGY ...................... 54
2.8.1 Definition and benefits of debranning................................................................................ 54
2.8.2 Debranning processes......................................................................................................... 55
2.9 SUMMARY ................................................................................................................................ 57
CHAPTER 3 UNDERSTANDING THE NATURE OF WHEAT BREAKAGE ........................ 59
3.1 INTRODUCTION ...................................................................................................................... 59
3.2 THE BREAKAGE EQUATION FOR ROLLER MILLING OF WHEAT ................................ 59
3.2.1 The Normalised Kumaraswamy Breakage function........................................................... 63
3.3 BIOCHEMICAL MARKERS AND “FINGERPRINTS” .......................................................... 70
3.3.1 Microscopical methods....................................................................................................... 70
3.3.2 Fluorescence techniques..................................................................................................... 72
3.3.3 Wet chemistry and Infrared methods coupled with multivariate analysis.......................... 73
3.4 COMPOSITIONAL BREAKAGE EQUATIONS ..................................................................... 77
3.11 SUMMARY .............................................................................................................................. 80
CHAPTER 4 MODELLING FIRST BREAK MILLING OF DEBRANNED WHEAT ............. 81
4.1 INTRODUCTION ...................................................................................................................... 81
4.2 MATERIALS .............................................................................................................................. 82
4.3 EXPERIMENT PROCEDURES ................................................................................................ 82
4.3.1 Characterisation of wheat by SKCS.................................................................................... 82
4.3.2 Conditioning of wheat......................................................................................................... 83
4.3.3 Debranning of wheat kernels.............................................................................................. 85
4.3.4 Milling of wheat.................................................................................................................. 86
4.3.5 Sieve Analysis of milled fragments.................................................................................... 87
Table of contents
4
4.3.6 Mathematical Analysis........................................................................................................ 88
4.4 RESULTS AND DISCUSSION ................................................................................................. 89
4.4.1 Bran removal....................................................................................................................... 89
4.4.2 Effect of debranning on the parameters of the DNKBF..................................................... 92
4.4.3 Parameter a......................................................................................................................... 93
4.4.4 Parameter α......................................................................................................................... 94
4.4.5 Parameters m1 and n1........................................................................................................... 95
4.4.6 Parameters m2 and n2........................................................................................................... 97
4.4.7 Overall effect of debranned wheat on the DNKBF............................................................. 98
4.4.8 Overall effect of roll gap................................................................................................... 101
4.4.9 The suggested nature of Type 1 and Type 2 particles...................................................... 106
4.5 SUMMARY .............................................................................................................................. 109
CHAPTER5 CHARACTERISATION OF TISSUES AND MILLED FRACTIONS ................ 110
5.1 INTRODUCTION .................................................................................................................... 110
5.2 MATERIALS ............................................................................................................................ 111
5.3 EXPERIMENTAL PROCEDURES ......................................................................................... 111
5.3.1 Dissection of wheat components into botanical tissues.................................................... 111
5.3.2 Moisture content................................................................................................................ 112
5.3.3 Sample preparation for FTIR analysis.............................................................................. 113
5.3.4 Fourier Transform Infrared (FTIR) spectroscopy............................................................. 115
5.3.5 Mathematical analysis....................................................................................................... 117
5.5 MANUAL ISOLATION OF WHEAT KERNEL TISSUES .................................................... 118
5.6 FTIR-ATR ANALYSIS OF GROUND BOTANICAL TISSUES ........................................... 120
5.7 PRINCIPAL COMPONENT ANALYSIS ON GROUND ISOLATED TISSUES ................. 121
5.7.2 Improving the separation of the clusters by using mathematical pre-treatment tools...... 129
5.7.3 PCA analysis of milled samples....................................................................................... 132
5.8 CALCULATING THE PROPORTION OF BOTANICAL TISSUES..................................... 135
Table of contents
5
5.9 SUGAR QUANTIFICATION BY HPLC ................................................................................ 139
5.9.1 Stock and standard solutions............................................................................................. 139
5.9.2 Sample preparation for chemical hydrolysis..................................................................... 140
5.9.3 HPLC analysis.................................................................................................................. 140
5.10 SUMMARY ............................................................................................................................ 146
CHAPTER 6 DEVELOPING A COMPOSITIONAL BREAKAGE EQUATION ................... 148
6.1 INTRODUCTION .................................................................................................................... 148
6.2 EXPERIMENT PROCEDURES .............................................................................................. 148
6.2.1 Sample preparation for FTIR analysis.............................................................................. 149
6.3 DERIVATION OF A COMPOSITIONAL BREAKAGE EQUATION .................................. 149
6.4 COMPOSITIONAL BREAKAGE FUNCTIONS .................................................................... 156
6.5 FINDING THE CONCENTRATION FUNCTIONS USING THE DNKB FUNCTIONS...... 160
6.6 SUMMARY .............................................................................................................................. 191
CHAPTER 7 CONCLUSIONS AND RECOMMENDATIONS ............................................... 193
7.1 INTRODUCTION .................................................................................................................... 193
7.2 PROGRESS MADE IN THE CURRENT THESIS.................................................................. 194
7.2.1 Modelling first break milling of debranned wheat using the DNKBF............................. 194
7.2.2 Characterisation of wheat botanical tissues and milled fractions with FTIR and sugar
profiles using HPLC.................................................................................................................. 195
7.2.3 Development of a compositional breakage equation for wheat milling........................... 196
7.3 RECOMMENDATIONS FOR FUTURE WORK ................................................................... 197
REFERENCES ............................................................................................................................... 201
APPENDICES ................................................................................................................................ 214
APPENDIX 1 .................................................................................................................................. 214
APPENDIX 2 .................................................................................................................................. 215
Table of contents
6
APPENDIX 3 .................................................................................................................................. 219
APPENDIX 4 .................................................................................................................................. 220
APPENDIX 5 .................................................................................................................................. 244
List of Tables
7
LIST OF TABLES
Table 2.1 Chemical composition of the whole wheat grain and its various components. ................ 41
Table 2.2 Main systems of the Mill flow sheet ................................................................................. 50
Table 2.3 Comparison among some existing debranning processes.. ............................................... 56
Table 4.1 Average and standard deviation (SD) values of SKCS kernel weight, diameter,
hardness and moisture content of Mallacca and Consort wheats. ............................................. 84
Table 5.1 Summary of the milled fractions analyzed. .................................................................... 114
Table 5.2 Moisture content of both wheat types and the botanical components.............................119
Table 5.3 Example of height data introduced in the spreadsheet to quantify the relative
proportion of each botanical component in the unknown sample..........................................137
Table 5.4 Sugars composition of two wheat grain tissues, two milled fractions from Mallacca
wheat and the results reported by Barron et al. (2007)............................................................143
Table 6.1 Particle size distributions and composition of size fractions following milling of
Mallacca and Consort wheats .................................................................................................. 161
Table 6.2 Fitted DNKBF parameters. ............................................................................................. 165
List of Figures
8
LIST OF FIGURES
Figure 2.1 Structure of the wheat kernel.. ......................................................................................... 32
Figure 2.2 Hyperspectral imaging of transversal cuts of Consort and Mallacca. ............................. 34
Figure 2.3 Ideal relation between botanical components and milled fractions.. ............................... 35
Figure 2.4 Scanning electron micrograph of Consort (soft) wheat.. ................................................. 37
Figure 2.5 Fluorescence micrograph of a transversal cut of Mallacca wheat.. ................................. 38
Figure 2.6 Micrograph of pure dissected germ tissue from Mallacca wheat.. .................................. 40
Figure 2.7 Average carbohydrate composition of whole mature wheat kernels. .............................. 43
Figure 2.8 Saddlestone. ..................................................................................................................... 45
Figure 2.9 A Quern stone from the Iron Age. ................................................................................... 46
Figure 2.10 Possible evolution of the milling devices. ..................................................................... 47
Figure 2.11 Simplified diagram of a typical flour milling process.. ................................................. 49
Figure 2.12 Flour milling process.. ................................................................................................... 51
Figure 2.13 Illustration of roll dispositions. ...................................................................................... 52
Figure 3.1 Non-cumulative and Cumulative form of the Double NKBF. ......................................... 68
Figure 4.1 Mallacca, hard wheat, Consort, soft wheat ...................................................................... 82
Figure 4.2 Single Kernel Characterization System. .......................................................................... 83
Figure 4.3 The Satake TM-05C (laboratory debranner).. ................................................................. 85
Figure 4.4 Satake STR-100AU Test Roller Mill. ............................................................................. 86
Figure 4.5 The Satake PLSB Series 2000 Laboratory Sifter. ........................................................... 87
Figure 4.6 Percentage of wheat mass removed at different debranning times and the rate of
wheat mass removal.. ................................................................................................................. 91
Figure 4.7 Malacca and Consort wheats at 0, 30 and 60 s of debranning. ........................................ 92
Figure 4.8 Values of collapsing parameter a at different debranning times. .................................... 94
Figure 4.9 Parameter α as a function of debranning time ................................................................. 95
Figure 4.10 Parameters m1 and n1 as a function of debranning time................................................. 97
Figure 4.11 Parameters m2 and n2 as a function of debranning time................................................ 98
Figure 4.12 PSD as described by the DNKBF from Mallacca and Consort wheats. ...................... 100
List of Figures
9
Figure 4.13 Experimental data, cumulative and non-cumulative DNKBF for Mallacca wheat
at 0 s and 60 s of debranning .................................................................................................. 102
Figure 4.14 Experimental data, cumulative and non-cumulative DNKBF for Mallacca wheat
at 0 s and 60 s of debranning. .................................................................................................. 103
Figure 4.15 Experimental data, cumulative and non-cumulative DNKBF for Consort wheat
at 0 s and 60 s of debranning. .................................................................................................. 104
Figure 4.16 Experimental data, cumulative and non-cumulative DNKBF for Consort wheat
at 0s and 60s of debranning. .................................................................................................... 105
Figure 4.17 Example of particles resulting from first break milling of Consort wheat. ................. 107
Figure 4.18 Effect of debranning on wheat breakage. .................................................................... 108
Figure 5.1 Hand dissections of wheat botanical tissues.. ................................................................ 112
Figure 5.2 Ball Mill used for grinding the botanical tissues. .......................................................... 113
Figure 5.3 Falling number hammer mill. ........................................................................................ 114
Figure 5.4 Spectrum two (FTIR-ATR) spectrometer. ..................................................................... 116
Figure 5.5 The four major wheat components dissected from Mallacca wheat. ............................. 118
Figure 5.6 FTIR-ATR spectra of milled Mallacca dissected wheat grain. ..................................... 120
Figure 5.7 PCA scores of the four botanical tissues. ...................................................................... 123
Figure 5.8 Loading vectors associated to PC1 and PC2 ................................................................. 124
Figure 5.9 Baseline corrected and normalised FTIR-ATR spectra and second derivative
computed for each botanical tissue .......................................................................................... 126
Figure 5.10 Baseline corrected and normalised FTIR-ATR spectra and second derivative
computed for each botanical tissue. ......................................................................................... 127
Figure 5.11 PCA scores of the four botanical tissues. Whole spectra (4000 to 550 cm–1
). ............ 129
Figure 5.12 PCA scores of the four botanical tissues. Spectral range from 810 to 1800 cm–1
. ...... 130
Figure 5.13 Spectra of different milled fractions.. .......................................................................... 132
Figure 5.14 Base line corrected and normalised spectral data and Second derivative
computed. Range (1800-810 cm-1
).. ........................................................................................ 133
Figure 5.15 PCA scores of wheat milled samples........................................................................... 134
Figure 5.16 Loading vectors associated with PC1. ......................................................................... 135
Figure 5.17 HPLC equipment used. ................................................................................................ 141
Figure 5.18 HPLC chromatograms of pericarp and endosperm after hydrolysis. ........................... 142
List of Figures
10
Figure 5.19 HPLC chromatograms of the sugars present in the two samples of Mallacca
wheat milled. ........................................................................................................................... 143
Figure 6.1 Contrived example. ........................................................................................................ 151
Figure 6.2 Non-cumulative form of the contrived example. ........................................................... 152
Figure 6.3 Broken wheat particle.....................................................................................................158
Figure 6.4 Cumulative particle size and component distributions for Mallacca (S-S). ................. 163
Figure 6.5 Non-cumulative particle size and component distributions for Mallacca (S-S). ........... 164
Figure 6.6 Concentration functions for Pericarp, Intermediate layer, Aleurone and starchy
endosperm. ............................................................................................................................... 169
Figure 6.7 Cumulative particle size and component distributions for Mallacca (D-D). ................. 171
Figure 6.8 Non-cumulative particle size and component distributions for Mallacca (D-D). .......... 172
Figure 6.9 Concentration functions for Pericarp, Aleurone, Endosperm and Intermediate layer... 173
Figure 6.10 Cumulative particle size and component distributions for Consort (S-S). ................. 175
Figure 6.11 Non-cumulative particle size and component distributions for Consort (S-S). ........... 176
Figure 6.12 Concentration functions for Pericarp, Intermediate layer, Aleurone and starchy
endosperm. ............................................................................................................................... 179
Figure 6.13 Cumulative particle size and component distributions for Consort (D-D). ................. 181
Figure 6.14 Non-cumulative particle size and component distributions for Consort (D-D). .......... 182
Figure 6.15 Concentration functions for Pericarp, Aleurone, Intermediate layer and Endosperm. . 183
Figure 6.16 Pericarp, Intermediate layer and Aleurone distributions for Mallacca and Consort.... 185
Figure 6.17 Pericarp, Intermediate layer, Aleurone and starchy endosperm distributions for
Mallacca and Consort. ............................................................................................................. 188
Figure A5.1 Typical data set for a PLS.. ......................................................................................... 244
Abstract
Silvia Galindez PhD Thesis August 2014 11
A COMPOSITIONAL BREAKAGE EQUATION FOR
FIRST BREAK ROLLER MILLING OF WHEAT
ABSTRACT
The particle size distribution produced from first break roller milling of wheat determines
the flows through the rest of the mill and hence the quality of the final flour, and is affected
by debranning and by the operation of the roller mill. The Double Normalised
Kumaraswamy Breakage function (DNKBF) gives a quantitative basis to describe
breakage during first break milling of wheat and to interpret effects. Previous work
developed and extended the breakage equation in order to understand and predict wheat
breakage based on distributions of the grain characteristics and the operating parameters of
the mill. However, broken particles vary in composition as well as size; therefore the
primary objective of the current work was to extend the DNKBF during first break milling
to include particle composition, using fingerprints of pericarp, aleurone, endosperm and
germ. Meanwhile, debranning is a technology that has enhanced flour milling in recent
years, leading to improvements in quality that are not well understood but that start with
the effect on milling. A second objective of the current work was therefore to apply the
DNKBF to describe and interpret the effects of debranning on wheat breakage and, in so
doing, to clarify the physical significance of the DNKBF parameters.
Samples of Mallacca (hard wheat) and Consort (soft wheat) were debranned for nine
different times, at three roll gaps and under S-S and D-D dispositions. The DNKBF
successfully described the normalised particle size distribution at different debranning
times. The DNKBF describes wheat breakage in terms of Type 1 and Type 2 breakage,
where Type 1 describes a relatively narrow distribution of mid-sized particles, whilst Type
2 describes a wide size range of predominantly small particles extending to very large
particles. The proportion of Type 1 breakage increased at longer debranning times, while
Type 2 breakage decreased, for both wheats under both dispositions. S-S milling tended to
produce more Type 1 breakage than D-D. A mechanism of wheat breakage is proposed to
explain the co-production of very large and small particles via Type 2 breakage, and hence
the effect of debranning. The proposed mechanism is that small particles of endosperm
arise from scraping of large flat particles of wheat bran under the differential action of the
rolls; removal of the bran reduces the production of the large bran particles and thus
reduces the opportunity for the scraping mechanism that produces the very small particles.
The composition of broken particles can be characterised considering the four major wheat
components, pericarp, aleurone, endosperm and germ. Kernels of Mallacca and Consort
wheats were manually dissected to isolate these components. FTIR spectroscopy was able
to distinguish the different components in milled fractions. However, attempts to quantify
the relative contribution of each wheat component in milled fractions (by measuring
specific peak heights and by Partial Least Squares, PLS) were compromised by technical
limitations. An alternative approach aimed to fingerprint the components using sugar
analysis by HPLC, with some success; however the technique was too complex and limited
by the detection limit of HPLC, in particular for arabinose and xylose.
Abstract
Silvia Galindez PhD Thesis August 2014 12
Instead, the botanical distributions within eight milled fractions of Mallacca and Consort
wheats milled under S-S and D-D dispositions were analyzed by PLS models developed by
Barron (2011). The concentration functions were then found by applying the DNKBF to
the particle size distributions and to the compositional distributions, the ratio of the
DNKBFs giving the concentration function. The DNKBF was able to describe the data
well for the four botanical components studied in both wheats: pericarp, aleurone,
intermediate layer and starchy endosperm. The analysis clarified the nature of the particles
produced on breakage, showing that for Mallacca wheat, the pericarp and aleurone layer
compositions mostly varied with particle size in similar ways. Intermediate layer showed
broadly similar results to those for pericarp and aleurone in the Mallacca wheat despite
being the least accurate component predicted. However, for Consort wheat, the
intermediate layer behaved differently from pericarp and aleurone, suggesting a different
breakage mechanism, perhaps associated with how the wheat hardness affects breakage of
the bran and the production of large flat bran particles. Creation of pericarp/intermediate
layer/aleurone dust during milling was notable, in particular for Mallacca wheat. The
relative uniformity of the Mallacca compositions in relation to pericarp, intermediate layer
and aleurone, which varied in consistent ways with particle size, was also notable. By
contrast, for Consort wheat, the relative proportions of these three components appear to
vary substantially in particles of different size, pointing to very different breakage origins.
It seems that in the hard wheat, the breakage patterns are dominated by the endosperm
physical properties, while for the soft wheat, the behaviour of the large bran particles
produced is dictated much more by the properties and structure of the bran layers than by
the hardness of the endosperm.
The approach presented is practical to describe, quantify and interpret the effects of
breakage on component distributions, in order to understand the fate of kernel components
during milling and hence the origins of flour quality.
Abbreviations
13
ABBREVIATIONS
AACCI American Association of Cereal Chemists International
AFM Atomic force microscopy
AG Arabinogalactan
ANOVA Analysis of Variance
Ara Arabinose
Asp Asparagine
ATR Attenuated Total Reflectance
AX Arabinoxylans
BG β-glucans
DB Dry basis
D-D Dull-to-Dull disposition
DEFRA Department of Environment, Food and Rural Affairs
DHD Ferulic acid dehydrodimers
DNKBF Double Normalised Kumaraswamy Breakage function
D-S Dull-to-Sharp disposition
ELSD Evaporative light scattering detector
FA Ferulic acid
FAO Food and Agriculture Organization of the United Nations
FE-SEM Secondary field emission scanning electron microscopy
FTIR Fourier Transform Infrared
Glc Glucose
Glu Glutamine
HI Hardness Index
HPLC High Performance Liquid Chromatography
ICP-OES Inductively Coupled Plasma-Optical Emission Spectrometry
INRA French National Institute for Agricultural Research
IR Infrared
MIR Mid-Infrared
NABIM National Association of British and Irish Millers
NKBF Normalised Kumaraswamy Breakage function
Abbreviations
14
p-CA Para-coumaric acid
PDF Probability Density Function
PC Principal Components
PCA Principal Component Analysis
PLS Partial Least Square
Pro Proline
PSD Particle Size Distribution
S-D Sharp-to-Dull disposition
SD Standard Deviation
SEM Scanning electron microscopy
S-S Sharp-to-Sharp disposition
SKCS Single Kernel Characterization System
SIMS Secondary ion mass spectrometry
SM Spectromicroscopy
STXM Scanning Transmission X-ray Microscope
t-SA Trans-sinapic acid
TEM Transmission electron microscopy
WB Wet basis
XPS X-ray photoelectron spectroscopy
Xyl Xylose
Nomenclatures
15
NOMENCLATURES
α Proportion of Type 1 breakage ()
a Collapsing parameter ()
ai, bi, ci, di Polynomial breakage function fitted parameters ()
Ai Height of peak i in the aleurone spectrum (cm)
Am,n Height of peak n in m botanical component (cm)
At Aleurone fraction (%)
Ch Height calculated for each peak (cm)
k Al-Mogahwi and Baker breakage function fitted parameter ()
D Kernel thickness (mm)
Ei Height of peak i in the endosperm spectrum (cm)
Et Endosperm fraction (%)
G Roll gap (mm)
Gi Height of peak i in the germ spectrum (cm)
Gt Germ fraction (%)
M Moisture content of sample (%)
Mi Initial moisture content before conditioning (%)
Mt Target moisture content for conditioning (%)
m1, n1 Type 1 breakage parameters from the DNKBF ()
m2, n2 Type 2 breakage parameters from the DNKBF ()
ρ Non-cumulative inlet particle size distribution to first break (m)
ρ2 Non-cumulative outlet particle size distribution from first break
(m)
P2 Cumulative outlet particle size distribution from first break (%)
Pi Height of peak i in the pericarp spectrum (cm)
Pmax Cumulative probability corresponding to the maximum particle size
(%)
Pmin Cumulative probability corresponding to the minimum particle size
(%)
Pt Pericarp fraction (%)
W Initial wheat weight before conditioning (g)
W1 Weight of empty container for Moisture content analysis (g)
Nomenclatures
16
W2 Weight of container plus sample before heating for Moisture content
analysis (g)
W3 Weight of container plus sample after heating for Moisture content
analysis (g)
xmax Maximum particle size (m)
xmin Minimum particle size (m)
Xi Mass proportions of botanical component i ()
Normalised output particle size (m)
max Maximum normalised output particle size (m)
min Minimum normalised output particle size (m)
yi Ratio between the mass of the botanical component i in particles in
the size range x and x + dx and the total mass in particles in the same
range ()
Yi Proportion (by mass) of each botanical component ()
Y* Ratio between the total mass of the botanical component i in
particles smaller than size x and the total mass of particles smaller
than size x ()
z Fully normalised output particle size ()
Declaration and Copyright Statement
17
DECLARATION
I declare that no portion of the work referred to in this thesis has been submitted in support
of an application for another degree or qualification of this or any other university or
institute of learning.
COPYRIGHT STATEMENT
1. The author of this thesis (including and appendices and/or schedules to this thesis)
owns any copyright in it (the “Copyright) and he has given The University of
Manchester the right to use such Copyright for any administrative, promotional,
educational and/or teaching purposes.
2. Copies of this thesis, either in full or in extracts, may be made only in accordance
with the regulations of the John Rylands University Library of Manchester. Details
of these regulations may be obtained from the Librarian. This page must form part
of any such copies made.
3. The ownership of any patents, designs, trademarks and any and all other intellectual
property rights except for the Copyright (the “Intellectual Property Rights”) and
any reproductions of copyright works, for example graphs and tables
(“Reproductions”), which may be described in this thesis, may not be owned by the
author and may be owned by third parties. Such Intellectual Property Rights and
Reproductions cannot and must not be made available for use without the prior
written permission of the owner(s) of the relevant Intellectual Property Rights
and/or Reproductions.
4. Further information on the conditions under which disclosure, publication and
exploitation of this thesis, the Copyright and any Intellectual Property Rights and/or
Reproductions described in it may take place is available from the Head of School
of Chemical Engineering and Analytical Sciences.
Dedication
18
DEDICATION
To my lovely fiancé and future husband Luis Gomez Palacin for all his unconditional love
and support along this nearly four years of relationship.
Acknowledgements
19
ACKNOWLEDGEMENTS
I gratefully acknowledge the National Council on Science and Technology of Mexico
(CONACYT), the Mexican Government and the Ministry of Public Education (SEP) for
financial support to undertake this project.
The Satake Corporation of Japan is gratefully acknowledged for its support in establishing
the activities of the Satake Centre for Grain Process Engineering.
I would like to express my gratitude to my research supervisor Dr Grant Campbell for his
support, guidance, friendship, cheerfulness and patience throughout these four years. For
always being available despite the massive amount of work that always had. For his
encouragement and help to overcome all the problems I was facing during my
experimental work and for helping me to improve my scientific writing skills. I know there
is still a long way to go with this, but at the end we have published one paper based on my
PhD project!, and the second one will come soon.
I would like to thank to Prof Colin Webb as my new supervisor for your guidance, help,
feedback and support during these last months as a PhD student. I’ve been enjoying our
meetings so much that I’m really going to miss them.
To Ben Perston and Kelly Palmer from Perkin Elmer for their technical assistance with the
Spotlight and Infrared systems and to Fred Warren from King’s College London and
Perkin Elmer for his kind help in the collaboration performed.
To Dr Cécile Barron and Dr Valérie Lullien-Pellerin from INRA, Montpellier, France, for
their kind help in the FTIR analysis of the milled fractions. Without your tremendous
effort, we could not have achieved the main objective of this PhD project.
To all the technicians working in The Mill, in particular to Liz Davenport, Roy, Shahla for
their kind support in the HPLC equipment and analysis, lab material and fixing equipment.
To my family, for all their love and support in these four years. I know it has been difficult,
but we did it! Thank you very much!! I love you so much!!!
To my friends and colleagues from the Satake Centre: Ruth, Nikolina, Stavrus, Amit,
Hosam. Thank you for your company and feedback. Ruth, I would like to acknowledge
you in particular because we have struggled together during these almost four years. Thank
you for all your help and collaboration in the hydrolysis and HPLC and for all the great
moments that we spent together in the office and in the lab.
To all my friends and colleagues in The Mill, for all the help, fun and great moments that
we spent together every day at lunch, parties and Conferences!
Chapter 1
20
CHAPTER 1
MODELLING OF WHEAT BREAKAGE
1.1 Wheat, an important cereal
Wheat is a unique seed closely linked with human food uses in the course of history.
Wheat has been the basic food for thousands of years of the major civilizations developed
in Europe, West Asia and North Africa and continues being the most important grain
source for humans. This importance has enabled wheat to be grown on more land than any
other crop (FAO, 2014).
Humans changed their hunter-gatherer life to settled agriculturalist life, which led to the
gradual development of agriculture and livestock equipment (Storck and Teague, 1952;
Evans, 1993). But what caused this shift in life style? Perhaps extreme forces such as
climate change, or cultural reasons, or maybe by population pressure or even the
combinations of these three events could answer this question (Evans, 1993). Many
historians have their own conclusions; for example, Childe (1934) based on historical
studies in the Near East and China, suggested that climate changes in the Pleistocene
occurred nearly at the same time as cultural changes that originated the “Neolithic
revolution”. Excavations in Oaxaca, Mexico shown that the transition to agriculture was
tremendously gradual, suggesting that agriculture has evolved gradually, irregularly and
independently in different regions (Adams, 1966). Boserup (1965) and others suggested
that maybe an increase of population in some areas led to more intensive and regular
cultivation, although there is still open the question if population pressure was responsible
of the development of agriculture.
A very important event that took place thousands years ago was crop domestication, which
earliest archaeological signs were found in Mexico, Near East and North China. Different
types of crops were domesticated in all these centres. The genetic changes that occurred in
crops perhaps from unconscious selection enabled their domestication (Evans, 1993;
Zohary et al., 2012). Wheat is a clear example of this, since only one tiny mutation in the
genetic makeup of the wild wheat einkorn allowed the cells between the stalk and the seed
to remain together, avoiding the seed to fall down after being ripened, as happens in the
Chapter 1
21
non-mutated wild wheat. With this new mutation, the collection of more grains was
possible, but even more important, its cultivation (Wrigley, 2009; Zohary et al., 2012).
Along other cereals, wheat provides important quantities of nutrients and energy to sustain
human and animal populations. However, among all the plants, wheat is the only grain
that contains gluten proteins. Their unique properties (i.e dough-forming for bread due to
its gas-retention ability) have enabled the production of a great variety of food products
(van Vliet et al., 1992; Decker et al., 2002; Dobraszczyk et al., 2000, 2003; Campbell,
2008; Hamer et al., 2009; Campbell and Martin, 2011). Nowadays, wheat has become for
humanity more than only a raw material for wheat-based products, it has become along
with other cereals a raw material for production of biofuels in order to cover the demand of
modern society to look for alternative energies to reduce the environmental impact.
The global population is now facing the increase of the demands of food supply in which
wheat plays an important role. Cereals in general and wheat in particular are required to
cover these needs and a non-food use (i.e biofuels) since the global population is
continuously growing and demands more food and energy (Dixon, 2007). On these bases,
wheat must be processed effectively in order to continue to make its central contribution to
meeting the world’s food needs whilst also leading the way towards sustainable production
of chemicals and energy based on biorefineries.
The processing of wheat begins with the separation of its botanical components. Flour
milling separates the wheat kernel into flour, hence an important component of the future
usage of wheat is the efficiency of the milling process to fulfil the different demands that
the modern society is increasingly facing.
1.2 Debranned wheat and modelling of wheat roller milling
Food products arising from cereals constitute a major part of the daily diet in the world´s
population. The wheat-food products are principally made from refined white flour from
which the outer layers of the wheat grain are removed. However, these layers (i.e pericarp,
the seed coat, the nucellar epidermis, aleurone), which are eliminated in the milling
fraction called by the millers as “bran”, contain most of the fibre, phytochemicals and
micronutrients of the wheat kernel that could contribute to increasing the nutritional quality
of human food if added in flours or used as food ingredients (Hemery et al., 2007).
Debranning process involves the removal of the peripheral layers of the cereal grain, by
abrasion and friction, using modified rice polishers. The debranned kernels are then
Chapter 1
22
recovered for being processed in successive stages. The removal of the bran before
grinding has resulted in increased flour and semolina extraction, as well as a considerable
reduction of steps in the sub-sequent milling process (Dexter and Wood, 1996). Several
advantages are associated with wheat debranning (or pearling), such as the removal of
contaminants that are present in the peripheral tissues of wheat grains (i.e mycotoxins,
pesticides residues and heavy metals) (Mousia et al., 2004; Laca et al., 2006; Bottega et al.,
2009; Posner, 2009; Delcour et al., 2012), or the removal of intrinsic components that are
detrimental to flour quality (e.g. instability of germ lipids) (Gys et al., 2004a,b; Beta et al.,
2005; Liu et al., 2008; Delcour et al., 2012; Sovrani et al., 2012; Sapirstein et al., 2013).
Flour produced from debranned wheat kernels exhibits different and better characteristics
and quality that results in a better bread in terms of organoleptic quality (soft texture, more
spongy and palatable taste) compared to flour produced from conventional milling (Mousia
et al., 2004; Singh and Singh, 2010; Delcour et al., 2012). More information about
debranning of wheat is detailed in Chapter 2. At least part of the effect of debranning on
flour quality arises through the effect of debranning on the milling process itself.
Therefore, to understand the process engineering consequences of debranning, in terms of
the interaction of debranned grains with the milling process, it is helpful to have models of
wheat kernel breakage. Such models have been developed and applied, so far, to
understand the breakage of whole wheat kernels.
As described in more detail in Chapter 3, the breakage equation that describes first break
roller milling was introduced by Campbell and co-workers using a breakage function. This
function considered and related kernel characteristics (i.e size, hardness, moisture content)
with processing parameters (i.e roll gap, roll disposition) (Campbell and Webb, 2001;
Campbell et al., 2001a; Fang and Campbell, 2003a,b; Campbell et al., 2007). The
polynomial breakage function enabled prediction of the particle size distribution obtained
by first breakage milling. Mateos-Salvador et al. (2011) introduced the Normalised
Kumaraswamy Breakage function (NKBF) in order to provide a simpler and more
meaningful breakage function. Campbell et al. (2012) formulated the Double Normalised
Kumaraswamy Breakage function (DNKBF), allowing two patterns of breakage to be
distinguished, with α indicating the proportion of Type 1 breakage. Fuh et al. (2014)
reformulated the DNKBF to account for the effect of roll gap more accurately whilst
retaining simplicity, with the function found adequate to describe breakage of a wide range
of wheats. However, in the current work, the simpler form of the DKNBF is applied to
Chapter 1
23
both sets of independent milled wheat fractions (debranned and whole wheats), for reasons
that are discussed in Chapter 4.
The breakage equation approach relates kernel properties and the mill operation to the
output particle size distributions. However, broken particles vary in composition as well as
size. In first break milling, the large particles produced are related to bran while the small
particles are associated to endosperm (Campbell, 2007). Therefore, a model of first break
milling that considers particle composition as well as size is required. However, it needs to
be remembered that the milling fraction called “bran” consist of several tissues that form
part of the whole wheat kernel, i.e the word “bran” is a practical milling term rather than a
botanical term.
The aim of the milling industry consists in separating and recovering flour relatively free
of bran. However, wheat milling generates many different fractions of imprecise origin that
can be remixed with flours in order to increase their nutritional content. To develop more
efficient milling processes, it would be useful to be able to monitor and determine the
histological composition of the milled fractions generated throughout the wheat milling.
The challenge to identify the botanical components contributing to particles of a given size
is considerable. In principle, this could be achieved by identifying specific biochemical
markers that are exclusively associated with particular botanical components in the kernel.
For example, Barron et al. (2007) developed a quantitative method based on carbohydrate
and phenolic acid content, leading to identification of potential biochemical markers to
monitor grain tissue proportions in fractions obtained from a conventional milling process.
Hemery et al. (2009) evaluated the grain tissue proportions in fractions of different
composition. Flour, bran and aleurone-rich fractions produced from milling and
debranning processes were analysed and good results were obtained although the germ
quantification did not enable the precise quantification of this tissue in milled fractions.
With these bases, Barron (2011) predicted the relative tissue proportion in wheat mill
streams by FTIR spectroscopy and Partial Least Square (PLS) analysis. PLS models were
developed to predict the proportion of the botanical tissues in the milled streams, achieving
a very good prediction for all botanical tissues.
Choomjaihan (2008) took a different approach to quantify the relative proportions of wheat
kernel components in size fractions following milling. Instead of identifying unique
markers, he aimed to demonstrate unique profiles or ‘fingerprints’ of minerals in the
Chapter 1
24
different kernel components. The mineral profile of a given milled fraction could thus, in
principle, be related to the distinctive profiles of the individual kernel components.
However, the mineral profiles were insufficiently distinct between components to allow the
proportion of each component in different size fractions to be deduced accurately.
Regression analysis consists on developing a model from a data set that would be able to
predict a desired response. There are plenty of regression methods that could be used for
quantitative analysis and Partial Least Squares regression is one of them. FTIR is a widely
used and non-destructive technique (the sample remains intact after analysis) that provides
a precise measurement without an external calibration (Esbensen et al., 2002). As
described in Chapters 3, 5 and 6, applying these techniques and appropriate analysis can
lead to accurate predictions, providing fast and easy bases for the analysis and
quantification of the large number of samples required for the extension of the breakage to
include composition.
A wide range of saccharides are present in all the wheat tissues but in different proportions
depending on the biochemical activities of each particular tissue (Barron et al., 2007; Stone
and Morell, 2009). By applying high performance liquid chromatography (HPLC), which
is a chromatographic technique that can separate a mixture of compounds (Fanali et al.,
2013) the identification and quantification of the sugar content in each wheat tissue and
milled fractions obtained can be possible. This technique is sufficiently rapid to construct
the basis for analysis and quantification of large number of samples.
Fistes and Tanovic (2006) defined a mathematical correlation in the form of a matrix
equation for predicting compositional properties of flour as well as their size distribution
following first break milling. Predicted results were compared with actual particle
compositional distribution obtained by experimental milling, and exhibited high accuracy
between them, showing the potential of the breakage matrix approach for predicting
compositional properties of flour stocks and their particle size distribution. However,
breakage matrices are discrete while continuous functions are more generally applicable
and more readily interpretable, thus yielding greater predictive power and greater
mechanistic insights regarding wheat breakage.
Knowing the botanical distribution in milled samples may help to divert certain flows in
the milling process to use them as a raw material for different processes or to enrich the
nutritional content of the final flour.
Chapter 1
25
The inclusion of composition in the breakage equation would make it more powerful in
general and it could make it particularly suited to understand, for example, the effects of
debranning.
The current work therefore aims to explore an alternative approach to that of Choomjaihan
(2008) for distinguishing grain components in milled fractions, and to find a suitable and
simpler function to describe not only the output particle size distribution from first break
milling, but also to describe compositional breakage functions. Infrared (IR) methods
alongside multivariate analysis, as well as sugar profile techniques are explored to find
specific fingerprints. Thus, the primary objective of the current work is to extend the
DNKBF during first break roller milling to include particle composition, characterized by
the fingerprints of pericarp, aleurone, endosperm and germ. This would help to indicate
and predict the distribution of these components within different size fractions obtained
during the roller milling operation. Equally, in the current work, the continuous equivalent
of the discrete compositional breakage matrices introduced by Fistes and Tanovic (2006) is
formulated and analyzed in wheat milled fractions. A second objective of the present work
is to apply the DNKBF to describe and interpret the effects of removal of bran on wheat
after first break milling and to determine the physical significance of the DNKBF
parameters.
1.3 Scope of the current work
The particles resulting from breakage during roller milling of wheat vary in size and in
composition. A breakage equation that includes composition would help to fully
understand the nature of wheat breakage and the fate of kernel components during milling.
The DNKBF is a flexible function that has given new insights into the nature of wheat
breakage in terms of the size of broken particles; the current work aims to address the issue
of particle composition through application of the DNKBF. Meanwhile, debranning is a
technology that has dramatically influenced wheat milling and flour quality in recent years.
The current work aims to understand the effect of debranning on first break milling, whilst
at the same time clarifying the physical significance of DNKBF parameters and hence the
nature of wheat kernel breakage and the origins of variations in particle composition.
Chapter 2 describes the general definition and uses of wheat, modern wheat milling and
debranning technologies. Chemical composition and location of nutrients within the wheat
Chapter 1
26
grain are reviewed.
Chapter 3 describes the development of the breakage equation and the recent introduction
of the Normalised Kumaraswamy Breakage function (NKBF), the Double NKBF and its
extended form. Biochemical markers and “fingerprints” used to identify tissues in milled
fractions are reviewed. The literature survey thus leads to the objectives of the current
work, to understand the effects of debranning on breakage within the framework of the
DNKBF, and to extend the breakage equation to include particle composition as well as
size.
Chapter 4 discusses the effect of debranning on wheat breakage; the wheat varieties used in
this work, their conditioning, debranning and milling procedures, the measurement of the
particle size distribution by sieve analysis, and the fitting of the DNKBF are described. A
mechanism of wheat breakage is proposed to explain the co-production of very large and
small particles via Type 2 breakage, and therefore the effect of debranning.
Chapter 5 discusses how botanical tissues were obtained by hand dissection of wheat
kernels, the collection of the FTIR spectra from the four major wheat components and
milled fractions followed by their multivariate analysis. Equally, it is explained how was
built up the calibration curve for sugar profiles and the HPLC analysis for both, botanical
tissues and milled fractions, as an attempt to the botanical composition of milled fractions.
Chapter 6 describes the extension of the breakage equation to include composition. The
mathematical formulation of the compositional breakage equation is derived.
Experimental data for the composition of different size fractions following milling is then
interpreted within this formulation, in order to find the form of the compositional breakage
function and to draw new insights regarding the nature of kernel breakage in terms of
particle composition.
Chapter 7 summarises the progress made in the current work and the conclusions drawn,
and points to recommendations for future research studies and industrial application.
Much of the current work has been presented at local and international conferences. Based
on the content of Chapter 4, a scientific paper has been accepted for publication in Cereal
Chemistry. The content of Chapter 6 is currently being prepared for submission as a
journal paper in Cereal Cereal Chemistry. The following list shows the Journal papers
accepted and in preparation, the conference works presented and the awards obtained.
Chapter 1
27
Journal papers
Galindez-Najera SP and Campbell GM (2014). Modelling first break milling of
debranned wheat using the Double Normalised Kumaraswamy Breakage
function. Cereal Chemistry (Accepted for publication). DoI: 10.1094/CCHEM-
02-14-0028-R.
Choomjaihan P, Galindez-Najera SP, Barron C, Lullien-Pellerin V and
Campbell GM. A compositional breakage equation for wheat milling. In
preparation for Cereal Chemistry.
Conferences
Galindez-Najera SP and Campbell GM. New insights from compositional analysis
and modelling. AACC International Annual Meeting, October 5–8, 2014 in
Providence, Rhode Island, US.
Galindez-Najera SP and Campbell GM. Milling of debranned wheat described with
the Double Normalised Kumaraswamy Breakage function (DNKBF). Particulate
Systems Analysis 2014, September 15-18, 2014 in Manchester, UK.
Galindez-Najera SP and Campbell GM. Milling of debranned wheat described with
the Double Normalised Kumaraswamy Breakage function (DNKBF).
ChemEngDayUK, April 7 - 8, 2014 in Manchester, UK.
Galindez-Najera SP and Campbell GM. Applying the Double Normalised
Kumaraswamy Breakage function (DNKBF) to describe the effect of debranned
wheat on first break milling. ChemEngDayUK, March 25 - 26, 2013 in London,
UK.
Galindez-Najera SP., Warren, F., Perston, B., Palmer, K., and Campbell GM.
Hyperspectral imaging of debranned wheat. Cereals and Europe Spring Meeting,
May 29 - 31, 2013 in Leuven, Belgium.
AACC International Annual Meeting, September 30 - October 3, 2012 in
Hollywood, Florida, US. Developing the compositional breakage equation using
FTIR spectroscopy to characterise wheat components and milled fractions.
Galindez-Najera SP and Campbell GM. Developing the compositional breakage
equation using FTIR spectroscopy to characterise wheat components and milled
fractions. Postgraduate research Conference, 2012, in Manchester.
Galindez-Najera SP and Campbell GM. Modelling first break milling of debranned
wheat using the Double Normalised Kumaraswamy Breakage function.
Postgraduate research Conference, 2011, in Manchester.
Travel and awards obtained
Travel award obtained for attending the AACC International Annual Meeting,
September 30 - October 3, 2012 in Hollywood, Florida, US.
Prizes and awards obtained
Second best Poster in the Postgraduate research Conference, 2012, in Manchester.
Chapter 2
28
CHAPTER 2
UNDERSTANDING WHEAT, MILLING AND DEBRANNING
2.1 Introduction
Wheat is an edible cereal grain that has fed human beings since ancient times, with milling
the key technology for releasing the nutrients in wheat and rendering them in a form
suitable for creating an array of staple food products. Meanwhile, wheat milling has
influenced other processing technologies that have contributed to the technology, cultural
and social evolution of Western civilisation, while debranning of wheat, adapted from rice
milling, is the latest technological development to impact on flour milling. The general
definition and uses of wheat, milling and debranning technologies are described in the
current chapter. Chemical composition and location of nutrients within the wheat grain are
reviewed. Modern wheat milling is an efficient process that economically fractionates the
wheat kernel to recover high quality flour. The use of debranning technology prior to
conventional roller milling is a key recent development. This fundamental understanding
of wheat leads into the specific focus of the current work, to enhance the process
engineering understanding of roller milling of wheat.
2.2 Wheat: a unique cereal grain
Wheat has accompanied mankind since ancient times, providing him with food and, along
with other cereals, enabling the transition from the hunter and gatherer nomad to the settled
agriculturalist (Storck and Teague, 1952; Wrigley, 2009).
In the regions of the ancient Eastern region of Egypt, Mesopotamia and the Levant it is
believed that wheat had its origins, around 7000 B.C., although the evidence is scarce
(Storck and Teague, 1952). There is indication that wild emmer, wild einkorn and durum
(related to emmer) are the oldest wheat types. Eventually, these species evolved to non-
brittle spike types, enabling their domestication (Storck and Teague, 1952; Wrigley, 2009).
Indeed, the evidence indicates that emmer and/or einkorn were first cultivated in
Mesopotamia, which is considered the birthplace of agriculture, and from there, the bread
wheats were quickly spread to Egypt, Iran (where they were somehow hybridized),
Chapter 2
29
China, Russia, India and Western Europe (Storck and Teague, 1952; Tainter, J., 1988;
Evans, L., 1993). The Durum variety was adopted many years later in comparison with the
other two types, which may be perhaps to the reduced agronomic adaptability and because
of the difficulty in milling the very hard Durum wheat grain (Wrigley, 2009).
After maize and rice, wheat is the third most cultivated cereal, with more than 700 million
tonnes harvested annually (FAOSTAT, 2014). The main world wheat producers are China,
India, USA, Russian Federation and France, which together contribute around 46% of the
total world wheat production (FAOSTAT, 2014). Most of the wheat used by UK millers is
grown in the UK, with an annual production of over 12 million tonnes in recent years,
while countries such as Canada, USA, France and Germany supply the majority of UK
imported wheat (NABIM, 2014; DEFRA, 2014). Although the UK has been recently a net
exporter of wheat, it requires import of higher protein wheats from these countries in order
to enhance the relatively poorer bread-making quality of the home-grown wheat. In recent
years, the amount of UK wheat exported has decreased as a result of the emergence of
wheat biorefineries producing fuel ethanol from lower protein wheat (for which the UK
climate is particular favourable) (Campbell, 2007).
Currently, in the UK, 31 companies are operating with 53 mills located throughout Great
Britain and Ireland (NABIM, 2014). In 2013, the flour production was around 5.1 million
tonnes, from which 49.2% was flour suitable for white bread-making, 11.6% for biscuits,
6.3% for wholemeal bread making, 2.4% for cakes, 2.9% for pre-packed household, 1.8%
for brown bread-making, 3.7% as food ingredients and the remaining 28.4% for other uses
(e.g starch manufacture) (NABIM, 2014). Wheat, as a traded commodity, is subject to
wide variations in price and quality due to different harvesting times around the world and
variations in grading systems (Carson and Edwards, 2009; Wrigley, 2009); part of the skill
of the miller is to yield a consistent quality product from a variable feedstock, while the
skill of the wheat trader and buyer is to maximise profits against a constantly changing
supply status.
Wheat is highly adaptable and has high yield potential, which contributes to making this
cereal so successful. Furthermore, wheat provides essential amino acids, vitamins,
minerals, antioxidants and dietary fibre components to the human diet. However, the
unique feature of wheat, that makes it pre-eminent among the cereals, is its ability to
produce raised bread (Belderok et al., 2000; Decker et al., 2002; Campbell, 2008).
Chapter 2
30
Wheat flour, when mixed with water, forms dough that is capable of retaining fermentation
gases produced by yeast in order to produce a highly aerated and palatable baked loaf
(Dobraszcyk et al., 2000; Campbell, 2008). No other cereal is able of doing this (with the
exception of rye, but its gas-retention ability is considerably inferior to that of wheat). This
unique ability comes from the gluten proteins of wheat flour which, on hydration, form a
strain hardening network that expands and contains inflating gas bubbles to produce bread
and other bakery products that are distinguished by an attractive aerated structure (van
Vliet et al., 1992; Hamer and van Vliet, 2000; Decker et al., 2002; Dobraszcyk et al., 2003;
Campbell, 2008; Hamer et al., 2009; Campbell and Martin, 2011).
The rheology involved in the breadmaking process is quite complex. For example, mixing,
sheeting, fermentation and baking are four critical steps in the breadmaking process.
During mixing, the viscoelastic properties of the wheat gluten protein are developed,
besides air is incorporated affecting the rheology and texture of the dough (Dobraszcyk et
al., 2000; Dobraszcyk and Morgenstern, 2003). The sheeting operation is different for each
product since its purposes depend on the desired product characteristics. In this context, for
bread dough for example, sheeting controls the bubble size distribution and shapes the
dough. The gluten network can be developed in bread dough by repeated sheeting.
Meanwhile for biscuit dough, sheeting is used for both to form the dough and to develop
the gluten network. During fermentation and baking, the gases produced by yeast are
trapped by the gluten network (Dobraszcyk et al., 2000; Dobraszcyk and Morgenstern,
2003; Campbell, 2008). The rheological properties of the bread dough affect the baking
quality. For example, if the dough fails to retain air and fermentation gases as a result of
the failure in the strain hardening properties and failure strain of cell walls (van Vliet et al.,
1992; Dobraszcyk et al., 2000; Dobraszcyk and Morgenstern, 2003), the resulting baked
bread has a very hard texture and not a particularly palatable taste.
Chapter 2
31
2.3 Classification of wheat
Although these days thousands of varieties of wheat are grown throughout the world,
currently, the two most prominent types of wheat are common wheat (Triticum aestivum
L.), used for bread-making and other bakery products, and for animal feed and industrial
(non-food) uses, and durum wheat (T. turgidum L. var. durum), a much harder wheat
suitable for pasta. Approximately 90-95% of the wheat crop is common wheat which is a
hexaploid and mainly selected by farmers for its superior properties. These species include
classes with spring and winter growth habitat (known as spring and winter wheats), hard
and soft endosperm, and red and white pericarp (seedcoat) (Dubcovsky and Dvorak, 2007;
Shewry, 2009). In fact, these three features, growth habit, kernel texture and bran colour,
are the most important criteria for classifying wheat (Gooding, 2009; Wrigley, 2009).
The uses of the varieties of common wheat are extensive. In general hard wheats have
higher protein contents and are more suited to breadmaking, while softer wheats are higher
yielding with lower protein contents that are more suited to cakes, biscuits, animal feed and
bioethanol production.
The grain of Durum is extremely hard and contains a high amount of protein, making it
suitable for pasta products (macaroni, spaghetti and other noodles) and semolina products
such as “couscous” and "bulghur", but in some localities it is also used for bread
(Pomeranz, 1988; Paulsen and Shroyer, 2004).
The oldest hulled wheat species einkorn (Triticum monococcum L.) (used for bread in
some regions), emmer (T. dicoccon Schrank) (bread and porridge uses), and spelt (T. spelta
L.) (used for bread) are now crops of minor economic importance and less frequently
cultivated (Stallknecht et al., 1996; Piergiovanni and Volpe, 2002). However, the interest
in these ancient species has been increasing again over the last years due to their high
adaptability to poor soils, attractive nutritional attributes, potential therapeutic properties,
to prepare alternative and new foods, and as a source of useful genes (Piergiovanni and
Volpe, 2002).
Plant breeders usually create new wheat lines more resistant to plagues or diseases and
equally to show yield improvements, so that commercial wheat types usually have a quite
short life cycle (NABIM, 2014). Therefore every year in the National Association of
British and Irish Millers (NABIM) web site, a list of both new and established wheat
Chapter 2
32
varieties is published. For example, currently there is one new winter variety that may be
suitable for bread-making (Chilton) and three new winter varieties that are appropriate for
biscuit and cakes (Delphi, KWS Croft and Monterey) (NABIM, 2014).
2.4 The structure of the wheat kernel
Figure 2.1 shows the structure of the wheat kernel, including its main components. Wheat
belongs to the grass family (Gramineae) that produces an edible caryopsis (Paulsen and
Shroyer, 2004). The embryo and endosperm are surrounded by seed coats and pericarp
(Dexter and Sarkar, 2004; Bechtel et al., 2009). The terms grain and kernel are commonly
used to describe the caryopses of cereals (Evers and Millar, 2002), hence both are used in
the current work.
Figure 2.1 Structure of the wheat kernel. Longitudinal (A) and transversal (B) view. Adapted and
modified from Dexter and Sarkar (2004), van der Kamp (2011) and Brouns et al., (2012).
Brush hairs
Endosperm (80-85%)
Aleurone layer (6-9%)
Nucellar tissue (Hyaline layer) Seedcoat (Testa 1%)
Tube cells Cross cells Hypodermis Epidermis
Inner pericarp
Outer pericarp
4-5%
Germ (3%)
Starch and
gluten
proteins
Alkylresorcionols
Insoluble & soluble
dietary fibre
(arabinoxylan,
β-glucan)
Proteins
Phenolic acids
Vitamin E
B vitamins
Minerals
Phytic acid
Enzymes
Lipids
Antioxidants
Vitamin E
B vitamins
Minerals
Sterols
Proteins
Insoluble
dietary
fibre
(xylan,
cellulose,
lignin)
Phenolic
acids
bound to
cell walls
Bran
Germ
Bran
Crease Endosperm
A
B
Pigment strand
Chapter 2
33
As Figure 2.1 shows, the wheat kernel has a lengthened oval form. Its length measures
from 4 to 10 mm, its width and depth from 2.5 to 4.5 mm and its average weight is around
30 to 35 mg. The wheat kernel is widest about the middle of its large axis (Evers and
Millar, 2002; Dexter and Sarkar, 2004; Shewry, 2009; Delcour and Hoseney, 2010). The
dorsal face is more rectilinear, and is traversed along its length by a broad and deep crease.
The grain resembles an oblate spheroid in shape, being incomplete as a result of the crease.
The crease is a deep furrow that runs almost the entire length of the kernel and extends the
bran layers deep into the kernel; this not only makes the separation of the bran from the
endosperm in milling difficult but also forms a perfect place for dust accumulation and
microorganism growth (Shewry, 2009; Bechtel et al., 2009; Delcour and Hoseney, 2010;
van der Kamp, 2011; Brouns et al., 2012). By contrast, the rice kernel does not exhibit a
crease, and bran and germ can be removed simply by polishing, such that rice endosperm
is eaten in a pure and intact form. Barley, oats and rye also have a crease, while maize,
sorghum and millet kernels do not have a crease. The presence of the crease in the wheat
kernel requires breaking open the kernel to separate bran from endosperm in the milling
process (Evers and Millar, 2002; Dexter and Sarkar, 2004; Shewry, 2009). The grain
shape, depth of the crease and the kernel size vary from one variety to another as well as
between kernels from the same variety. The higher end of the kernel is covered with small
hairs, called brushes. On the opposite end, on the dorsal (crease) side, is the germ area
where the embryo is covered with a folded membrane (Pomeranz, 1988; Bechtel et al.,
2009; Delcour and Hoseney, 2010; van der Kamp, 2011; Brouns et al., 2012).
Figure 2.2 shows hyperspectral images of transversal cuts of soft (Consort, A) and hard
(Mallacca, B) wheat types. On these images it can be observed some of the different
botanical constituents that make up the wheat grain, such as pericarp, aleurone and
endosperm. The spectra shown from each botanical component reflect their distinguishable
composition, enabling their identification.
Chapter 2
34
Figure 2.2 Hyperspectral imaging of transversal cuts of Consort (A) and Mallacca (B) wheats. Data
acquired with a Perkin Elmer Spotlight 400 ® system attached to a Frontier Spectrometer. Principal
Component Analysis (PCA) is carried out on the spectra collected; a false colour image is
generated from an overlay of each of the main principle components’ spatial locations. Image taken
by Silvia Galindez.
In practice, the milling industry aims to extract the nutritious components of wheat by
separating three fractions: flour (arising mainly from the endosperm), bran and germ; these
three main constituents comprise around 80-85, 12-18 and 2-3% respectively (Posner and
Benjamin, 2003; Campbell, 2007; Brouns et al., 2012). For this, roller milling is used
because it breaks the wheat kernel in a way that bran layers stay as large particles from
which the endosperm can be scraped by the differential action of the rollers (Campbell,
2007). Figure 2.3 presents the ideal relation between botanical components and the main
milled fractions obtained from roller milling (Bechtel et al., 2009). The typical “extraction
rate or yield” of white flour is 72% (Delcour and Hoseney., 2010), compared with a
theoretical maximum of up to 85%.
Waxy cuticle
Aleurone cell walls
Protein bodies
Starchy endosperm
Pericarp cell
walls
Aleurone cell walls
A
B
Chapter 2
35
Figure 2.3 Ideal relation between botanical components and milled fractions. Adapted from Bechtel
et al., (2009).
Inner pericarp
Intermediate cells
Cross cells
Tube cells (inner epidermis)
BEESWING
Pericarp (fruit coat)
Outer pericarp
Outer epidermis (pericarp)
Hypodermis
Thin-walled cells-remnants over most of grain; cell walls remain
in the crease; includes vascular
tissue in crease
Seed
Seed coat (testa) and
pigment strand
Nucellar epidermis (hyaline
layer)
Endosperm
Aleurone layer
Starchy endosperm
BRAN
( (8%)
BOTANICAL
COMPONENT
MILL FRACTION
GERM (2-3%)
WHITE FLOUR (80-85%)
BRAN
(12-18%)
Embryo
Embryonic axis Scutellum
Chapter 2
36
2.4.1 The Endosperm
The endosperm is the major part of the wheat grain and is composed of about 64-75%
starch (which is the stored form of energy of the grain and is the food source for the
germinating plant, until it germinates and is able to start synthesising its own food), 11-
16% protein, 0.5-1% β-Glucan, 2-6% Pentosans, 1.5% fats, 0.5% minerals and 1.5% of
dietary fibres (Belderok et al., 2000; Bechtel et al., 2009; Cui and Wang, 2009).
Arabinoxylans (or pentosans) are the main components of wheat endosperm cell walls
(70%), containing minor levels of β-D-Glucans (20%), β-glucomannan (7%) and cellulose
(2%) (Bechtel et al., 2009). Among the proteins present in the endosperm are found
albumins and globulins (25%); glutenins and gliadins (75%) (Belderok et al., 2000), the
latter are the proteins responsible for the gluten complex formation during dough making
and its ability to retain the gas produced by fermentation, enabling the production of
aerated products (Belderok et al., 2000; Dobraszcyk et al., 2000; Dobraszcyk and
Morgenstern, 2003; Campbell, 2008; Campbell 2008). A protein matrix (mainly gluten, a
storage protein) embeds starch granules in cells; the stronger protein-starch bond in hard
wheats “wets” the starch surface. When hard kernels are crushed during the milling
process, broken starch granules are produced. Conversely, soft wheats have a weak
protein-starch bond that breaks easily on milling and does not produce fractured starch
granules (Delcour and Hoseney, 2010). Figure 2.4 illustrates the starch and protein bodies
in Consort (soft) wheat. Starch granules are not broken because the bond between them
ruptures easily. The inner part of the kernel that yields high quality white flour is the
starchy endosperm, which is extracted and separated from the bran and germ during flour
milling (Figure 2.3). The aleurone layer embraces the endosperm, and provides amylase
enzymes during seed germination which degrade the starch to glucose units, allowing the
development of the embryo, roots, and shoots (Wrigley, 2004, Delcour and Hoseney,
2010). During conventional milling the aleurone layer adheres strongly to the outer bran
layers and is therefore removed as a part of the bran. The inclusion of the aleurone in flour
can increase milling yield and the nutritional quality of the flour, for this, millers aim as
much as possible to recover aleurone material without also causing bran to enter the flour
(Campbell, 2007).
Chapter 2
37
Figure 2.4 Scanning electron micrograph of Consort (soft) wheat. Bar is 20 µm. Image obtained
using a Philips XL30 FEGSEM microscope. Image was acquired with accelerated voltage 10.0 kV.
Image taken by Silvia Galindez.
2.4.2 The Bran
Bran is not a botanical term but is a milling term that refers to the product that consists
principally of two components of the kernel: the pericarp layer and the aleurone layer. As
noted in Figure 2.3, although the aleurone is botanically part of the endosperm, in practice
it is generally removed during milling along with the pericarp, nucellar tissue and seed coat
to form the fraction known by the millers as bran (Campbell, 2007; Bechtel et al., 2009;
Delcour and Hoseney, 2010). The ash (mineral) content of bran is 10-20 times that of the
endosperm (Posner and Hibbs, 2005), hence, the ash measurement is an indicator of high
bran content; high ash levels indicate the degree of undesirable bran contamination in flour
(Evers, 2004; Greffeuille et al., 2005).
Bran is a protective, structural component of the wheat kernel, in contrast to endosperm,
the purpose of which is nutritional. Wheat bran composition consists of 53% dietary fibres
(cellulose and arabinoxylans), 16% proteins, 16% carbohydrates, 7.2% minerals, 5% fats
and the remaining 2.8% are other components (Belderok et al., 2000; Courtin and Delcour,
2002; Maes and Delcour, 2002; Guttieri et al., 2008). The saccharides reported in wheat
bran include glucose, xylose, arabinose, sucrose, galactose and raffinose (Beaugrand et al.,
2004; Barron et al., 2007).
Chapter 2
38
Figure 2.5 shows a fluorescence micrograph of a transversal cut of Mallacca (hard) wheat,
in which the pericarp layer covers the whole grain; the next layer is the aleurone, which
surrounds the endosperm.
Figure 2.5 Fluorescence micrograph of a transversal cut of Mallacca wheat. Image obtained using
an Olympus BX51 microscope with a 10x objective and captured using a Coolsnap ES camera
(Photometrics) through MetaVue Software (Molecular Devices). Filter sets for DAPI (blue), FITC
(green) and TRITC (red). Image processed using ImageJ software (http://rsb.info.nih.gov/ij). Image
taken by Silvia Galindez.
Pericarp
The pericarp layer completely covers the entire kernel. It is composed of nearly 70-72%
non-starch polysaccharide, 20% cellulose, 6% protein, 2% ash and 0.5% fat. The outer
pericarp is made of hypodermis and epidermis cells which tend to break to form a product
known by the millers as the “beeswing”, illustrated in Figure 2.3. The inner pericarp is
composed of cross cells and tube cells (Figures 2.1 and 2.3) (Bechtel et al., 2009; Delcour
and Hoseney, 2010).
Seed coat and Nucellar tissue
The seed coat is attached on its outer side to the tube cells and on its inner side to the
nucellar epidermis. It is made of a thick outer cuticle and a thin inner cuticle. In coloured
wheats, it also has a layer containing pigments. The nucellar tissue (or hyaline) is firmly
joined to both the seed coat and aleurone, as shown in Figure 2.1 (Bechtel et al., 2009;
Delcour and Hoseney, 2010).
Aleurone cell walls Pericarp cell
walls
Starchy endosperm
Chapter 2
39
Aleurone
As noted above, the aleurone layer belongs to the endosperm from a botanical standpoint,
being tightly bound to the endosperm. It is considered the outermost layer of the starchy
endosperm, observed in Figure 2.5. The aleurone tissue is relatively rich in insoluble and
soluble dietary fibres such as arabinoxylan and β-glucan; proteins, phenolic acids, vitamin
E, B vitamins, minerals (K, P, Mg, Mn, Ca, Fe, Zn), phytic acids such as ferulic acid, p-
coumaric acid, vanillic acid, sinapic acid, syringic acid, alkylresorcionols, lignin and
lignans; and enzymes (Bechtel et al., 2009; Delcour and Hoseney, 2010; Brouns et al.,
2012).
2.4.3 The Germ
As illustrated earlier in Figure 2.1, the embryo contains the embryonic axis and the
scutellum (transport, digestive and absorbing organ), which together form the milling
fraction known as germ (Figure 2.3), representing 2-3%. The germ contains approximately
24-27% protein, 17% sugars (sucrose and raffinose mainly, although not negligible
amounts of arabinose and xylose can be found (Barron et al., 2007)), oil (16% in the
embryonic axis and 32% in the scutellum) 7% triglycerides, 5% ash and B vitamins and
vitamin E (Bechtel et al., 2009, Delcour and Hoseney, 2010). Germ, botanically, is the
baby plant that grows into a new plant, and the endosperm, provides the food source for the
germinating plant until it emerges from the ground and is able to start synthesising its own
food (Bechtel et al., 2009). This is why the wheat kernel is full of food and is therefore of
interest to human beings and to other life-forms, hence the protective layer of bran. The
germ in particular is a rich source of micronutrients such as B vitamins and vitamin E,
making it suitable in the preparation of vitamin concentrates, and oil, which can cause
wheat products to become rancid and decrease palatability; for this reason, germ (which is
eliminated with the outer layers during milling) needs to be separated. Although wheat
germ is problematic if left in flour, it is valuable in its own right and is used in many
products (Bechtel et al., 2009; Posner, 2009; Delcour and Hoseney, 2010).
Chapter 2
40
Figure 2.6 (A) shows a micrograph of pure dissected germ tissue from Mallacca wheat and
(B) the image reconstructed of scutellum from hyperspectral analysis. The characteristic
spectra of the oil and protein content present in the scutellum can be distinguished.
Figure 2.6 (A) Micrograph of pure dissected germ tissue from Mallacca (soft) wheat. Bar is 1000
µm. Image obtained with a Stereomicroscope. (B) Hyperspectral imaging of transversal cut of
scutellum from Consort wheat. Data acquired with a Perkin Elmer Spotlight 400 ® system attached
to a Frontier Spectrometer. Principal Component Analysis (PCA) is carried out on the spectra
collected; a false colour image is generated from an overlay of each of the main principle
components’ spatial locations. Image taken by Silvia Galindez.
1000 µm
Oil
Proteins
A
B
Chapter 2
41
2.5 Chemical composition of wheat
Carbohydrates (around 65-75% starch and fibre), proteins (around 7-12%), lipids (around
2-6%), water (around 12-14%), and micronutrients such as minerals (especially
magnesium) and vitamins (B and E) can be found in the wheat grain (Pomeranz, 1988;
Belderok et al. 2000; Hemery et al., 2007; Shewry, 2009; Delcour and Hoseney, 2010;
www.fao.org), making it a nutritious food of high energy value.
The chemical composition of the whole wheat grain and its various parts are shown in
Table 2.1. Most of the endosperm consists of energy reserves (mainly starch). The contents
of minerals (ash) and dietary fibre are low except in bran and germ.
Table 2.1 Chemical composition of the whole wheat grain and its various components (converted
to percentages on a dry matter basis (DB)). Adapted from Pomeranz (1988) and Belderok et al.
(2000). Values may change depending on the wheat type. These values are the most general
reported.
Whole grain
(%)
Endosperm
(%)
Bran (%) Germ (%)
Carbohydrates 68 82 16 40
Dietary fibre
(ash)
11 1.5 53 25
Proteins 16 13 16 22
Fats 2.0 1.5 5.0 7.0
Minerals (ash) 1.8 0.5 7.2 4.5
Other
components
1.2 1.5 2.8 1.5
Total 100 100 100 100
Figure 2.7 shows the average carbohydrate composition of mature whole-wheat grains.
Mature grains consist of 85% carbohydrate (around 75% is starch), found in the starchy
endosperm; around 7% low molecular mass carbohydrates (i.e glucose, sucrose) and
oligosaccharides are present in the aleurone, endosperm and germ (in the embryo
specifically); and around 12% cell wall polysaccharides (i.e cellulose, arabinoxylan,
glucan) are present in all grain tissues (Dexter and Sarkar, 2004; Cui and Wang, 2009;
Stone and Morell, 2009; Shewry, 2009).
2.1 Table 2.1 Chemical composition of the whole wheat grain and its various components (converted
to percentages on a dry matter basis (DB)). Adapted from Pomeranz (1988) and Belderok et al.
(2000). Values may change depending on the wheat type. These values are the most general
reported.
Chapter 2
42
In mature grain, the low molecular mass carbohydrates (glucose, fructose, sucrose,
raffinose, kestose, maltose and melobiose) are found in germ tissues (embryo and
scutellum), aleurone and endosperm, but are essentially absent from the pericarp-seed coat
tissues (Stone and Morell, 2009; Shewry, 2009). (As discussed later in this thesis, the
varying sugar profiles could form the basis for quantifying the kernel components in wheat
and in milled fractions, hence the interest in identifying them at this point.) Fructans
(inulin type) account for 1.3-2.5%, and are present predominantly in the embryo but also in
the endosperm. Cellulose makes up around 2% of the primary cell walls of endosperm and
aleurone; its highest concentrations (around 30%) are in the secondary cell walls of
pericarp-seed coat tissues and intermediate layers. By contrast, glucomannans are minor
components of cell walls of endosperm and aleurone. Arabinoxylans, as mentioned above,
are quantitatively the major non-cellulosic polysaccharides of both primary and secondary
cell walls, whether from endosperm, aleurone, germ tissues (scutellum, embryo) or
pericarp-seed coat (Dexter and Sarkar, 2004; Barron et al., 2007; Stone and Morell, 2009;
Shewry, 2009, Delcour and Hoseney, 2010).
Chapter 2
43
Figure 2.7 Average carbohydrate composition of whole mature wheat kernels. Adapted from Stone
and Morell (2009).
Cereals are deficient in lysine, one of the essential amino-acids for human; hence cereal
grains are commonly consumed along with other sources of protein with amino-acid
compositions that complement that of the grains. The relatively low fat content of the
cereal grains constitutes a dietary benefit for humans, in particular when cereals are
consumed with the levels of non-starch polysaccharides, acting as important sources of
fibre (Fabriani and Lintas, 1988; Wrigley, 2004). For animal nutrition, the low fat and high
fibre are less advantageous, as livestock nutritionists aim at maximising weight gain. This
is opposite to the emphasis of human nutrition which highlights the benefits of wholegrain
diets for combating obesity and its associated diseases.
0.03-0.09 %
0.09-0.15 %
0.54-1.55 %
0.05-0.18 %
%
0.19-0.68 %
0.26-0.41 %
0.94-2.50 %
71.4-75.0 %
2.00 %
0.60-1.00 %
0.27-0.38 %
5.80-6.60 %
0.60-1.00 %
<1 %
Carbohydrates
Monosaccharides
Disaccharides
Oligosaccharides
Fructans
Starch
Cell wall
polysaccharides
Arabinogalactan-
peptides
Phytic acid
Glucose Fructose
Sucrose Maltose
Raffinose
Fructose oligosaccharides
Cellulose Arabinoxylan
(1→3, 1→4)-β-D-Glucan
Glucomannan
Chapter 2
44
2.6 Uses of wheat
A wide range of edible products can be produced from wheat such as bread, pasta, puff
pastries, biscuits, crackers, pretzels, doughnuts and breakfast cereals (Bekes et al., 2004;
Campbell, 2008; Hamer et al., 2009).
From wheat straw, a substitute of wood called strawboard is produced, which is
particularly useful in kitchen cabinets, ready-to-assemble furniture, roofing and other
building materials (www.kswheat.com). Straw is also used without further processing for
bedding and feeding livestock and for composting (Paulsen and Shroyer, 2004). Although
wheat is also used as food for poultry, the production of this type of wheat has been
affected by the increase in wheat price over the last 10 years (NABIM, 2014) hence the
price of this meat increases.
In the presence of heat and mechanical action, the cross-linking in wheat gluten increases,
resulting in a high viscosity protein that is useful for extrusion processes such as the
production of dry pet food (Chantapet et al., 2013; Atchison and Gates, 2010).
The main component of wheat is starch, and its extraction enables its use in several
industrial processes. For example, starch is used as a loading agent in the manufacture of
plastics and resins; as an adhesive on postage stamps; as a surface coating agent in the
production of paper; as gelling agent or emulsifier in the manufacture of paint; to hold the
bottom of grocery sacks; as a fermentation substrate in the production of hormones and
antibiotics in the pharmaceutical industry; or in the production of biopolymers for plastic
bags, packaging materials, plastic films, eating utensils and molded items (Jackel, 1995;
Paulsen and Shroyer, 2004; Atchison and Gates, 2010; www.kswheat.com). From
processed milled flour, glucose or maltodextrin can be obtained, both are important
additives found in snack foods and sweets.
Wheat starch is also used as a replacement of fat, as a substitute of egg white, as a
replacement of milk, as a co-binder in packaging of food and non-food products and as a
“carrier” of controlled liberation of flavours (www.kswheat.com).
The wheat grain is used in the production of alcohol for the drinks’ industry, cleaners,
industrial textiles, inks, pharmaceuticals, propellants for perfumes (Atchison and Gates,
Chapter 2
45
2010), and more recently for bio-ethanol and biofuel refinery (Batchelor et al., 1993; Du et
al., 2009; Misailidis et al., 2009; Martinez-Hernandez et al., 2013).
Wheat in its natural form is of little direct use; in order to transform wheat into useful food
or non-food products, it must be processed, usually via dry milling to produce flour.
2.7 Wheat Milling
Wheat milling is an ancient industry that has been active for thousands of years and
continues be a very important activity part of the modern food industry. Milling technology
has slowly evolved, beginning with the mortar and pestle and the stone milled used by
primitive cultures, moving towards the invention of the millstone in Roman times, to the
very sophisticated roller milling machines of modern times (Storck and Teague, 1952;
Campbell, 2007).
The oldest device known and used for flour milling is the saddlestone, which is a cradle-
shaped piece, made of hard stone. Coarse flour is obtained by shearing forces between the
cylindrical stone and the saddlestone (Storck and Teague, 1952), as seen in Figure 2.8.
Figure 2.8 Saddlestone. Image obtained from http://www.cerealsdb.uk.net/.
Throughout the passing of the years, the need for more efficient tools to produce flour was
evident, such that advanced civilizations developed rotary querns, which consist of two
circular stones; one is a rotating runnerstone and overlays a static bedstone. The grain is
Chapter 2
46
introduced in the centre, in a hole of the runnerstone, as the grain grinds, this travels to the
edges, obtaining coarse flour (Storck and Teague, 1952). Figure 2.9 illustrates an example
of an ancient quern found in England.
Figure 2.9 A Quern stone from the Iron Age (about 400-300 B.C.), found in Burton Agnes, East
Yorkshire, England. Image obtained from The British Museum website
(http://www.britishmuseum.org/).
From the saddlestone to the invention of the quern and later the millstone, several devices
were developed, each one trying to make the flour production easier and more efficient,
even when the main objective was to produce enough flour to feed a family. Perhaps the
lever mill was the device that changed the path of milling. From that moment, flour milling
became a profitable operation (Storck and Teague, 1952).
Figure 2.10 shows the evolution of milling devices, from the saddlestone to the millstone.
Chapter 2
47
Figure 2.10 The evolution of the milling devices. Adapted from Storck and Teague (1952).
Saddlestone Slab Mill Push Mill
Lever Mill
Hourglass Mill
Delian Mill
Quern
Millstones
Chapter 2
48
International innovations that occurred at the end of the 19th century generated a milling
revolution, switching the millstones for roller mills (Campbell, 2007). The first setting up
of the “modern” flour milled was around 1880, and three places claim this installation with
similar validity: Manchester, UK; Minneapolis, U.S., and Budapest, Hungary. The
invention of roller mills rapidly displaced millstones, having the ability to mill hard wheats
more effectively, while offering versatility in adjustments in roll gap, roll disposition,
different speeds, roll surfaces, etc., in order to produce high quality and consistent flour
relatively free of bran (Campbell, 2007).
The modern roller milling process breaks up the kernel and scraps the endosperm from the
bran. The endosperm is gradually reduced into flour by a series of grindings; sifters and
purifiers separate intermediate products throughout the whole process (Campbell, 2007).
This dry milling process separates endosperm from bran and germ to produce a high yield
of relatively pure white flour using a dry (and hence cheap) process. The main objective of
the flour milling industry is to manufacture flour of a high and consistent quality, as
efficient as possible, always trying to meet the specifications of the buyers, and at the
lowest cost for the milling activity (Sudgen, 2001; Yuan et al., 2003).
Figure 2.11 illustrates a basic diagram showing how flour is produced. Table 2.2 describes
the inputs, outputs and objectives of the main stages in the milling process, starting with
the cleaning stage. The yield varies depending on the wheat type processed and from mill
to mill.
Chapter 2
49
Figure 2.11 Simplified diagram of a typical flour milling process. Adapted from NABIM, 2014.
Conditioning
Cleaning
Gristing
Rolls (up to 4)
Reduction
Rolls (up to 12)
Sieves
Sieves
White
flour
Wheat
germ
Wheat
feed
Bran
Break
Chapter 2
50
Table 2.2 Main systems of the Mill flow sheet. Adapted from Posner and Hibbs (2005), Posner
(2009) and NABIM (2014).
Stages Input Output Objective
Cleaning Wheat containing contaminants
such as strings, straw, parts of
bags, wood, stones or metal
Cleaned wheat, free of
impurities
Removal of undesirable
material using
differences in size,
shape, specific gravity,
air resistance, etc
Conditioning Different moisture content in
grains
Uniform grain moisture content Soften endosperm but
toughen bran preventing
its breaking up and
improving its separation
from the floury
endosperm
Gristing Wheat with different
characteristics
Specific types of wheat
blended
Blending batches of
wheat together (gristing)
produces a mix suitable
to make specific flours
Break system Clean conditioned wheat Bran, sizings, and germ to
purifiers; pure endosperm to
reduction system; some flour
A series of up to four
fluted rolls to break
open endosperm,
cleaning the bran from
endosperm.
Grading system Mixture of sizings and flour
from breaks
Sizings to purifiers and sizing
rolls, and flour
Separate flour from
sizings
Purification
system
Mixed of pure endosperm,
compound endosperm/bran,
bran and germ
Germ, large compound bran
with endosperm, small
compound bran with
endosperm, pure endosperm
Maximize pure
endosperm particles to
reduction. Divert fine
breaks material and
sizings for further
processing
Sizing system Mixed particles of
bran/endosperm from breaks
and purifiers
Pure bran and germ as by-
products; pure endosperm to
reduction, compound
endosperm/aleurone to low-
grade system; some flour
Reduce endosperm
particles to a required
size and separate
between endosperm
particles and adhering
bran
Reduction
system
Relatively clean endosperm
from the breaks, grading, and
purification systems
Flour; germ fragments; mixture
of endosperm/aleurone/bran to
low grade
Reduce pure endosperm
to flour. Endosperm
chunks pass through up
to twelve sets of smooth
grinding rolls producing
finer particles. Redirect
not-pure endosperm to
sizings and low-grade
systems.
Low-grade
system
Mixed
endosperm/aleurone/bran
particles from reduction,
sizings, and remaining of the
breaks systems
Low-grade flour, mixed
aleurone/bran/endosperm
Remove the last flour
from by-products
2.2 Table 2.2 Main systems of the Mill flow sheet. Adapted from Posner and Hibbs (2005), Posner
(2009 and NABIM (2014)
Chapter 2
51
Although milling is a complex process that is made up of several stages, it can be broadly
divided into break and reduction systems. Figure 2.12 illustrates the typical modern flour
milling process flow-sheet using the gradual reduction system. As can be observed, the
break system consists of four breakages, from which wheat enters the mill at first break,
and the particles size produced at this stage dictates the flows throughout the rest of the
milling process. The purpose of the break system is to separate endosperm from bran, with
some production of flour. Meanwhile, the reduction system aims to reduce the size of
remaining endosperm particles to be small enough for flour. Flour is produced at numerous
different points and having different properties resulting from their different paths through
the process; the properties of the final combined flour depends on the proportions and
properties of the flour of each stream (Campbell, 2007).
Figure 2.12 Flour milling process. Source: Campbell (2007).
Each roller mill unit has a pair of counter-rotating fluted rolls. The angles of the flutes are
asymmetric, thus providing a dull and a sharp edge, indicated by βD and βS, respectively in
Figure 2.13(a).
Chapter 2
52
The rolls rotate at different speed, which facilitates opening up the wheat kernel and
scraping endosperm from bran, and also reduces the power drawn by the mills. The speed
differential is set at about 2.5:1 or 2.7:1 in commercial mills for cleaning the bran in the
break system (Posner and Hibbs, 2005; Campbell, 2007). Operating the rolls under
differential allows arrangements such that the sharp and dull faces on the fast and slow
rolls meet in different ways, allowing the rolls to be operated under four different
dispositions: Sharp-to-Sharp (S-S), Dull-to-Dull (D-D), Dull-to-Sharp (D-S) and Sharp-to-
Dull (S-D). Figure 2.13(b) shows the four roll dispositions, with the double arrows
indicating the faster roll.
The different dispositions result in different output particle size distributions (PSD) from
first break. Dull-to-Dull (compressive action) reduces the cutting action on the bran and
gives a wider PSD, with more large branny particles and small endosperm particles and
fewer in the mid-size range, than Sharp-to-Sharp (shearing action), in which the bran and
endosperm break together resulting in a relatively even distribution of broken particles
(Fang and Campbell, 2003a; Posner and Hibbs, 2005). This wider distribution facilitates
separation of bran from endosperm; thus most millers operate under a Dull-to-Dull
disposition at first break.
(a)
(b)
Figure 2.13 (a) Typical flute profile of a first break milling roll, where βD and βS indicate the Dull
and Sharp angles respectively; and (b) Illustration of roll dispositions arising from the differential
and fluting of break rolls: Sharp-to-Sharp (S-S), Sharp-to-Dull (S-D), Dull-to-Sharp (D-S) and
Dull-to-Dull (D-D).The double arrow indicates the fast roll. Source: Adapted from Fang and
Campbell (2002b).
βD
βS
S-S S-D D-S D-D
Chapter 2
53
As noted above, the PSD produced from first breakage directly affects the rest of the mill,
such as the machine settings and the system arrangement, determining the yields of flour
free of bran produced (Sudgen, 2001; Fang and Campbell, 2003a). Hence, the first break is
a critical control point in the milling process (Hsieh et al., 1980; Fang and Campbell,
2002a; Campbell, 2007). The purpose of the breakage equation for first break roller milling
was to relate the input characteristics of the wheat grain to the output PSD obtained. The
development of and modifications to the breakage equation are described in detail in
Chapter 3.
Flour milling aims to separate flour from bran. To achieve this, roller mills break the wheat
in such a way that the bran usually stays as large particles, while endosperm breaks into
small particles. These small endosperm particles are easily separated from the large bran
particles using sifting. By contrast, in rice milling, bran is separated from endosperm by
just polishing the bran off the outside of the grain, as the rice kernels do not have a crease.
However, a recent development has seen the adaptation of rice milling technology to wheat
milling, in which the wheat kernel has been debranned before entering the conventional
milling process.
Chapter 2
54
2.8 Debranning of wheat – A new fractionation technology
2.8.1 Definition and benefits of debranning
Debranning or pearling consists in the removal of the outer bran layers before it enters the
break system. This technology has been introduced and rapidly adopted within flour
milling, particularly in the UK, enabling a range of benefits in the milling, baking and
pasta manufacturing industries (McGee, 1995, 1996; Dexter and Wood, 1996; Bradshaw,
2004, 2005; Campbell, 2007; Hemery et al., 2007). Debranning technology is based on the
pearling process used in the preparation of rice kernels, which has been effectively adapted
for application to wheat. The first system that applied the rice pearling technology to wheat
in the late 1990s was the Peritec process (Satake Corporation, McGee, 1995), described in
Table 2.3. However, unlike rice, the wheat kernel features the crease, a deep furrow located
throughout the length of the grain that prevents the complete removal of bran layers by
pearling (McGee, 1995; Laca et al., 2006; Posner, 2009). Thus debranning of wheat is
followed by conventional milling to open up the wheat grain and separate the floury
endosperm from the remaining bran.
Debranning of wheat prior to milling produces better or more consistent quality flour from
lower quality wheat (Dexter and Wood, 1996; Mousia et al., 2004). Debranning common
wheat could effectively reduce its enzymatic activity in sprouted wheat (Hareland, 2003;
Gys et al., 2004a, b). Improvements in semolina colour, speck count, physico-chemical
characteristics and pasta cooking performance by using debranning technology have been
reported for common wheat varieties (McGee, 2006; Posner, 2009) and durum wheat
varieties (Singh and Singh, 2010). Undesirable adhering material such as bacteria, mould,
pesticides and heavy metals are reduced through the use of debranning processes (Mousia
et al., 2004; Laca et al., 2006; Bottega et al., 2009; Posner, 2009; Delcour et al., 2012). An
improvement in wheat gluten characteristics has been attained by debranning, because bran
absorbs water, restricting the hydration and development of the gluten network, resulting in
a better bred (Belderok et al. 2000; Posner, 2009).
The research on debranning thus far has tended to focus on changes in composition,
characteristics and quality of flour (Dexter and Wood, 1996; Hemery et al., 2007;
Izydorczyk et al., 2011), the antioxidant benefits and reduction of enzymatic activities
(Gys et al., 2004a,b; Hareland, 2003; Beta et al., 2005; Liu et al., 2008; Delcour et al.,
2012; Sovrani et al., 2012; Sapirstein et al., 2013) and changes in performance in relation
Chapter 2
55
to bread quality (Sun et al., 2007). However, the characteristics of the final flour depend
on the diverse streams that flour stocks take through the milling process, thus, the effect of
debranning on flour quality is firstly a process engineering matter before it becomes a
compositional one. In order to understand the interaction of debranned wheat kernels with
the milling process, mathematical models of wheat breakage are needed. Chapter 3
describes the development of such models which have, thus far, been applied extensively
to understand the breakage of whole wheat kernels. The current work takes these models
and applies them to understand the effects of debranning on wheat breakage, as a first step
to understand the consequences of debranning for flour and bread quality. Ultimately the
effect of wheat breakage is compositional as well, and the current work also investigates
extending models of breakage to include composition.
2.8.2 Debranning processes
New developments in technology are related to wheat debranning and the peeling of the
wheat outer layers from the grains before milling (Dexter and Wood, 1996). The main two
operations carried out in the debranning process are friction (peeling) and abrasion
(pearling); however, in many debranning processes the combination of both operations is
applied (McGee, 1995; Eugster, 2006, Hemery et al., 2007). Debranning by friction
consists in rubbing the wheat kernels against each other as they pass through the machine,
removing the peripheral layers. Conversely, debranning by abrasion consists in rubbing the
wheat kernels against an abrasive stone, causing the removal of the bran tissues (Hemery et
al., 2007).
During the last 20 years, different debranning processes have been developed, presenting
various characteristics, some of them allow debranning in wheat, oats, barley and rye
(McGee, 1995; Eugster, 2006). Among the existing debranning processes are TrigoTec
process (Tkac and Timm Enterprises Ltd., Tkac, 1992), the PeriTec process (Satake
Corporation, McGee, 1995), Pearling process (ConAgra Inc., Wellman, 1992) and the
Peeling process (Buhler A.G., Eugster, 2006).
Table 2.3 shows a description of the characteristics of the four most known debranning
processes. The PeriTec process marketed by Satake Corporation and the Peeling process
marketed by Buhler A. G., are the currently most used with some updating made to satisfy
the requirements of the millers.
Chapter 2
56
Table 2.3 Comparison among some existing debranning processes. Adapted from Hemery et al.
(2007).
Process Conditionin
g of grains
Description of
the process
% of
debranning
Characteristics of the
final fractions TrigoTec
process.
Tkac &
Timm
Enterprises
Ltd. (1992)
Addition of 2%
of water by
weight, 30-60s
before the
friction step.
Spraying of the
grains with 0.25-
0.35% water by
weight before
the second
friction
operation
Debranning of wheat.
Two successive friction
operations, followed by
two successive abrasion
operations.
Brushing of the pearled
kernels after the second
abrasion operation to
loosen the germ and
remove bran powder from
the crease.
Separate collection and
storage of the different
bran layers removed
during each operation.
>12%of grain
by weight
Friction kernel to kernel: blend of
the most outer layers.
First abrasion: blend of testa,
nucellar layer, aleurone and
endosperm.
Second abrasion: blend of the
remaining testa, nucellar layer,
aleurone and endosperm.
PeriTec
process.
Satake
Corporation.
(1995).
Conditioning of
the grains up to
16–17% water
by weight,
during 6–48 h
(depending on
the grains).
Quick water
addition before
pearling to
increase
the water % by
1–2%
Debranning of wheat, rye
or barley.
First bran removal (up to
the testa) by abrasion in a
vertical pearler using
abrasive stones.
Second bran removal by
agitation resulting in grain
on grain friction
(polishing of the kernels).
Possibility to combine
successive abrasion and
friction passages.
Up to 10% of
the kernel by
Weight
Polished grains with large parts
of aleurone layer still adhering.
Mix of pericarp fragments and
other bran particles
Pearling
process.
ConAgra
Inc. (1992).
Conditioning of
the grains
after the 1st
pearling step,
during 4 h up to
14.5%
(soft wheat) or
15% (hard
wheat) water by
weight
Debranning of wheat
First bran-removal step by
pearling (abrasion).
After conditioning of
grains, second bran-
removal step with a
similar pearling machine
(abrasion).
Grain milling with
conventional roller mills
Approx. 6–
10% of the
wheat kernel
by weight
Pearled kernels, with large parts
of aleurone layer still adhering.
Mix of peripheral layers and
fragments of germ.
Half of the germ is removed after
the first pearling step
Peeling
process.
Buhler A.G.
(2006).
Grains that were
previously
cleaned, are
wetted and
conditioned.
Addition of 2%
more water
before
processing to
condition
Debranning of wheat, oat,
barley and rye.
Removing by friction of
the outermost skin of the
grain only (mostly
pericarp), to remove
wheat contaminants.
Mainly by grain-on grain
friction,+gently
interaction of the grains
with the rotor and the
screen jacket.
Approx. 4% of
the kernel
by weight
Peeled wheat, with testa and
aleurone layers still adhered.
Pericarp fragments and other bran
particles: ground and pressed into
pellets for use in
non-food applications
2.3 Table 2.3 Comparison among some existing debranning processes. Adapted from Hemery et al.
(2007).
Chapter 2
57
The current chapter has introduced the structure of the wheat kernel and the nature of roller
milling as applied to fractionate that structure, with debranning a recent introduction that
offers an additional fractionation technology that has led to improved flour quality. The
breakage equation has been alluded to, as an approach that has led to a greater
understanding of roller milling of wheat in a process engineering sense; however, the effect
of debranning on breakage has not been studied previously using this approach.
Meanwhile, there is also scope to include composition within the breakage equation. The
next chapter therefore reviews the development and evolution of the breakage equation,
along with work on tissue identification and characterisation in wheat and milled fractions,
leading to the objectives for the current work.
2.9 Summary
Wheat is an ancient and widely cultivated cereal. Although nowadays thousands of
varieties of wheat are grown throughout the world, the two types most widely used are
common wheat (Triticum aestivum L.) and durum wheat (T. turgidum L. var. durum). In
general, soft wheat is mainly used for cakes, biscuits, animal feed and industrial processing
(i.e for bioethanol), hard wheat is used for bread-making, and very hard wheat is used for
pasta products. The wheat grain is made up of four major parts, pericarp, aleurone,
endosperm and germ, which are separated by milling to recover mainly the endosperm for
use in food applications.
In the wheat grain are found carbohydrates, fibre, proteins, lipids and micronutrients. The
kernel is mainly rich in carbohydrates from which the starch is the principal one, with
cellulose, β-D-Glucans, arabinoxylans and free sugars also present. The most predominant
non-starch polysaccharides of cell walls of wheat grain are arabinoxylans (AX) that consist
predominantly of pentose (5-carbon) sugars arabinose and xylose, and therefore are often
referred to as pentosans.
Although wheat is widely used in the production of bread, biscuits, doughnuts, snacks,
pretzels, crackers, breakfast cereals, puff pastries, and pasta, it is also important in the
Chapter 2
58
production of non-food products, such as a surface coating agent, as gelling agent or
emulsifier and as a fermentation substrate.
Wheat milling is a very ancient industry that has been active for thousands of years and
continues to be an important part of the modern food industry. The modern wheat roller
milling process consists of breaking open the kernel and scraping the endosperm from the
bran; bits of endosperm are gradual reduced into flour by a series of grindings. Sifters and
purifiers are used for separation of intermediate products.
Debranning or pearling consists of the removal of the hull layers of the wheat by abrasion
and friction, using modified rice polishers, before it enters the break system, enabling the
production of more consistent and higher quality flour from lower quality wheat. Some of
the most important advantages of debranning technology are the intensive removal of
bacteria, mould, pesticides, enzymes and heavy metals.
First break milling is a critical point in the overall milling process because wheat enters the
mill at first break, and the broken particles produced dictate the flows through the rest of
the mill, affecting the rest of the milling process and thus determining the yields and
quality of flour obtained. Therefore, the next chapter describes recent developments in the
process engineering understanding of wheat milling and the development of the breakage
equation that describes the first break step of the milling process.
Chapter 3
59
CHAPTER 3
UNDERSTANDING THE NATURE OF WHEAT BREAKAGE
DURING ROLLER MILLING
3.1 Introduction
The breakage equation was formulated to provide a modelling framework for studying
wheat breakage during milling; in its more developed form, it includes the Double
Normalised Kumaraswamy Breakage function (DNKBF) to describe particle size
distributions resulting from wheat breakage. This chapter introduces the breakage equation
and the most recent formulation of the DNKBF. Tissue bio-markers and their application
in process monitoring are summarised in order to understand the structure and chemistry of
wheat components and how these behave during dry roller milling processes. Studies based
on microscopy techniques, fluorescence properties, wet chemistry and infrared methods
along with multivariate analysis are explored. The objectives of the present work are
presented: to understand the effects of debranning on breakage within the framework of the
DNKBF, and to extend the breakage equation to include particle composition as well as
size.
3.2 The breakage equation for roller milling of wheat
A critical challenge for flour millers is to ensure a consistent flour quality, in order to fulfil
the needs of their major customers, the bakers, since feedstock is continuously varying
(Campbell, 2007). Consequently, a relation between the variable characteristics of the
grains entering the mill and the particle size distribution of the milled grains is required. To
achieve this, models of particle breakage during milling have been developed, mostly
focused on relating the input particle characteristics and the operation of the mill to the
output particle size distribution. Campbell and colleagues developed the breakage equation
to describe first break roller milling of wheat using a breakage function that included grain
characteristics such as diameter and hardness, and processing parameters like roll gap
Chapter 3
60
(Campbell and Webb, 2001; Campbell et al., 2001a; Fang and Campbell, 2003a,b;
Campbell et al., 2007). The breakage equation for roller milling of wheat is defined as:
dDDDxBxP
D
D
)(),( 1
0
2 (3.1)
Equation 3.1 describes the relation between the input and output particle distributions,
where P2(x) is the cumulative outlet PSD, B(x,D) is the cumulative breakage function
which quantifies the proportion of material smaller than size (x) in the output, arising from
particles of size D, and ρ(D) is the probability density function describing the input PSD.
Originally, the lower limit of integration in the equation above was D= x, in agreement
with earlier formulations such as reviewed by Austin (Austin, 1971; Austin et al., 1982;
Austin and Rogers, 1985), which assume that inlet and outlet PSDs are measured in the
same way and that an outlet particle of size x can only have originated from an inlet
particle larger than x. However Fang and Campbell (2003a) changed the lower limit of
integration to D=0 on the basis that for roller milling of wheat, D and x are measured by
different instruments and it cannot be directly compared; D is a measure of grain width or
thickness as reported by the Single Kernel Characterisation System (SKCS)*, whilst x is
the size of the smallest square aperture through which a particle can pass during sieve
analysis. Thus, although they may both have length dimensions, it is meaningless to write
D= x, as they mean physically different and not directly comparable things. Also, the
nature of wheat breakage, in which the kernel is opened up to produce large flattened bran
particles, means that it is entirely possible to produce, for example, a 4 mm bran particle
from a kernel originally 3 mm in diameter (Campbell, 2007).
* The Single Kernel Characterization system (SKCS) is an instrument that provides
measurements of weight, diameter, hardness and moisture content by testing three hundred
wheat kernels in 3-5 mins. The system feeds individual kernels into a crushing mechanism;
the kernels are crushed between the inclined crescent and the toothed rotor. The crush
response profile is measured by a load cell and converted to a hardness index (Osborne and
Anderssen, 2003).
Chapter 3
61
The breakage function describes the particle size distribution arising from a particle
originally of size D; the breakage equation then multiplies this by the prevalence of
particles of size D as indicated by the probability density function ρ1(D). Fang and
Campbell (2003a) used the following empirical polynomial relation for the breakage
function:
3
0
2
000/, xdxcxbaDGxB
G
Dxdxcxba )( 3
1
2
111
2
3
2
2
222 )(
G
Dxdxcxba (3.2)
where ai, bi, ci and di are fitted coefficients, and G/D is the milling ratio (dimensionless), a
parameter that considers that wheat breakage depends on the ratio between the roll gap (G)
and the size of the wheat kernel (D). Typical G/D value is 0.16, considering G=0.5 mm
and D =3 mm, although this value of G/D could vary between the range 0.07-0.3
(Campbell et al., 2001a), depending on the wheat type and the configuration of the value of
G to obtain flour with specific properties. The breakage function (Equation 3.2), quadratic
in G/D and cubic in x, was adequate to describe the PSDs arising from all four dispositions,
Sharp-to-Sharp, Sharp-to-Dull, Dull-to-Sharp and Dull-to-Dull, over the range 0-2000 µm.
With this equation, the output PSD for breakage of wheat can be calculated from knowing
the wheat kernel size distribution and the gap that separates the rolls (roll gap).
Further work was undertaken to introduce more characteristics to the polynomial breakage
function, such as hardness, moisture, kernel shape and degree of debranning. Fang and
Campbell (2003b) introduced a quadratic function to allow prediction of the effect of
moisture on breakage. Campbell et al. (2007) showed that the various coefficients of
Equation 3.2 could be modelled as linear functions of hardness, allowing construction of
an extended breakage equation that allowed the effects of kernel size and hardness
distributions, as reported by the SKCS, to be predicted. Thus an unknown wheat sample,
or even a mixture of wheats of unknown origin, could be tested in the SKCS and their
subsequent breakage at any roll gap predicted directly from the SKCS hardness and
diameter distributions. However, including these additional parameters requires an
increasing number of fitted coefficients, thus becoming more complicated hence, as
Chapter 3
62
Campbell et al. (2007) quoted “the polynomial nature of the fitting can yield physically
infeasible results in certain situations”. Although this experimental function developed
from polynomial fits was satisfactorily flexible to describe the range of PSD found in
common milling operations, the inconvenience of the high number of coefficients needed
for the calculations, makes their physical meaning interpretation more difficult. Besides, an
enormous amount of experimental data is required (Mateos-Salvador et al., 2011).
Meanwhile, Al-Mogahwi and Baker (2005) suggested normalises the output particle size
(x) against the roll gap (G), with the assumption that larger roll gaps give larger output
particles. A breakage function was formulated to fit the results with only one parameter, k:
G
x
kexp
G
x
kGxB
1111, (3.3)
The approach was successfully applied to different flour streams produced from the first
break of commercial flour milling, collapsing the data points from different breaks or
reductions. Mateos-Salvador et al. (2011) used a more extensive milling data set of
Campbell et al., (2007), and it was demonstrated that the model of Al-Mogahwi and Baker
(2005) only gives good results on hard wheat samples milled under D-D disposition, and
was not sufficiently flexible to describe the full range of particle size distributions under
other dispositions. Also, the function did not consider any of the input characteristics of the
wheat such as kernel size or hardness. Thus, the approach of Al-Mogahwi and Baker was
inadequate as an input–output model for a roller mill unit operation. However, a
normalization using the ratio x/(G/D) instead of x/G was investigated by Mateos-Salvador
et al. (2011), who proposed raising the milling ratio to a power of a, the “collapsing
parameter”; an appropriate value of the collapsing parameter could pull all the
experimental data points together onto a single curve. Equation 3.4 shows the breakage
function tested by Mateos-Salvador et al. (2011):
aa
G
D
k
xexp
G
D
k
xDGxB 11/, (3.4)
Chapter 3
63
With this function, the data was more successfully collapsed than using the Al-Mogahwi
and Baker function (Mateos-Salvador et al., 2011). However, this function only yielded
acceptable results for soft wheat under S-S disposition and hard wheat under D-D
disposition.
3.2.1 The Normalised Kumaraswamy Breakage function
Mateos-Salvador et al. (2011), after investigating several probability density function
equations for one- and two-parameter distributions, such as the Exponential distribution
(one parameter distribution, Equation 3.5) and Gompertz distribution (two parameter
distribution, Equation 3.6), proposed the Kumaraswamy equation as suitably flexible for
use in describing the PSDs that arise from roller milling of wheat, The Kumaraswamy
equation (Kumaraswamy, 1980) is described as a “flexible Probability Distribution
Function (PDF) for Double Bounded processes” that uses only two shape parameters. This
PDF was originally developed to fit hydrological random variables such as daily runoff,
daily recharge and atmospheric temperatures, among others (Kumaraswamy, 1980). Being
a proper probability density function, rather than a polynomial fit, meant that it was
constrained to be more physically realistic, while having fewer fitted coefficients meant
that the interpretation of the physical meaning of those coefficients would be more
tractable.
XePP 1max Exponential distribution (3.5)
XeX eePP 1max Gompertz distribution (3.6)
The Normalised Kumaraswamy Breakage function was introduced by Mateos-Salvador et
al. (2011) as an adequately simple and flexible function to describe roller milling of wheat.
The Kumaraswamy equation in its non-cumulative form is:
11 1
nmm zznmz (3.7)
Chapter 3
64
which on integration gives the associated cumulative form:
nmzzP 11 (3.8)
where ρ(z) and P(z) are, respectively, the non-cumulative and cumulative probability
distributions of the independent variable z, which lies within the interval [0,1]; m and n are
the parameters that determine the shape of the Kumaraswamy equation, where both m and
n are positive numbers.
In order to turn this into a breakage function for use in describing wheat breakage, z needs
to be defined as a non-dimensional parameter covering the range [0, 1]. Drawing on their
earlier conclusion that data from different roll gaps could be collapsed onto a single roll
gap using a suitable collapsing parameter, and on the even earlier work showing that the
milling ratio determined breakage, Mateos-Salvador et al. (2011) proposed a parameter :
aDG
x
)/( (3.9)
such that
aDG
x
)/( min
maxmax (3.10)
and
max
z (3.11)
where, x is the size of the broken particles determined by sieve analysis, G is the roll gap,
D is the size of the wheat kernel to be milled, max is based on the maximum sieve size
and minimum roll gap, and a is the collapsing parameter, which pulls the data from
different roll gaps together onto a single curve.
The Kumaraswamy function, as it was originally developed, presumes that the distribution
is well known throughout its whole range. However, sieve analysis of broken stocks is
likely to result in some particles remaining on the largest aperture sieve, hence giving an
incomplete distribution with a maximum probability lower than 100%. Consequently, it is
possible to add a parameter Pmax, which is the cumulative probability corresponding to the
maximum particle size ( max ), and accounts for incomplete distributions (Mateos-Salvador
Chapter 3
65
et al., 2011). Thus, Equations 3.12 and 3.13 become then the Normalised Kumaraswamy
Breakage function (NKBF):
11
max 1
nmm zznmPz (3.12)
nmzPDxBzP )1(1),()( max (3.13)
Note that Equations 3.12 and 3.13 may be extended to the case were Pmin is not zero, such
that its inclusion gives more flexibility when the minimum particle size min is different to
zero, but for the purposes of the current work is not necessary the use of Pmin because this
is zero; however, if Pmin is required, then the NKBF becomes in its non-cumulative and its
cumulative forms respectively:
11
minmax 1
nmm zznmPPz (3.14)
nmzPPPDxBzP )1(1)(),()( minmaxmin (3.15)
The Normalised Kumaraswamy Breakage function (NKBF) was used by Mateos-Salvador
et al. (2011) to re-examine the experimental data of Campbell et al. (2007), giving
adequate predictions of output particle size distributions over the range 0-2000 µm.
Further investigation showed that the NKBF was able to describe the entire particle size
distribution 0-4000 µm, by splitting into two ranges, fine (0-2000 µm) and coarse (2000-
4000 µm), and fitting a separate NKBF to each. 2000 µm was selected as the cut off point
between both ranges because the particles entering into the Second break in a commercial
mill are usually larger than 2000 µm (Mateos-Salvador, 2010).
The NKBF successfully predicted the fine range PSD (0–2000 μm) produced from first
break milling of wheat at different roll gaps and dispositions. Only three parameters were
required instead of the 12 previously needed by the polynomial breakage function. If it was
Chapter 3
66
necessary, the coarse range 2000-4000 μm could be fit with a second NKBF (Mateos-
Salvador et al., 2011). The slightly reduced accuracy obtained with the NKBF compare
with the original polynomial breakage function (caused by the reduction of the number of
parameters from 12 to 3), was considered acceptable, because of the much simpler function
that in principle allows the meaning of each parameter to be better understood.
Since the introduction of the breakage equation and various breakage functions for roller
milling of wheat, it has been applied primarily to first break milling, on the basis that this
is the critical control point in the milling, where the particle size distribution produced
dictates the flows through the rest of the mill. First break is also the only point where the
input material can be characterised by measuring its size and hardness with the SKCS.
However, the breakage equation could be applied to the subsequent break and reduction
systems, in order to develop more complete models and simulations of flour milling.
Therefore, Mateos-Salvador et al., (2013) extended the NKBF to apply it to Second break
milling. As above, the PSD was split into coarse (2000-4000 µm) and fine particles (0-
2000 µm) and independent NKBF functions were applied to each range. This led to new
insights regarding the nature of Second break milling and how it differs from first break,
particularly in relation to the milling ratio, which no longer dictates breakage as it does for
first break. Instead this work concluded that coarse particles produced after Second break
are strongly dependent on the input particle size and independent of the roll gap;
conversely, fine particles are independent of input particle size but highly dependent on the
roll gap. These mathematical relationships illuminated and clarified the nature of Second
break milling, indicating physical mechanisms in which the endosperm material (small
particles) is scraped off the bran particles (large particles) that are aligned with the roll gap,
such that the scraping of these small particles depends strongly on the roll gap, while the
bran material remains largely intact, independent of roll gap, but therefore strongly
dependent on its input size.
The above work by Mateos-Salvador et al. (2011, 2013) applied the NKBF separately to
coarse and fine ranges of particles, separated at 2000 µm. Campbell et al. (2012) proposed
instead combining two NKBF functions in parallel to formulate the Double NKBF
(DNKBF), which was found to describe the entire 0-4000 µm range adequately for a wide
range of wheats, and pointed towards two populations of broken particles arising from
what was labelled as Type 1 and Type 2 breakage:
Chapter 3
67
Breakage2 Type Breakage1 Type
max2
22
11 11111),(
nmnmzzPDxBzP (3.16)
Breakage2 Type
1
22
Breakage1 Type
1
11max2
222
111 11)(11
nmmnmmzznmzznmPz (3.17)
Here α is the proportion of the breakage than can be described as Type 1 breakage, and m1
and n1 are the Kumaraswamy shape parameters that describe the shape of the PSD arising
from Type 1 breakage. The quantity (1–α) gives the proportion of Type 2 breakage, whilst
m2 and n2 describe the form of Type 2 breakage.
Figure 3.1 illustrates the cumulative and non-cumulative forms of the Double NKBF for a
typical wheat sample. Campbell et al. (2012) concluded that Type 1 breakage seems to
describe a fairly narrow distribution of mainly large particles, while Type 2 breakage
seems to describe predominantly small particles whilst extending to include the few large
particles. During roller milling of wheat, bran tends to stay as large particles and the
endosperm as small ones, suggesting two apparent breakages, bran and endosperm
breakage, respectively, but Campbell et al. (2012) warned that “it is necessary to be careful
in giving a physical meaning to a suitable mathematical description”. Subsequent work of
which the current thesis forms a part, has aimed at elucidating the physical nature and
meaning, if any, of Type 1 and Type 2 breakage.
Chapter 3
68
Figure 3.14 Non-cumulative (left) and Cumulative (right) form of the Double NKBF. Adapted
from Sharp (2010).
Campbell et al. (2012) applied the Double NKBF to describe breakage of 45 samples of
wheat covering a wide range of hardness, under S-S and D-D dispositions and at five roll
gaps. They concluded that under a S-S disposition, the collapsing parameter a appeared to
take a constant value of about 0.4, while under a D-D disposition a was constantly around
0.5. A constant value of a would provide a simpler breakage equation with fewer fitted
coefficients. It also suggests a physical relationship between output particle size and roll
gap that is more or less constant for different types of wheat, arising from the constant
geometry. The value of a of 0.5 under D-D implies a square-root relationship, while the
higher value of a for D-D compared with S-S suggests milling is more sensitive to roll gap
under D-D, in agreement with the findings of Campbell et al. (2007).
Campbell et al. (2012) demonstrated that the DNKBF parameters could be correlated with
wheat hardness (as measured by the SKCS), to allow construction of a universal breakage
function that could predict breakage of any wheat based on its hardness. The resulting
predictions were very good. However, although wheat hardness is the major factor
affecting breakage, it was of interest to know whether wheat kernel shape affected
breakage. Campbell et al. (2012) tried to relate the residual variation in breakage (that not
predicted by hardness) to kernel shape (as indicated by the ratio of mass to diameter-cubed,
as reported by the SKCS). The results showed small but statistically insignificant effects,
in part because of the limited shape variation in the wheat kernels used in the study.
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0 z
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0.0 0.2 0.4 0.6 0.8 1.0 z
Type 1 Breakage
Type 2 Breakage
DNKBF
Type 1 Breakage
Type 2 Breakage
DNKBF ρ
(z)
P (
z)
3.1 Figure 3.1 Non-cumulative (left) and Cumulative (right) form of the Double NKBF. Adapted from
Sharp (2010)
Chapter 3
69
Fuh et al. (2014) performed a similar study using a sample set obtained by cross-breeding a
spherical wheat, T. sphaerococcum, with the typical wheat Cappelle, thus having a broader
range of shapes than the samples sets used by Campbell et al. (2012). This work
confirmed the findings of Campbell et al. (2012) and gave stronger evidence that shape
affects breakage; more elongated kernels, which contain relatively more bran than more
spherical kernels of the same diameter, tended to break to give larger particles than
predicted from their hardness, which makes physical sense as bran tends to stay in larger
particles.
Fuh et al. (2014) also reformulated the DNKBF to make α a function of roll gap, to give:
22
11 115.01115.0 2121max
nmnm zGzGPzP (3.18)
The extension by Fuh et al. (2014) gives the most recent formulation of the DNKBF,
combining accuracy with simplicity, with the function having been found adequate to
describe breakage of a wide range of wheats, allowing more precision in the prediction at
the cost of a single extra parameter. However, in the current work in which the DNKBF
was applied to debranned wheats, this extra parameter introduced instability, thus, the
simpler form of the DKNBF was applied for the work described later in Chapters 4 and 6.
The breakage equation has been developed to allow accurate prediction of the PSD arising
from first break milling of wheat directly from hardness and diameter distributions as
reported by the SKCS, for different roll gaps and dispositions. Work has also shown that
the other two parameters reported by the SKCS, mass and moisture content, can also be
introduced into breakage functions to give more accurate predictions. The latest
formulation of the breakage function, the DNKBF, suggests two populations of particles
produced from first break milling of wheat. Work is ongoing to elucidate the nature of
these populations, in order to understand their physical origins in terms of breakage
mechanisms. However, the output particles from first break milling of wheat differ not
only in size but also in composition. A complete understanding of wheat milling therefore
requires knowledge of particle composition as well as size within a “compositional
breakage equation”. Particle composition reflects the botanical origin of a particle, and can
be characterised in terms of distinctive biochemical species that represent the different
components of the wheat kernel.
Chapter 3
70
3.3 Biochemical markers and “fingerprints”: Potential tools to identify
wheat features and wheat tissues in milled fractions
The basis for roller milling of wheat is that bran tends to break to produce large particles
and endosperm to produce small particles, such that these compositionally different
components of the wheat kernel can be separated based on size. The breakage equation
gives a basis for describing the size distribution of particles produced from roller milling of
wheat, but does not indicate the composition of those particles. Extending the breakage
equation to include particle composition as well as size would give a fuller and more
powerful model of wheat roller milling, but represents a substantially more challenging
problem.
The wheat grain is a complex structure that is made up of several botanical components, as
detailed in Chapter 2. The challenge to identify the botanical components contributing to
particles of a given size is considerable. In principle, this could be achieved by identifying
specific biochemical markers that are uniquely associated with particular botanical
components in the kernel. Several studies have been performed to find specific features
that led to specific biomarkers, enabling the identification and quantification of botanical
tissues and their structural changes taking place during milling and processing. The
challenge to find suitable bio-markers is that most of them are not found only in one part of
the wheat (Pomeranz, 1988; Robert et al., 2005; Barron et al., 2007; Barron and Rouau,
2008; Hemery et al., 2009; Barron, 2011).
The following sections summarise the most relevant works related to understand the
structure and chemistry of wheat components (in common and durum varieties) and how
these behave during dry roller milling processes. Some of these works are based on
microscopy techniques, others on fluorescence properties, others based on wet chemistry
and infrared methods along with multivariate analysis.
3.3.1 Microscopical methods
Several analytical microscopy techniques such as fluorescence microscopy, scanning
electron microscopy (SEM), atomic force microscopy (AFM), transmission electron
microscopy (TEM); and analytical techniques that combine both microscopy and
spectroscopy such as Infrared (IR) and Raman microspectroscopy, X-ray
Spectromicroscopy (SM), secondary ion mass spectrometry (SIMS) among others, have
Chapter 3
71
increased the knowledge of the features and characteristics of the wheat grain, including its
surface and each layer. All this knowledge has lead to the development of experimental
techniques and mathematical models, which combined, can help to understand the
behaviour of wheat during debranning and milling processes and how the botanical
components are distributed throughout the different streams. In this field, Heard et al.
(2002) developed an imaging secondary ion mass spectrometry system, which was able to
map the distribution of elements or ions, followed by their overlapping on the image of the
aleurone and sub-aleurone tissues of mature wheat grains generated by ion-induced
secondary electrons. The technique showed a high spatial resolution of O±, PO2
±, Mg
+,
Ca+, Na
+ and K
+ in the granules of phytate from the aleurone layer, while CN
± was related
to the protein content and C2 ± was associated with the starch content in the endosperm.
Karunakaran et al. (2009) determined the distribution of biopolymers and their assemblies
in different parts of pericarp, aleurone and endosperm by Scanning Transmission X-ray
Microscope (STXM), while Waduge et al. (2013) used AFM for the analysis of the surface
of developing starch wheat granules, showing that at different stages of maturity the starch
granules have different properties. Neethirajan et al. (2008) used AFM to characterize the
surface morphology of wheat starch granules of vitreous and non-vitreous durum wheat
grains, finding that in the latter (non-vitreous), the starch granules were larger in size and
in number compared to the former (vitreous), while the non-vitreous starch had more
amylopectin than amylase. Scudiero and Morris (2010) used four techniques to image and
investigate soft and hard wheat grain endosperms and their resultant roller milled flours:
Secondary field emission scanning electron microscopy (FE-SEM), AFM, Raman
spectroscopy and X-ray photoelectron spectroscopy (XPS). All of the techniques showed
different features between soft and hard varieties, for example, hard wheat presented a
more granular texture and denser endosperm, while soft wheat exhibited a granular
clustering and less dense endosperm. The resulting flours from both wheat types showed
differences in characteristics such as the concentration of carbon, oxygen and nitrogen.
As has been described in this subsection, microscopical methods such as fluorescence
microscopy, atomic force microscopy, scanning electron microscopy, transmission electron
microscopy among others have played an important role in the knowledge of the features
and characteristics of wheat, from its surface and through each tissue, enabling the
development of both experimental techniques and mathematical models.
The identification of the distribution of elements and biopolymers and their assemblies in
different wheat layers has been possible with imaging secondary ion mass spectrometry
Chapter 3
72
and STXM system. AFM has proved to be useful in the characterization of surfaces in
order to get insights into the morphology. Combining AMF with FE SEM, Raman and
XPS, can exhibit the different features of soft and hard wheat varieties. All these
techniques applied alone or in combination can help to understand the behaviour of wheat
during debranning and milling processes and give a better clue of how the botanical
components are distributed throughout the different milled streams.
3.3.2 Fluorescence techniques
Pericarp, aleurone cell walls and endosperm have autofluorescence properties due to the
presence of certain constituents such as carotenoids, ferulic acid and tryptophan
respectively. These properties along with statistical models have enabled the quantification
of these botanical tissues in wheat milled fractions produced by decortication and ball
milling (Jensen et al., 1982). This method was validated by measuring starch (for
endosperm), fibre (for pericarp) and ash (for aleurone) in the samples as indicators of these
tissues, resulting in a high correlation of the autofluorescence method with starch and fibre
determinations, and low correlation with ash content, indicating that ash is not a specific
indicator of aleurone, whereas ferulic acid showed its potential as biomarker for the same
layer. With these bases, Jensen and Martens (1982) decided to quantify the botanical
tissues in the same milling fractions analyzed by Jensen et al. (1982) using the
fluorescence method, but now by amino acid composition and multiple linear regression.
The botanical tissues identified and quantified in the milled fractions were pericarp,
aleurone, endosperm and germ. Pericarp and endosperm showed good correlation
compared with the previous autofluorescence method along with the starch and fibre
determinations, whereas aleurone showed a lower correlation with the same
autofluorescence method and equally for ash content. Germ, which was only estimated by
the amino acid content, showed a relatively high correlation with niacin (although this
vitamin was used as an indicator of aleurone because nearly 82% of niacin is located in the
aleurone layer), perhaps because of the nature of the decortication system, which separates
both aleurone and germ from endosperm. Because of the autofluorescence properties of
aleurone and pericarp, the former produced by the high content of ferulic acid in this layer
and the latter associated with the carotenoid and flavonoid pigments; these wheat tissues
have shown their potential for on-line monitoring indicators in mill-streams of flour
refinement.
Chapter 3
73
Although these methods offer non-destructive, easy and rapid ways for monitoring pericarp
and aleurone contamination in flours, sometimes they have limitations in sensitivity that
need to be considered. In order to support the results obtained from these methods, it is
necessary to combine them with other analytical techniques such as ash content, HPLC,
among others; however, the disadvantage of these techniques is the time required for the
sample preparation and their analysis (Symons and Dexter, 1991, 1992, 1993, 1994).
In this subsection the advantages of the autofluorescence properties of certain wheat tissues
such as pericarp, aleurone cell walls and endosperm has been described, showing their
potential as biomarkers for monitoring these components during the milling process. The
species responsible for the autofluorescence of aleurone and pericarp are the ferulic acid
and the carotenoids’ content, respectively, and for endosperm the presence of tryptophan.
The fluorescence methods coupled with statistical models can allow the quantification of
these botanical tissues in wheat milled fractions. However, these techniques need other
time-consuming analytical methods (wet chemistry) to support the results obtained such as
ash content, colour determination, HPLC and micro-spectrofluorimetry.
3.3.3 Wet chemistry and Infrared methods coupled with
multivariate analysis
Microscopy and fluorescence methods are simple and convenient techniques, but
ultimately compositional analysis of biological tissues comes down to wet chemistry.
Peyron et al. (2002) evaluated the separation of starchy endosperm, aleurone layer and
pericarp within milling fractions of eight durum wheat samples. The phenolic content
analysis of these tissues showed a low content in ferulic acid (FA) in the starchy
endosperm, a high content in trans-sinapic acid (t-SA) in the aleurone, and a high content
of ferulic acid dehydrodimers (DHD) in the pericarp, showing the potential of these three
biomarkers to differentiate these three major botanical tissues in milling processes, in
particular, to evaluate the efficiency of separation between endosperm and bran. Jaillas et
al. (2012) studied the histological origin of durum wheat in 18 milled fractions. The system
used was an in-house imaging system, in which multispectral images were collected and
analyzed with multivariate analysis.
Chapter 3
74
The results shown that spectral fingerprints of the botanical tissues of the wheat grain and
the milled products were similar, suggesting their potential for on-line monitoring in wheat
milling processes. Antoine et al. (2004) investigated the distribution of wheat tissues in
three different bran fractions using biochemical markers. The proportion of starchy
endosperm, aleurone cell contents, aleurone walls, intermediate layer and outer pericarp
within bran fractions was quantified using starch, phytates, p-coumaric acid (p-CA) and
dehydrodimers of ferulic acid (DHD) as markers. The method successfully determined the
botanical composition in the milling fractions, showing the potential of these biomarkers to
monitor individual tissues during milling process. Similarly, Robert et al. (2005)
investigated the cell wall polysaccharides (mainly arabinoxylan (AX), β-glucans (BG) and
arabinogalactans (AG)) from cereal grains using FTIR and statistical analysis. It was
shown that the method and model used were appropriate to distinguish AX, BG and AG in
a mixture based on their spectral characteristics, allowing the estimation of relative
proportions of polymers and the fine structure of AX in complex mixtures like the cell
wall.
With these bases, Barron et al. (2007), isolated outer pericarp, aleurone layer, endosperm,
hyaline layer, embryonic axis, scutellum, and a composite layer made up of testa + hyaline
layer + inner pericarp, from two common wheat varieties, and determined their
carbohydrate and phenolic acid contents with wet chemistry techniques. The information
was used to identify specific composition patterns in each tissue and generate possible
biochemical markers as tools for control of milling processes. It was found that all the
peripheral layers showed a high amount of cell wall polysaccharides (Arabinose [Ara] and
Xylose [Xyl] primarily), whereas the lowest amount of these sugars were in the scutellum
and embryonic axis. Conversely, scutellum exhibited a higher phenolic acid content than
embryonic axis. Hyaline layer was mainly composed of arabinoxylan (AX) and high
amounts of ferulic acid. Aleurone showed lower AX content and lower amounts of ferulic
acid than outer pericarp. Extending this work, Barron and Rouau (2008) hand-dissected
outer pericarp, aleurone, hyaline, and a composite layer made up of testa + hyaline layer +
inner pericarp, from three common and one durum wheat varieties, and analyzed both sides
of the tissues by ATR-FTIR and Raman spectroscopy. Results were compared with
biochemical analysis of the same samples, enabling the identification of specific
composition patterns and features in each layer. Multivariate analysis of FTIR spectra was
successful in identifying pericarp, aleurone, hyaline and testa, based primarly on the
polysaccharide spectral signature.
Chapter 3
75
Hemery et al. (2009) developed a quantitative method based on biochemical markers for
testing grain tissue proportions in fractions obtained from a conventional milling process.
The botanical components quantified were the outer pericarp, the intermediate layer (testa
and hyaline), the aleurone cell walls and the aleurone cell contents, the endosperm and the
germ. Dissected tissues of wheat kernels were analyzed. To identify the outer pericarp, the
ferulic acid trimer was quantified; alkylresorcinols, for the intermediate layer; p-coumaric
acid and phytic acid for aleurone cell walls and the aleurone cell contents, respectively;
starch, for endosperm; and for germ, the wheat germ agglutinin was determined. The
effectiveness of this method was tested by measuring the grain tissue proportions in
fractions of different compositions. Flour, bran and aleurone-rich fractions obtained from
milling and debranning processes were used for the analysis. The composition of the
generated products enabled the study of the distribution of every tissue within the fractions
produced from both processes, showing the practicality of this method for process
monitoring and improvement. The results showed the efficacy of the method although the
germ quantification through wheat germ agglutinin did not enable the accurate
quantification of this tissue in milled fractions. Based on this, Barron et al. (2011) applied
the same biomarkers (with the exception of starch) to quantify the peripheral wheat grain
tissues and germ proportions in milling fractions from different wheat cultivars. The results
obtained showed a wide variability of the biochemical markers quantified for each
botanical tissue among cultivars. Equally, the methodology was limited, for the
quantification of both aleurone cell walls and cell content, only to processes that do not
affect their structure much. Overall, the method was more effective for the quantification
of phytic acid and alkalyresorcinols, related to aleurone and intermediate layer content.
However, although these two works presented good results with the use of biochemical
markers to identify and quantify different botanical tissues in milled fractions, they
involved a lot of wet chemistry, which is time consuming and very expensive. Based on
these results, Barron (2011) predicted the relative tissue proportions in wheat mill streams
by FTIR spectroscopy and PLS analysis. In this study, wheat tissues (the aleurone layer,
intermediate layer, outer pericarp and starchy endosperm) were isolated as in previous
works from the same author from various common wheat cultivars. Milled streams
(debranning, conventional milling, and bran fractionation) were produced from only two
wheat types, Crousty and Tiger. The spectra of botanical tissues and milled fractions were
collected by FTIR coupled with an ATR device. The fingerprints technique developed by
Hemery et al. (2009) was used as the reference method.
Chapter 3
76
The prediction of the botanical tissue proportion within milled fractions was performed
using Partial Least Square models. Regardless of the botanical tissue tested, the predictions
obtained were good.
This subsection has explained the advantages of combining Infrared methods with
multivariate analysis to develop models that are able to monitor the fate of botanical tissues
within milled fractions. The Infrared results are supported by wet chemistry methods
which are generally used as reference methods. This means that in order to have a
successful prediction using IR and multivariate analysis, it is important to have an
independent wet chemistry method that can precisely quantify the components of interest.
Then, based on these results, a calibration curve can be constructed for the IR methods,
because now there is a reference of the amounts expected of each component of interest.
Having a successful calibration curve in the IR method is crucial for a successful
prediction. For example, considering the previous works described, to identify the outer
pericarp, intermediate layer, aleurone cell walls and the aleurone cell contents, endosperm
and germ, the following biomarkers can be used, respectively: ferulic acid trimer,
alkylresorcinols, p-coumaric acid and phytic acid, starch and wheat germ agglutinin. After
quantifying the amounts of these biomarkers in the botanical tissues, the results constitute
the reference of how much of these biomarkers are contained in each tissue. Based on these
amounts, a calibration curve of botanical tissues can be constructed and their spectra
collected. The spectra are processed before computing PLS or other multivariate analysis
in order to remove the noise. If the reference method was successful and the calibration
curve built up was reliable, then the results predicted are accurate, and further analysis of
milled streams can be based just on IR and multivariate analysis, because the bases behind
the prediction are consistent. Indeed, successful predictions have been described in this
section.
As has been described in this subsection, there are numerous studies based on
microscopical, fluorescence, wet chemistry, Raman and IR methods coupled with
multivariate analysis that have demonstrated the potential of the biochemical markers to
identify and quantify certain botanical tissues during the milling process. All have
advantages and disadvantages; for example, these methods need wet chemistry to support
the results obtained and/or as reference method for multivariate analysis, and wet
chemistry is time consuming and expensive. However, once this drawback is overcome, all
these techniques are simple and convenient, showing their potential not only to understand
Chapter 3
77
features of the wheat grain but also for monitoring the fate of botanical tissues in milled
fractions.
This section has reviewed work on the use of biochemical markers to identify features in
the wheat kernel that can enable the individual identification and quantification of the
botanical components that make up the wheat grain in milled and debranned fractions. The
biomarkers and the maths exist, but they have been brought together just to a limited extent
like in the works of Fistes and Tanovic (2006), and Choomjaihan (2008), in which aimed
to formulate mathematical models to predict compositional properties of flour as well as
their size distribution following first break milling. However, both approaches have
limitations that are explained in the following section.
3.4 Compositional breakage equations
Identifying specific biochemical markers that are uniquely associated with particular
botanical components in the kernel is challenging, but as has been described in the
previous sections, there are many works that have achieved good results finding specific
features that led to specific biomarkers, enabling the identification and quantification of
botanical tissues and their structural changes taking place during milling (Barron et al.,
2007; Hemery et al., 2009; Barron et al., 2011; Barron, 2011). Fistes and Tanovic (2006)
defined a mathematical relation in the form of matrix equation for predicting compositional
properties of broken particles as well as their size distribution following first break milling
of wheat.
Two wheat varieties showing different structural and physical characteristics were roller
milled, and for the eight different size fractions obtained, moisture, ash (indicator of bran)
and protein content (found in all the botanical tissues, but highly concentrated in the germ
and aleurone) were determined. Predicted results were compared with actual PSD obtained
by experimental milling, and exhibited high accuracy between them, showing the potential
of the breakage matrix approach for predicting compositional properties of flour stocks and
their particle size distribution. However the matrix approach, based on discretisation of the
breakage equation, is not as powerful as the continuous form for formulating functions that
can be readily turned into predictive equations.
Chapter 3
78
Choomjaihan (2008), instead of identifying unique biochemical markers, attempted to
demonstrate distinctive profiles of minerals in the different grain components. The mineral
profile of a particular milled fraction could, in principle, be related to the characteristic
profiles of the individual grain components. Choomjaihan used Inductively Coupled
Plasma-Optical Emission Spectrometry (ICP-OES) to quantify the concentrations of nine
different minerals (Al, Fe, K, Mg, Mn, Na, P, Se and Zn) in pericarp, aleurone, endosperm
and germ, isolated by hand-dissection. The starchy endosperm showed a slightly higher
concentration of Al, aleurone layer was found to have higher concentration of Mn
compared with the other botanical tissues, while Na seemed to be high in germ. However,
the mineral profiles were not sufficiently distinctive between components to enable the
proportion of each component in different size fractions to be calculated. Also, the process
of soaking in order to allow hand dissection appeared to allow movement of minerals
between tissues, such that the mineral profiles in the dissected components were no longer
directly comparable to those in size fractions following dry milling.
Meanwhile, Choomjaihan (2008) formulated the compositional form of the breakage
Equation, which allows each size fraction to be described in terms of the relative
proportions of pericarp, aleurone, endosperm and germ:
i
x
i
i i
x
iiii
dxDGxyDGx
dxDGxXDGxYXDGxP
0
2
0
2
)·/,()·/,(
)·/,(·)/,(·)/,(
(3.19)
where i represents the different botanical components (pericarp, aleurone, endosperm, and
germ), and yi(x,G/D) is “compositional breakage function” of each botanical tissue.
The full derivation of the compositional breakage equation is described in detail in Chapter
6 of the present work, where these two pieces of the puzzle, the mathematical formulation
and the experimental compositional data are brought together.
The extension of the breakage Equation to include composition could make it more
powerful in general and it could make it particularly suited to understand, for example, the
effects of debranning. The current work therefore aims to explore an alternative approach
to that of Choomjaihan (2008) for distinguishing grain components in milled fractions, and
to find a suitable and simpler function to describe not only the output particle size
distribution from first break milling, but also to describe the compositional breakage
Chapter 3
79
functions that previously established. Infrared (IR) methods alongside multivariate
analysis, as well as sugar profile techniques are explored to find specific fingerprints.
Thus, the main objective of the present work is to extend the Double Normalised
Kumaraswamy Breakage function (DNKBF) during first break roller milling to include
particle composition, by using the fingerprints of pericarp, aleurone, endosperm and germ.
This would help to indicate and predict the distribution of these components within
different size fractions produced during roller milling. It would also give greater insight
into the physical nature of the breakage mechanisms by which different types of particles
are produced. The continuous equivalent of the discrete compositional breakage matrices
introduced by Fistes and Tanovic (2006) and developed by Choomjaihan (2008) is
formulated and applied to compositional data for wheat milled fractions. A second
objective of the present work is to apply the DNKBF to describe and interpret the effects of
debranning on wheat after first break milling and to determine the physical significance of
the DNKBF parameters. Between these two objectives, it is expected to gain clearer
insights into the nature of the physical mechanisms of wheat breakage and hence the
origins of the particles produced on breakage.
Chapter 3
80
3.11 Summary
The particle size distribution resulting from first break roller milling is affected by
debranning and milling processes, affecting the quality of the final flour. To ensure a
constant flour quality, in order to satisfy the requirements of the bakers, a relationship is
required between the variable characteristics of the grains entering the mill and the particle
size and compositional distributions of the milled grains obtained.
Microscopical methods such as fluorescence microscopy, atomic force microscopy,
scanning electron microscopy, transmission electron microscopy among others;
fluorescence techniques (leveraging the autofluorescence properties of certain grain tissues
such as aleurone); infrared and Raman spectroscopy methods coupled with multivariate
analysis; wet chemistry techniques for quantifying phenolic contents, alkylresorcinols,
sugar contents among others are reported as lab techniques suitable for the identification of
specific features in wheat grains that can be used as biochemical markers, have shown their
potential to monitor certain botanical tissues during the milling process. However, few
studies have aimed to formulate mathematical models that enable to predict compositional
properties of flour as well as their size distribution following the first break milling.
The breakage equation, including the Double Normalised Kumaraswamy Breakage
function, was developed and extended in order to understand and predict wheat breakage
based only on distributions of the grain characteristics and the operating parameters of the
mill. However, particle composition (characterised by biochemical markers or by sugar or
mineral profiles) has not been successfully included yet. Therefore the main objective of
the current work is to extend the DNKBF to include particle composition using fingerprints
of pericarp, aleurone, endosperm and germ during first break roller milling. This is
expected to help in the identification and prediction of the distribution of botanical tissues
within different size fractions obtained during the first break roller milling operation.
Equally, the continuous equivalent of the discrete compositional breakage matrices
introduced by Fistes and Tanovic (2006) is formulated and analyzed in the current work.
Meanwhile, the DNKBF is also applied in the next chapter to describe and interpret the
effects of debranning on wheat after first break roller milling and to determine the physical
significance of the DNKBF parameters.
Chapter 4
81
CHAPTER 4
MODELLING FIRST BREAK MILLING OF DEBRANNED
WHEAT USING THE DOUBLE NORMALISED
KUMARASWAMY BREAKAGE FUNCTION
4.1 Introduction
The breakage equation has been introduced and refined over several years in order to
quantify the effects of kernel properties and mill operation on wheat milling. Most
recently the Double Normalised Kumaraswamy Breakage function (DNKBF) has been
formulated to describe wheat breakage sufficient simply and flexibly to allow more
mechanistic studies. Modelling of first break milling using the DNKBF has identified two
populations of broken particles, described as Type 1 and Type 2, where Type 1 describes a
relatively narrow distribution of mid-range particles, whilst Type 2 describes a broad
distribution of predominantly small particles along with very large particles. However, the
physical significance and mechanistic origins of these two populations of particles is not
clear. Whilst recognising that the DNKBF is a convenient mathematical description that
does not automatically imply physical mechanisms, nevertheless, it is of interest to study
the factors that affect the production of Type 1 and Type 2 particles, and to try to elucidate
mechanisms that make sense of observed effects. As wheat milling depends critically on
the patterns of breakage of the bran layers, the effects of removal of the bran could throw
light on the physical meaning of parameters in the DNKBF. Meanwhile, debranning is an
interesting technological development in its own right that has given rise to flour of
superior quality for bread-making and pasta making; understanding how debranning affects
the initial breakage of the wheat is relevant to understanding the origins of the enhanced
functionality of the resulting flour.
The effect of debranning on wheat breakage was investigated by debranning two
representative hard (Mallacca) and soft (Consort) wheat varieties to different extents, and
milling the debranned kernels at different roll gaps and under Sharp-to-Sharp and Dull-to-
Dull dispositions. The resulting particle size distributions were described and interpreted
Chapter 4
82
using the DNKBF. This chapter describes the wheat varieties used in this work, their
conditioning, debranning and milling, the measurement of the particle size distributions by
sieve analysis, and the fitting of the DNKBF. A mechanism of wheat breakage is then
proposed to explain the co-production of very large and small particles via Type 2
breakage, and therefore to explain the effect of debranning. Galindez and Campbell, (2014)
describe the content of the current Chapter.
4.2 Materials
Two representative UK wheat types harvested in 2006 were used (Figure 4.1), Mallacca
(hard, SKCS Hardness Index = 52.52 after conditioning) and Consort (soft, SKCS
Hardness Index = 33.27 after conditioning).
Figure 4.15 Mallacca, hard wheat (left), Consort, soft wheat (right)
4.3 Experiment procedures
4.3.1 Characterisation of wheat by SKCS
Wheat samples were characterised with the Perten Single Characterisation System (SKCS,
Model 4100, Perten Instruments AB, Sweden) shown in Figure 4.2. The SKCS was used to
measure the average hardness, weight, diameter and moisture content of both wheat types
before and after conditioning.
8 mm 8 mm
4.1 Figure 4.1 Mallacca, hard wheat (left), Consort, soft wheat (right)
Chapter 4
83
Figure 16 Single Kernel Characterization System.
4.3.2 Conditioning of wheat
The SKCS was used to determine the initial moisture content of the wheat samples, from
which the amount of water needed to increase the moisture content to 16%* wet basis
(WB) was calculated as:
(4.1)
where: x = amount of water to be added
W = weight of initial wheat before conditioning (g)
Mi = initial moisture content of wheat kernel before conditioning (%WB)
Mt = target moisture content of wheat kernel after conditioning (%WB)
* The moisture content of the wheat kernels is increased up to 16% because it softens the
endosperm but toughens the bran, preventing its breaking up and improving its separation
from the floury endosperm. This percentage of moisture content is an industrial
requirement applied in the conditioning step in the milling process.
t
it
M
MMWx
100
4.2 Figure 4.2 Single Kernel Characterization System.
Chapter 4
84
Samples were conditioned overnight as described by Mateos-Salvador et al. (2011):
approximately 2800 g of each wheat type was placed in a clean bucket and the quantity of
water calculated from Eq. 4.1 added The resulting mixture was agitated manually for
several minutes and then left overnight (for at least 16 hours) to allow the moisture to be
distributed evenly throughout the sample. The following day, approximately 20 g of the
conditioned wheat kernels were run through the SKCS in order to confirm that the
moisture content had reached the 16% moisture, and to determine the new parameters
(hardness, weight and diameter) for the conditioned wheat kernels.
Table 4.1 shows the values of SKCS weight, diameter, hardness and moisture of both
Mallacca and Consort wheats after conditioning.
Table 4.1 Average and standard deviation (SD) values of SKCS kernel weight, diameter, hardness
and moisture content of Mallacca and Consort wheats before and after conditioning.
Mallacca Consort Mallacca Consort
Before Conditioning After conditioning
Kernel
characteristics
Average SD Average SD Average SD Average SD
Weight (mg) 46.98 10.96 33.92 10.17 47.60 10.96 34.74 10.45
Diameter
(mm)
3.29 0.39 2.85 0.39 3.26 0.37 2.85 0.41
Hardness (HI) 56.46 17.56 36.42 17.05 52.52 17.54 33.27 17.05
Moisture
content
(% wb)
14.35 0.29 14.11 0.31 16.12 0.36 16.44 0.32
4.1 Table 4.1 Average and standard deviation (SD) values of SKCS kernel weight, diameter, hardness
and moisture content of Mallacca and Consort wheats before and after conditioning.
Chapter 4
85
4.3.3 Debranning of wheat kernels
Wheat samples were debranned using the TM-05C laboratory debranner (Satake
Corporation, Hiroshima, Japan), shown in Figure 4.3.
Figure 17 The Satake TM-05C (laboratory debranner). The right image shows a close view of the
abrasive roller.
The TM-05C Test Mill was originally developed to determine the total amount of white
rice recovered in the whitening process (www.satake.com.au/); however it can be also used
for debranning wheat. The debranning machine is equipped with a medium abrasive roller
No. 40, with a roller speed of 1450 rpm. At the front of the debranner, on the top, a feed
box is located; in the bottom part, a metal drawer is included, which is divided into three
sections; the bran is collected in the outer two sections, while the debranned wheat is
collected in the central one.
Conditioned wheats of Malacca and Consort were split into 100 g samples (stored in clear
sample bags labelled with the intended debranning time and the roll gap and disposition to
be used for milling). Samples of each wheat variety were debranned for nine different
debranning times (0, 5, 10, 20, 30, 35, 40, 50 and 60 s). Debranning was performed in
triplicate. The samples were debranned in a randomised order with respect to wheat
sample and debranning time, to avoid systematic errors. The samples were weighed out
after debranning in a Mettler Toledo balance B303 to an accuracy of 1 mg, in order to
calculate the quantity of removed bran and the rate of debranning.
Debranning machine
Motor
Control panel
Feeder box
Drawer
Figure 4.3 The Satake TM-05C (laboratory debranner). The right image shows a close view of the
abrasive roller.
Chapter 4
86
4.3.4 Milling of wheat
Wheat was milled using the STR-100 Test Roller Mill (Satake Corporation, Hiroshima,
Japan), shown in Figure 4.4.
Figure 18 Satake STR-100AU Test Roller Mill.
The STR-100 test roller mill is a “single pass roller mill”, which is completely variable
(roll fluting, gap, speed, feed rate). The first break rolls that contains are 250 mm in
diameter and 100 mm in length and 10.5 flutes per inched, which are similar to those
employed in typical milling operations (Campbell et al., 2001b).
Following the method described by Campbell et al. (2012), debranned wheat samples were
milled at three roll gaps, 0.3, 0.5 and 0.7 mm, under Sharp-Sharp (S-S) and Dull-Dull (D-
D) dispositions. Roll speeds were set at 600 rpm and 222 rpm using a tachometer, to give a
differential of 2.7. Roll gaps were set and verified with a feeler gauge, rotating the two
rolls to a constant position to ensure consistent setting of the roll gap. To avoid frequent
changes of roll speed (which would have introduced additional error), the milling was not
randomised with respect to roll disposition; the D-D milling was completed before
reversing the roll speeds for the S-S milling. For each run, the entire milled sample was
collected for sieve analysis.
Feed hopper and rolls
Milling rolls
Control panel
4.4 Figure 4.4 Satake STR-100AU Test Roller Mill.
Chapter 4
87
4.3.5 Sieve Analysis of milled fragments
Milled samples were sieved using the laboratory sifter (Satake PLSB Series 2000
Laboratory Sifter, Satake Corporation, Japan), shown in Figure 4.5 (left).
Figure 19 The Satake PLSB Series 2000 Laboratory Sifter.
The sifter rotates at 200 rpm with a throw of 300 mm. To assist an even distribution of the
particles across each sieve during the sifting operation, a triangular plastic sieve cleaner
was placed on each sieve (Figure 4.5, right). The samples were sieved first through a
coarse stack comprising sieves of aperture sizes 4000, 3550, 3350, 3150, 2800 and 2360
µm, for five minutes. The material collected in the bottom pan was then sieved through a
fine stack containing sieves of aperture sizes 2000, 1700, 1400, 1180, 850, 500 and 212 µm
for a further five minutes. Both stacks contained a collecting pan in the bottom. The mass
of material remaining on each sieve was weighed using a Sartorius BP 610 balance to an
accuracy of 0.005 g. In total 108 samples were produced: two wheats × nine debranning
times (including t = 0) × three roll gaps × two dispositions.
Figure 4.5 The Satake PLSB Series 2000 Laboratory Sifter.
Chapter 4
88
4.3.6 Mathematical Analysis
The Double Normalised Kumaraswamy Breakage function (DNKBF) was used to describe
the experimental particle size distribution (PSD) obtained after debranning followed by
milling the wheat samples. A spreadsheet was created in Microsoft Excel® (Microsoft
Corporation, USA), using a template based on the previous work of Sharp (2010) which
was adapted to make it more suitable to the current work. A new spreadsheet was used for
each type of wheat milled at each disposition, giving a total of four spreadsheets, with each
containing the data for the nine different debranning times and the three roll gaps.
Initially, the experimental data obtained for each wheat type, at different debranning times
and under both dispositions (S-S and D-D), were normalised using Equations 4.2 and 4.3
using an initial arbitrary value of a (collapsing parameter), resulting in a single normalised
particle size distribution.
aDG
x
)/( and
aDG
x
)/( maxmin
maxmax (4.2)
max
z (4.3)
where G is the roll gap (0.3 mm was the minimum roll gap used), D is the inlet size of the
wheat kernel (around 3 mm), xmax is the maximum sieve size used (4000 µm) and a is the
collapsing parameter. An initial value of a of 0.455 was used in order to give a starting
point that would yield a robust best fit solution. The DNKBF in its cumulative form (Eqn.
4.4) was then used to calculate the percentage of particles smaller than each sieve size.
Breakage2 Type Breakage1 Type
max2
22
11 11111),(
nmnmzzPDxBzP (4.4)
Chapter 4
89
where P(z) is the normalised particle size distribution, Pmax is the percentage of material
smaller than the largest sieve size max , α is the proportion of the breakage that can be
described as Type 1 breakage, and m1 and n1 are parameters corresponding to Type 1
breakage; the values of m1 and n1 determine the shape of the function describing that
proportion of the PSD. The quantity (1– α) gives the proportion of Type 2 breakage, while
m2 and n2 are the parameters that describe the form of Type 2 breakage. z is the normalised
experimental data. In the current work, the maximum sieve size was 4000 μm and the
minimum roll gap was 0.3 mm. For D of around 3 mm and a typically 0.5, max was
therefore around 12650 μm. The parameters α, m1, n1, m2, n2 were initially set using values
(α= 0.959, m1 =5.042, n1 = 101.104, m2 = 0.816, n2 =1.850 ) from Sharp (2010). Then, the
difference between the calculated normalised particle size distribution and the
experimental distribution was calculated and squared for each value of z. The sum of the
squared differences was then minimised by varying the values of a, α, m1, n1, m2, n2 using
Microsoft Excel ® and its optimization module, Solver, resulting in a “best fit” DNKBF.
Each parameter of the DNKBF was plotted against debranning time, to see how
debranning affected the values of these parameters, in order to gain insights into the
physical significance of the parameters in the DNKBF.
4.4 Results and discussion
4.4.1 Bran removal
Figure 4.6 shows the percentage of wheat mass removed at different debranning times,
based on 100 g batches, and the rate of wheat mass removal at different times, for Mallacca
(hard) and Consort (soft) wheats. Approximately 20% and 17% of the initial wheat mass
was removed after 60 s from Consort and Mallacca wheats, respectively, with a standard
deviation of less than 0.6% between replicate batches. These percentages represent
approximately 100% and 85% of bran removal on Consort and Mallacca wheats,
respectively. The rate of wheat mass removal showed a slight increased during the
debranning process, initially from around 0.3% s–1
up to 0.4% s–1
after 60 s approximately.
Work by Laca et al. (2006) from our laboratories and using the same Mallacca wheat gave
almost identical results to the current ones. In the only independent comparable study to
report rates of debranning, Gys et al. (2004a) used a different hard wheat variety, Meunier,
and the same debranning equipment as the current work.
Chapter 4
90
Their results, based on a larger batch size of 175 g, showed a faster reduction of around
22.4% of the total kernel weight after 36 s of debranning. Their results also showed a
slowing of the rate of wheat mass removal with time, in contrast to the increase in
debranning rate observed in the current work. It is not possible to draw firm conclusions
beyond the observation that the rate of wheat mass removal was slower in the current
work, probably because of the smaller batch size, but undoubtedly affected by the
differences in the wheat properties and in conditioning regimes. It is tempting, however, to
observe that the rates of debranning appear to be converging to around 0.4% s–1
,
suggesting that the differences in rates are dominated by differences in the bran layers,
such that the rate of removal of material becomes more similar once the abrasion action
reaches the endosperm. There is scope for much more work on the effect of kernel
properties and tempering and debranning conditions on rates of wheat mass removal; for
the moment, we simply observe that the rates were similar for the two wheats in the current
work and slower than for similar work reported elsewhere.
Laca et al. (2006) found that debranning to about 15% removal was sufficient to eliminate
the majority of the aleurone tissue, and beyond this percentage, the ash content did not
change dramatically. Other studies have shown that debranning about 5% of the kernel on
average removes most of the outer pericarp layer, exposing most of the aleurone tissue;
debranning beyond this point, to about 10-15% of the kernel, removes the majority of the
aleurone layer as indicated by the decrease in ash and phenolics content (Gys et al., 2004a;
Beta et al., 2005; Bottega et al., 2009; Singh and Singh., 2010; Sovrani et al., 2012;
Sapirstein et al., 2013).
MacMasters et al. (1971) indicate that nearly 7% of the total grain weight corresponds to
pericarp, testa and nucellar tissue, and another 9% to aleurone layer. Debranning removes
layers unevenly around the kernel, but these numbers would suggest that roughly speaking,
in the current work, pericarp was largely removed after about 30 s and aleurone after 60 s,
corresponding to 8 and 18.5 % of removal of the initial mass grain on average respectively
for both wheats.
Chapter 4
91
(a) (b)
Figure 20 (a) Percentage of bran removed at different debranning times; and (b) the rate of
debranning, for Mallacca (hard) and Consort (soft) wheats, compared with the results of Gys et al.
(2004a) and Laca et al. (2006).
Figure 4.7 illustrates the differences in appearance of non-debranned and debranned wheat
for this trial, showing on the top Malacca (hard) wheat, and at the bottom Consort (soft)
wheat before and after being debranned for 30s and 60s. It becomes evident that the
structure of the kernels is changing due to the removal of the peripheral layers; some
kernels appear to be more debranned than others due to debranning not being uniform over
the grain surface. However, after 60s of debranning, almost the entire bran was removed,
avoiding starch damage.
0
5
10
15
20
25
0 10 20 30 40 50 60 70 80
Wh
ea
t m
ass r
em
ov
ed
(%
)
Debranning time t (s)
MallaccaConsortGys et al.Laca et al.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0 5 10 15 20 25 30 35 40 45 50 55 60 65Ra
te
of
wh
ae
t m
ass r
em
ov
ed
(%
/s)
Debranning time t (s)
MallaccaConsortGys et al.Laca et al.
Figure 4.6 (a) Percentage of wheat mass removed at different debranning times; and (b) the rate of
wheat mass removed for Mallacca (hard) and Consort (soft) wheats, compared with the results of
Gys et al. (2004a) and Laca et al. (2006).
Chapter 4
92
Figure 21 Malacca wheat (top) and Consort wheat (bottom) at 0, 30 and 60 s of debranning.
4.4.2 Effect of debranning on the parameters of the DNKBF
The external layers of wheat are tougher than the starchy endosperm, on milling using
roller, these produce large particles that can be easily separated from the smaller
endosperm particles, which is, indeed, the basis of dry milling of wheat to produce
relatively pure white flour (Antoine et al., 2003; Campbell, 2007; Hemery et al., 2007).
Debranning prior to milling is likely to affect the production of large bran particles. The
effect of the removal of the outer layers on the overall wheat breakage can be understood
and quantified by applying the DNKBF.
The DNKBF was used to describe the particle size distribution obtained after milling of
debranned wheat samples as described in Section 4.3.6. The PSD data were normalised
using Equations 2 and 3, and the Solver facility in Microsoft Excel was used to minimise
the sum of the squared errors to find best fit values of a (the collapsing parameter), α (the
proportion of Type 1 breakage), m1 and n1 (the Kumaraswamy parameters describing Type
8 mm 8 mm
8 mm
8 mm
8 mm 8 mm
0s 30s 60s
Figure 4.7 Mallacca wheat (top) and Consort wheat (bottom) at 0, 30 and 60 s of debranning.
Chapter 4
93
1 breakage), and m2 and n2 (the Kumaraswamy parameters describing Type 2 breakage).
Appendix 1 contains all the DNKBF parameter values. The effects of debranning on these
parameters are considered in turn.
4.4.3 Parameter a
Parameter a (the collapsing parameter) is used to pull the data from different roll gaps
together onto a single curve that can be described by a single function. In so doing it allows
the relationship between the milling ratio and the output particle size to be quantified.
Previous work concluded that for whole wheats, the relationship between milling ratio and
output particle size is approximately a square root relationship (a = 0.5) under D-D, and
slightly less than a square root relationship (a = 0.4) under S-S, for a wide range of wheat
hardnesses (Campbell et al., 2012). The difference in a indicates that milling is more
sensitive to roll gap under D-D than S-S. Fuh et al. (2014) reported work on the effect of
wheat kernel shape on milling under just a D-D disposition and, based on the finding of
Campbell et al. (2012), set a at a constant value of 0.5. In the current work, there was no
a priori basis for assuming that a should have a constant value for wheats debranned to
different levels; a was therefore retained as a variable parameter within the fitting function.
Figure 4.8 shows the best fit values for a at different debranning times for Mallacca and
Consort wheat milled under S-S and D-D dispositions. For whole wheats, the values of a
are in the range 0.35-0.46, similar to values reported previously. For both wheats the value
of a was consistently higher under S-S than under D-D at all debranning times, except for
Consort at time = 0; the average values of a for Mallacca were 0.491 and 0.444, and for
Consort were 0.470 and 0.437, under S-S and D-D, respectively. This is in contrast to the
conclusion of Campbell et al. (2012) that for whole wheats the value of a tended to be
higher under D-D than S-S. The differences in the current work are not great and may be
specific to these wheat types (the data from Sharp (2010), on which Campbell et al. (2012)
is based, shows that for some wheats the value of a was greater under S-S than under D-
D), or may reflect a genuine difference in a under S-S and D-D when wheats are
debranned.
Meanwhile, although there is a suggestion in Figure 4.8 that a increased slightly with
debranning time, analysis of variance (ANOVA) showed that variations between
debranning times were not significant at the 1% level. In general it can be concluded that a
was similar for all the samples.
Chapter 4
94
Figure 22 Values of collapsing parameter (a) at different debranning times, under S-S and D-D
dispositions for Mallacca (hard) and Consort (soft) wheats.
4.4.4 Parameter α
The parameter α gives the proportion of the PSD that can be described as Type 1 breakage.
Figure 4.9(a) shows the variation of α with debranning time, for Mallacca and Consort
wheats under S-S and D-D dispositions. For both wheats α increased with debranning time,
indicating a greater proportion of Type 1 breakage as bran is removed. Type 2 breakage
(which is associated with small particles, resulting mainly from the endosperm, and with
very large particles) is dominant at low debranning times for Mallacca (hard) wheat under
both dispositions, but after 20 s debranning Type 1 becomes predominant. For Consort
(soft) wheat, Type 2 breakage is more dominant (lower values of α) and remains dominant
for longer. Thus, at 20 s of debranning (at which time ~5% of the initial wheat mass has
been removed), Type 2 breakage accounts for approximately 80% of the PSD under S-S
and almost 90% under D-D. However, under D-D there is a sudden increase in α between
20 and 40 s, such that Type 1 breakage suddenly becomes dominant for the Consort wheat.
Under S-S the increase is more gradual; nevertheless, after 60 s of debranning, the
proportion of Type 1 breakage is greater than 0.5 for both wheats under both dispositions.
As was explained in Chapter 3, Type 2 breakage describes a broad range of particles
covering both the very small and the very large particles, while Type 1 describes mid-
range particles. It is therefore perhaps not surprising that removal of bran, which tends to
stay as large particles on breakage, reduces the proportion of Type 2 breakage.
What is not yet clear is why this should also reduce the production of the very small
particles that are also captured by the Type 2 Kumaraswamy function. Figure 4.9(b) shows
0
0.1
0.2
0.3
0.4
0.5
0.6
0 10 20 30 40 50 60a
Debranning time, t (s)
Mallacca S-S
Mallacca D-D
Consort S-S
Consort D-D
Figure 4.8 Values of collapsing parameter (a) at different debranning times, under S-S and D-D
dispositions for Mallacca (hard) and Consort (soft) wheats.
Chapter 4
95
an example of how Type 1 breakage increases at different debranning times, the sample
illustrated is Consort (soft) wheat milled under D-D disposition. As explained later, Figure
4.12 illustrates the overall effects of wheat hardness, milling disposition and debranning on
the shape of the Type 1 and Type 2 curves that combine to give the overall particles size
distribution. Appendix 2 reports all the plots related to the PSD described by the DNKBF
with their corresponding Type 1 and Type 2 breakages, for both wheats, debranned at nine
different times and milled under both dispositions (the complement of Figure 4.12).
Equally, Appendix 3 contains all the Type 1 and Type 2 breakages profile for each wheat
type milled under both dispositions (complement of Figure 4.9(b)).
(a)
(b)
Figure 23 Param aa function of debranning time, under S-S and D-D dispositions for Mallacca
(hard) and Consort (soft) wheats.
4.4.5 Parameters m1 and n1
Parameters m1 and n1 describe (using a Kumaraswamy function) the shape of the part of
the particle size distribution curve that corresponds to Type 1 breakage. To help in
interpreting the trends, it is helpful to recall that values of m and n >1 give a peaked curve,
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0 10 20 30 40 50 60
α
Debranning time, t (s)
Mallacca S-S
Mallacca D-D
Consort S-S
Consort D-D
0.0
0.5
1.0
1.5
2.0
2.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage, Consort D-D0 s5 s10 s20 s30 s35 s40 s50 s60 s
Figure 4.9 (a) Parameter α as a function of debranning time, under S-S and D-D dispositions for
Mallacca (hard) and Consort (soft) wheats; and (b) profile of Type 1 breakage increase at different
debranning times, showing Consort milled under D-D disposition as an example.
Chapter 4
96
and an increase in m tends to move the peak to the right, indicating larger particles, while
an increase in n tends to move it to the left (Campbell et al., 2012).
Figure 4.10 shows the variation of m1 and n1 with debranning time for both wheats under
both dispositions, revealing some interesting trends and contrasts. The values of m1 and n1
describe a relatively sharp peak of mid-range particles (as illustrated in Figure 4.12 to be
discussed later). For Mallacca wheat under S-S, the value of m1 remains consistently at
around 4 for all debranning times, while n1 increases. Together these imply that the peak
moves to the left, towards smaller particles, as is evident in Figure 4.12(a). By contrast
under D-D the value of m1 decreases from 4 initially down to around 3 after 60 s
debranning, while n1 stays relatively constant; this also implies a move towards smaller
particles. This is again evident in Figure 4.12(a), but with differences in the details of the
shape of the curve, which is narrower under S-S. Consort under S-S similarly gives a
decrease in m1 over a wider range, from around 5 for whole wheat down to around 2 after
60 s debranning. Under D-D, however, the change in m1 for Consort is quite dramatic,
from around 6 initially down to around 1.6 by 35 seconds, where the value appears to
plateau for longer debranning times. Under S-S disposition for Mallacca wheat, n1
increases from an initial value of 177 up to 300; conversely, under D-D disposition n1 is
somewhat erratic but with no evident trend. For Consort under both dispositions, n1
decreases from 77 to 24 and from 60 to 10, under S-S and D-D, respectively. Again these
changes have the effect of moving the Type 1 particles to the left, this time quite
dramatically, as illustrated in Figure 12(b). This indicates that the soft wheat is particularly
sensitive to debranning, and more so under the crushing action of D-D milling, and that
debranning causes a dramatic decrease in the size of Type 1 particles.
Taken together, the results show that in general the effect of debranning is to increase the
proportion of Type 1 particles and to make those particles smaller. Type 1 particles are
initially larger for the soft wheat and for D-D milling, but move to much smaller particles
under the action of debranning, while increasing in proportion. For the hard wheat the
effect of debranning on the size of Type 1 particles is less, and less so under S-S milling;
the effect of debranning is for the hard wheat more on the proportion of Type 1 particles
than on their size.
Chapter 4
97
Figure 24.10 Parameters m1 (left) and n1 (right) as a function of debranning time, under Sharp-to-
Sharp and Dull-to-Dull dispositions for Mallacca (hard) and Consort (soft) wheats.
4.4.6 Parameters m2 and n2
Parameters m2 and n2 correspond to Type 2 breakage. Figure 4.11 shows the variation of
m2 and n2 with debranning time for the two wheats under both dispositions. It is noticeable
that m2 has a value around or just below 1, compared with m1 which was greater than 1.
For values of m2<1, the shape of the curve is monotonically decaying, while m2>1
introduces a small peak into the curve (Campbell et al., 2012). Meanwhile, n2 has values of
ranging from 1.6-6.6, much smaller than n1 which typically has values of around 200 under
S-S and up to around 150 under D-D (Campbell et al., 2012). Thus, while Type 1 breakage
describes a relatively narrow peak of mid-range particles, the Type 2 curve has a very
different shape, describing primarily small particles but extending over a broad range to
include the very large particles. ANOVA shows no effect of time on m2, but a significant
effect of time on n2. In both cases the effect of treatments was significant. The larger values
of both m and n tend to cancel each other to some extent, such that the shape of the Type 2
curve is not dramatically different for the two wheats and two dispositions; the greater
effect is on the proportion of Type 2 breakage. The effect of debranning time on m2 is
relatively minor, while debranning causes a small increase in n2 for both wheats and both
dispositions. This implies a shift towards smaller Type 2 particles as a result of debranning,
as illustrated in Figure 4.12. This makes physical sense; the Type 2 curve extends to
describe the very large particles that originate from bran, so removal of the bran eliminates
those large particles and shifts the Type 2 curve towards the very small particles.
0
1
2
3
4
5
6
7
0 10 20 30 40 50 60
m1
Debranning time, t (s)
Mallacca S-S
Mallacca D-D
Consort S-S
Consort D-D
0
50
100
150
200
250
300
350
400
0 10 20 30 40 50 60
n1
Debranning time, t (s)
Mallacca S-S
Mallacca D-D
Consort S-S
Consort D-D
Figure 4.10 Parameters m1 (left) and n1 (right) as a function of debranning time), under Sharp-to-
Sharp and Dull-to-Dull dispositions for Mallacca (hard) and Consort (soft) wheats).
Chapter 4
98
Figure 25 Parameters m2 (left) and n2 (right) as a function of debranning time, under Sharp-
to-Sharp and Dull-to-Dull dispositions for Mallacca (hard) and Consort (soft) wheats.
4.4.7 Overall effect of debranned wheat on the DNKBF
Figure 4.12 illustrates the overall effects of wheat hardness, milling disposition and
debranning on the shape of the Type 1 and Type 2 curves that combine to give the overall
particles size distribution. The effects of wheat hardness and roll disposition are consistent
with previous findings for whole wheat; Type 2 breakage tends to dominate overall, but the
proportion of Type 1 breakage increases with increasing wheat hardness and is greater
under S-S milling. The effect of debranning is to increase the proportion of Type 1
breakage and to move both Type 1 and Type 2 curves towards smaller particles. The
removal of bran eliminates the largest particles, hence giving smaller particles on average.
Debranning also appears to alter the nature of the breakage of the remaining kernel, to
favour the production of mid-range Type 1 particles. This change is more dramatic for the
soft wheat (under both dispositions); from the lower graphs in Figure 4.12, a very small
percentage of (relatively large) Type 1 particles from whole wheat transforms into a
dominant peak after 60 s debranning.
Two principal breakage mechanisms operate during first break milling of wheat: crushing
and shearing; bran and endosperm respond different in each one (Fang and Campbell,
2002a,b; Campbell et al., 2012). Based on the established understanding of the nature of
breakage for soft and hard wheats, combined with the new data from the current work, a
picture is emerging in which broken wheat particles are initially created via shattering of
the wheat kernel, and are then sheared to scrape small endosperm particles from large bran
particles.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0 10 20 30 40 50 60
m2
Debranning time, t (s)
Mallacca S-S
Mallacca D-D
Consort S-S
Consort D-D
0
1
2
3
4
5
6
7
0 10 20 30 40 50 60
n2
Debranning time, t (s)
Mallacca S-S
Mallacca D-D
Consort S-S
Consort D-D
Figure 4.11 Parameters m2 (left) and n2 (right) as a function of debranning time, under Sharp-to-
Sharp and Dull-to-Dull dispositions for Mallacca (hard) and Consort (soft) wheats.
Chapter 4
99
If the initial shattering does not produce large bran particles, which can align with the roll
movement to allow effective scraping, then the production of small particles is
compromised: no (or few) large bran particles, therefore no (or little) shearing, therefore no
(or few) small endosperm particles, and mid-range particles dominate. This is the pattern
for hard wheat, and more so under S-S milling which is less effective in producing large
bran particles. (It is not the complete pattern; even for hard wheats Type 1 breakage is
typically around 60% (Campbell et al., 2012), and there is still significant production of
large bran particles from which endosperm is scraped.) Conversely, the crushing action of
D-D favours the production of nice large flat bran particles, particularly for soft wheats
(Fang and Campbell, 2002a,b), which orient nicely between the rolls and are scraped
effectively. Removing bran eliminates the production of these nice flat bran particles from
which endosperm can be scraped. In that case, the crushing action favours shattering rather
than scraping, and hence favours production of mid-range particles, more so for soft wheat;
hence the dramatic increase in α seen in Figure 4.9 for the Consort wheat. For soft wheats
the endosperm shatters easily such that, for debranned wheat, these mid-range particles are
smaller than for hard wheats, as illustrated in Figure 4.12 and evident in Figures 4.13 and
4.15. In the absence of bran to hold endosperm together, crushing favours shattering of the
endosperm into mid-range particles, and shearing of these particles is no longer so
effective. In hard wheats the endosperm resists shattering and breaks together with the bran
(Pomeranz and Williams, 1990), such that in the current work the effect of debranning on
the DNKBF parameters was less dramatic for the harder Mallacca wheat than for the softer
Consort. Appendix 2 shows the complete set of PSDs as described by the DNKBF from
Mallacca and Consort wheats under both milling dispositions at 0, 5, 10, 20, 30, 35, 40, 50
and 60 s of debranning.
Chapter 4
100
(a) Mallacca
(b) Consort
Figure 4.126 PSD as described by the DNKBF from (a) Mallacca and (b) Consort wheats
under both milling dispositions at 0 and 60 s of debranning.
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage
Type 2 breakage
Combined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
S-S 0s S-S 60s
D-D 0s D-D 60s
S-S 0s S-S 60s
D-D 0s D-D 60s
S-S 0s S-S 60s
Figure 4.12 PSD as described by the DNKBF from (a) Mallacca and (b) Consort wheats
under both milling dispositions at 0 and 60 S of debranning. Note: as was previously
explained, the shape of the Type 1 and Type 2 curves is determined by the parameters m1, n1
and m2, n2 respectively. α gives the proportion of the PSD that can be described as Type 1 breakage (predominantly mid-range particles). 100
Chapter 4
101
4.4.8 Overall effect of roll gap
The DNKBF normalises the data from different roll gaps onto a single curve. It is helpful
to see how that curve compares with the original data for each roll gap. Figures 4.13 to
4.16 present the experimental data for each roll gap and the fitted DNKBF, in both
cumulative and non-cumulative forms, for the two wheats and two dispositions. The non-
cumulative presentation of the data clearly evidences peaks that are well described by the
Type 1 curve. Overall the combined curves describe the experimental data extremely well,
with odd exceptions; notably the 0.3 mm roll gap data for Consort under D-D (Figure
4.14(b)). Previous work has similarly found that the DNKBF is not well adequate for soft
wheats under D-D (Campbell et al., 2012) particularly in accounting for effects of roll gap
(Fuh et al., 2014). Overall, however, it can be concluded that the DNKBF describes the
experimental data well and allows effects of debranning on first break milling of wheat to
be distinguished and quantified. Appendix 4 reports the complete set of experimental data,
cumulative and non-cumulative DNKBF for Mallacca and Consort wheats at 0, 5, 10, 20,
30, 35, 40, 50 and 60 s of debranning at S-S and D-D dispositions, roll gap 0.3, 0.5 and 0.7
mm.
Chapter 4
102
(I)
(II)
Figure 27 Experimental data (I, II), cumulative (I) and non-cumulative (II) DNKBF for Mallacca
(hard) wheat at 0 s (top) and 60 s (bottom) of debranning at S-S. From left to right, roll gap 0.3, 0.5
and 0.7 mm.
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
Figure 4.13 Experimental data (I,II), cumulative (I) and non-cumulative (II) DNKBF for
Mallacca (hard) wheat at 0 s (top) and 60 s (bottom) of debranning at S-S. From left to right, roll
gap 0.3, 0.5 and 0.7 mm. Note 1: dots represent the experimental data, and the curve is the
DNKBF fit. Note 2: the DNKBF curve is the combined curve from Type 1 and Type 2 breakages
curves (add them up) previously described in figure 4.12. The shape of the DNKBF is dictated by
the shape of both, Type 1 and Type 2 curves, which are in turn affected by their corresponding
shape parameters, m1 and n1 for Type 1 and m2 and n2 for Type 2.
Chapter 4
103
(III)
(IV)
Figure 28 Experimental data (III, IV), cumulative (III) and non-cumulative (IV) DNKBF for
Mallacca (hard) wheat at 0 s (top) and 60 s (bottom) of debranning at D-D. From left to right, roll
gap 0.3, 0.5 and 0.7 mm.
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
) ρ
(z)
ρ(z
)
ρ(z
) ρ
(z)
ρ(z
)
Figure 4.14 Experimental data (III, IV), cumulative (III) and non-cumulative (IV) DNKBF for
Mallacca (hard) wheat at 0 s (top) and 60 s (bottom) of debranning at D-D. From left to right, roll
gap 0.3, 0.5 and 0.7 mm. Note 1: dots represent the experimental data, and the curve is the
DNKBF fit. Note 2: the DNKBF curve is the combined curve from Type 1 and Type 2 breakages
curves (add them up) previously described in figure 4.12. The shape of the DNKBF is dictated by
the shape of both, Type 1 and Type 2 curves, which are in turn affected by their corresponding
shape parameters, m1 and n1 for Type 1 and m2 and n2 for Type 2.
Chapter 4
104
(V)
(VI)
Figure 29 Experimental data (V, VI), cumulative and non-cumulative DNKBF for Consort (soft)
wheat at 0 s (top) and 60 s (bottom) of debranning at S-S. From left to right, roll gap 0.3, 0.5 and
0.7 mm.
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
ρ(z
) ρ
(z)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
Figure 4.15 Experimental data (V, VI), cumulative (V) and non-cumulative (VI) DNKBF for
Consort (soft) wheat at 0 s (top) and 60 s (bottom) of debranning at S-S. From left to right, roll
gap 0.3, 0.5 and 0.7 mm. Note: dots represent the experimental data, and the curve is the DNKBF
fit. Note 1: dots represent the experimental data, and the curve is the DNKBF fit. Note 2: the
DNKBF curve is the combined curve from Type 1 and Type 2 breakages curves (add them up)
previously described in figure 4.12. The shape of the DNKBF is dictated by the shape of both,
Type 1 and Type 2 curves, which are in turn affected by their corresponding shape parameters, m1
and n1 for Type 1 and m2 and n2 for Type 2.
Chapter 4
105
(VII)
(VIII)
Figure 30 Experimental data (VII, VIII), cumulative (VII) and non-cumulative (VIII) DNKBF for
Consort (soft) wheat at 0s (top) and 60s (bottom) of debranning at D-D. From left to right, roll gap
0.3, 0.5 and 0.7 mm.
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
4.16 Figure 4.16 Experimental data (VII, VIII), cumulative (VII) and non-cumulative (VIII) DNKBF
for Consort (soft) wheat at 0 s (top) and 60 s (bottom) of debranning at D-D. From left to right,
roll gap 0.3, 0.5 and 0.7 mm. Note 1: dots represent the experimental data, and the curve is the
DNKBF fit. Note 2: the DNKBF curve is the combined curve from Type 1 and Type 2 breakages
curves (add them up) previously described in figure 4.12. The shape of the DNKBF is dictated by
the shape of both, Type 1 and Type 2 curves, which are in turn affected by their corresponding
shape parameters, m1 and n1 for Type 1 and m2 and n2 for Type 2.
Chapter 4
106
4.4.9 The suggested nature of Type 1 and Type 2 particles
Figure 4.17 illustrates the suggested nature of Type 1 and Type 2 particles. The figure
shows particles resulting from first break milling of Consort under D-D with a 0.5 mm roll
gap, showing particles recovered on the 1700 and 850 µm sieves and the particles smaller
than 212 µm. Clearly the 1700 µm particles are flattened bran with adhering endosperm,
while the 850 µm particles are more angular. The suggestion here is that large flat particles
created by the initial breakage then orient along the direction of roll movement. The
relative movement of the fast and slow rolls causes scraping of the particles which releases
small endosperm particles while leaving the bran relatively intact. Hence, both large and
small particles are produced from this mechanism, as described by the Type 2 curve, and
illustrated in Figure 4.17(a) and (c). Meanwhile, the initial breakage of the kernel also
produces mid-range, more angular particles (Figure 4.17(b)). These proceed between the
rolls without much further breakage, emerging as a population of mid-range particles.
Clearly this is a simplification, but it explains the strange observation that Type 2 breakage
appears to describe both very large and very small particles; the proposed mechanism says
that the initial creation of very large flat particles allows effective scraping of those
particles, and hence the co-production of very small particles. Meanwhile, Type 1 breakage
produces mid-range particles which are not scraped to a significant extent, such that there
is no simultaneous or subsequent creation of associated small particles. It also explains the
observed effect of debranning, as illustrated in Figure 4.18. Removal of bran precludes the
initial creation of large bran particles. In the absence of such particles, there is no
opportunity to produce, via the scraping action, the very small particles. Hence the
proportion of Type 2 breakage (both large and small particles) decreases and Type 1 mid-
range particles comes to dominate.
The effect of debranning is more dramatic for the soft wheat, as soft wheats create more of
these large bran particles in the first place. It is more dramatic under D-D, as the crushing
action of D-D milling is also more effective at creating large bran particles from which
smaller endosperm particles can be scraped. It is also consistent with the observation of
Fuh et al. (2014) that reducing roll gap increases the proportion of Type 2 breakage; at
smaller roll gaps, the scraping mechanism would extend down to smaller “large” particles.
Thus the proposed mechanism explains several observed phenomena and gives insights
into the nature of roller milling of wheat, deduced from the quantitative descriptions
Chapter 4
107
afforded by the DNKBF.
(a) 1700 µm
(b) 850 µm
(c) <212 µm
Figure 317 Example of particles resulting from First break milling of Consort under D-D with a 0.5
mm roll gap, recovered on the 1700 and 850 µm sieves and the particles smaller than 212 µm.
From Choomjaihan (2008) with permission.
Figure 4.17 Example of particles resulting from first break milling of Consort under D-D with a
0.5 mm roll gap, recovered on the 1700 and 850 µm sieves and the particles smaller than 212 µm.
From Choomjaihan (2008) with permission.
Chapter 4
108
Figure 32 Effect of debranning on wheat breakage.
Top: Bran breaks into large flat particles, which align with the rolls, and from which very small
particles are scraped as a result of the differential action of the rolls. Mid-range particles produced
from the initial breakage proceed between the rolls without further breakage.
Bottom: Removal of the bran precludes the production of large flat particles, and hence precludes
the scraping mechanism that produces very small particles. Production of mid-range particles
therefore dominates, which are small enough to pass through the roll gap without further breakage.
Figure 4.18 Effect of debranning on wheat breakage.
Chapter 4
109
4.5 Summary
Two representative UK wheat types, Mallacca (hard) and Consort (soft), were used in this
study. After conditioning, both wheats were debranned for nine different times and then
milled at three gaps under S-S and D-D dispositions. The resulting particle size distribution
was modelled using the Double Normalised Kumaraswamy Breakage function (DNKBF).
Around 20% and 17% of the bran was removed from Mallacca and Consort wheats,
respectively, after 60 s of debranning time, suggesting that all pericarp tissue and the
majority of the aleurone layer were removed. For both wheats the proportion of Type 1
breakage increased at longer debranning times under both S-S and D-D dispositions. The
values of the parameter a (collapsing parameter) were similar to previous reports. Sharp-
to-Sharp milling tended to produce more Type 1 breakage than Dull-to-Dull, while the
Mallacca (hard) wheat broke to give more Type 1 mid-range particles than the Consort
(soft). A breakage mechanism is proposed that suggests that debranning prior to milling
reduces the creation of large bran particles from which very small endosperm particles can
be scraped, thus, increasing the proportion of Type 1 particles. The effect of debranning
was more dramatic for the soft wheat under the Dull-to-Dull disposition.
The proposed mechanism explains several observed phenomena and gives insights into the
nature of roller milling of wheat, deduced from the quantitative descriptions afforded by
the DNKBF. The next chapter explains the characterisation performed on wheat botanical
tissues and some of the milled fractions used in the current chapter, in order to get insights
for the further extension of the DNKBF to include particle composition during first break
roller milling.
Chapter 5
110
CHAPTER 5
CHARACTERISATION OF BOTANICAL TISSUES AND
MILLED FRACTIONS WITH FTIR AND THEIR SUGAR
PROFILES USING HPLC
5.1 Introduction
The identification of botanical components contributing to particles of a given size is a
considerable challenge. As was described in Chapter 3, this could be achieved using
specific biochemical markers that are exclusively associated with particular botanical
components in the wheat grain. Therefore, in the current work pericarp, aleurone,
endosperm and germ (the four major wheat components), along with different milled
fractions, were investigated by Fourier Transform Infrared (FTIR) spectroscopy coupled
with multivariate analysis. Botanical tissues were isolated and ball mill ground. Spectra of
milled fractions and ground tissues were collected. The resulting data set was analyzed
with Principal Component Analysis to identify correlated clusters of data. Similarly, a wide
range of saccharides are present throughout wheat tissues, in different proportions
depending on the biochemical activities of each particular tissue, hence, by applying
hydrolysis followed by High Performance Liquid Chromatography (HPLC) analysis, the
identification and quantification of the sugar content in each wheat tissue and milled
fractions obtained can be possible. Therefore, the isolated botanical tissues and the milled
fractions were also chemically hydrolyzed and analyzed using HPLC to determine sugar
profiles. A calibration curve was built up considering the most representative sugars
present in endosperm (glucose) and pericarp (arabinose and xylose) fractions, to quantify
the proportion of these sugars in the botanical tissues and some milled fractions. This
chapter describes how the botanical tissues were obtained, the collection of the spectra
from the four major wheat components and milled fractions, their multivariate analysis,
and the sugar profiles analysis of endosperm, pericarp and some milled fractions using the
HPLC.
Chapter 5
111
5.2 Materials
As described in Chapter 4, Mallacca, a representative UK hard wheat, and Consort, a
representative UK soft wheat, were used in this study (Figure 4.1).
5.3 Experimental procedures
5.3.1 Dissection of wheat components into botanical tissues
Botanical tissues were isolated from whole wheat kernels. The dissection procedure for
wheat kernels performed in this work followed the protocol reported by Choomjaihan
(2008):
1. Wheat grains were soaked in ultrapure distilled water for 7 hours with a ratio grain:
water of 1:10 by weight.
2. The grains were left in the beakers by draining the soaking water from them.
3. For dissection, each grain was left on a flat and clean bench. Both tips of the grain
were cut with the scalpel and three parts were separated, brush, germ and body part,
as illustrated in Figure 5.1 The three parts isolated were collected and stored in
Eppendorf tubes.
4. The germ was isolated by cutting the kernel in a V-shape as illustrated in Figure
5.1. Dissection forceps were used tu pull the pericarp layer from the germ.
5. The pericarp layer was separated from point 1 (blue dot) to point 2 (red dot), as
observed in Figure 5.1.
6. The aleurone layer was removed from the remaining body part from point 1 to
point 2 (Figure 5.1). Endosperm was scraped from aleurone by using the dull edge
of the scalpel, because endosperm and aleurone are strongly attached to each other.
7. The crease was separated from the remaining body part. Pure endosperm was
recovered with this operation, as shown in Figure 5.1.
8. Pericarp, aleurone, endosperm and germ from both Consort and Mallacca wheats
were placed separately into dry and clean Eppendorf tubes.
Approximately 100 whole wheat grains were soaked and then dissected following the
procedure described above, recovering (on a wet basis) about 160 mg of pericarp, 110 mg
of aleurone, 1600 mg of endosperm and 95 mg of germ. These values are quite similar to
Chapter 5
112
the values reported for the same botanical tissues by Choomjaihan (2008). The dissection
procedure allowed ten kernels to be dissected per day.
Figure 33 Hand dissections of wheat botanical tissues. Adapted from Choomjaihan (2008).
5.3.2 Moisture content
The moisture content of each botanical component was measured according to the AACC
standard method 44-19. Basically, samples were placed into a hot air oven at 135oC for 3
hours. The moisture content of the samples was then calculated as:
10012
32
WW
WWM (5.1)
PERICARP ALEURONE ENDOSPERM
CREASE
GERM
Whole
wheat
kernel
Germ is removed for further analysis. Hair
brush is removed and discharged
1 2
Pericarp was
removed from
point 1 to point
2. Aleurone was
removed from
the remaining
part from point
1 to point 2
Crease was removed and the remaining
pure endosperm was recovered
Figure 5.1 Hand dissections of wheat botanical tissues. Adapted from Choomjaihan (2008).
Chapter 5
113
where,
M = moisture content of the sample (%, wet basis)
W1 = weight of empty container (g)
W2 = weight of container plus sample before heating (g)
W3 = weight of container plus sample after heating (g)
5.3.3 Sample preparation for FTIR analysis
Botanical tissues
The four botanical tissues, pericarp, aleurone, endosperm and germ previously isolated, for
both wheat types were ground to a homogenous size (<100 µm) using a RETSCH Mixer
Mill MM 200 ball mill, (for which the size reduction principle is impact and friction) for 3
mins at a frequency of 30 s–1
. Acetone was used for cleaning both the grinding jars (Figure
5.2) and the balls prior to grinding each batch, in order to avoid contamination among
samples.
Figure 34 Ball Mill used for grinding the botanical tissues.
Grinding
jars
Figure 5.2 Ball mill used for grinding the botanical tissues.
Chapter 5
114
Milled fractions
The milled fraction samples that were analyzed are summarized in Table 5.1; these had
been previously prepared by Choomjaihan (2008) and stored frozen at –18oC.
Choomjaihan (2008) reduced the size of the milled fractions using the Falling Number
hammer mill (Perten Instruments AB, Sweden) (Figure 5.3) with a 500 µm sieve aperture,
in order to have consistent samples suitable for further analysis. A total of 192 samples
were analysed (two wheats × two dispositions × six rolls gaps × eight fractions = 192).
Figure 35.3 Falling number hammer mill.
Table 5.1 Summary of the milled fractions analyzed.
WHEAT TYPES Consort and Mallacca
ROLL DISPOSITION Sharp-to-Sharp (S-S) and Dull-to-Dull (D-D)
ROLL GAP (mm) 0.3, 0.4, 0.5, 0.6, 0.7, 0.8
PARTICLE SIZE (µm) <212, 212, 500, 850, 1180, 1400, 1700, 2000
Figure 5.3 Falling number hammer mill.
Table 5.1 Summary of the milled fractions analyzed.
Chapter 5
115
5.3.4 Fourier Transform Infrared (FTIR) spectroscopy
Infrared spectroscopy is a particularly useful technique that helps to identify functional
groups in an unknown compound (Chan et al., 2010; Smith, 2011; Chalmers et al., 2012).
To understand Infrared spectroscopy, it is necessary to understand first the chemical bases
behind it, which are associated with the vibrations of the chemical bonds of molecules.
Chemical bonds vibrate with different amounts of energy at a frequency that depends on
their strength and the masses of the atoms; therefore, light atoms give higher frequencies
and heavier atoms give lower frequencies. Similarly, stronger bonds give higher
frequencies than weaker bonds. Every bond such as N-H, O-H, C=O, etc, absorb energy at
a particular frequency, enabling their identification (Chan et al., 2010; Smith, 2011;
Chalmers et al., 2012; www.rsc.org).
FTIR is widely used to provide a precise measurement without destroying the sample. The
principle of the FTIR is as follows: the infrared emission (consisting of a range of
frequencies) from the infrared (IR) source, travels by a sequence of mirrors into the sample
(placed on a suitable holder). The frequencies are absorbed, some of them more than
others, depending on what bonds are present in the sample (Chan et al., 2010; Chalmers et
al., 2012; www.rsc.org). All the remaining radiation that was not absorbed by the sample
arrives in the detector. The interferogram (produced by the interferometer) contains the
information associated to the intensity of all IR radiation at all frequencies at once from the
infrared source, and then passes through the sample and goes to the detector. The
interferogram that reaches the detector is then decoded by the Fourier Transformation
algorithm, which calculates the intensity of the IR radiation at each frequency individually.
The computer processes the transformation given by the algorithm and produces a graph of
percentage Transmission or Absorbance against wavenumber; the spectral data represented
on the graphs is interpreted by the analyst. Wavenumbers are measured in cm–1
and are
inversely proportional to frequency (Field et al., 2007; Griffiths and De Haseth, 2007;
www.rsc.org).
A useful device attached to moderns FTIR instruments is the Attenuated Total Reflection
(ATR) diamond. Briefly, the sample is placed onto the crystal area and the IR beam is
directed to the sample by a mirror. The upper surface of the sample reflects back the IR
beam before being directed by a second mirror to the detector. The biggest advantage that
Chapter 5
116
the ATR method offers is that a low quantity of sample is needed for the analysis, and little
or no preparation is required for most samples (Griffiths and De Haseth, 2007).
In the current work, the spectra analysed were collected using a Spectrum Two
Spectrometer ® equipped with an attenuated total reflectance accessory (Perkin Elmer,
USA). The equipment was kindly lent by Perkin Elmer ® as part of a collaboration.
Figure 5.4 shows the FTIR equipment with the ATR accesory. Interferograms were
obtained in transmission mode and collected at 4 cm–1
resolution. Spectra were recorded
between 550 and 4000 cm–1
. Samples were placed onto the ATR surface, which featured a
circular diamond crystal of diameter 2 mm. Less than a milligram of sample was used to
cover the crystal area. Once the samples were placed on the crystal area, the pressure rod
was positioned over the crystal/sample area. The pressure rod locks into a precise position
above the diamond crystal. Force was applied to the sample by pushing it onto the diamond
surface until “clicked”, indicating the pressure limit. The forced applied was always the
same for all the samples. For each sample, spectra were collected in triplicate. An air-
background scan was recorded every six spectra. The ATR diamond surface and pressure
rod were cleaned with a cotton wool round pad wet with a mixture of water-ethanol before
each spectral recording.
Figure 36 Spectrum two (FTIR-ATR) spectrometer lent by Perkin Elmer ®.
Pressure
rod
ATR
diamond
surface
Figure 5.4 Spectrum two (FTIR-ATR) spectrometer lent by Perkin Elmer ®.
Chapter 5
117
5.3.5 Mathematical analysis
Multivariate statistics (Principal Component Analysis, PCA and Partial Least Squares,
PLS), were applied to raw spectral data, to baseline-corrected spectral data, and to
baseline-corrected, normalised and second derivative spectral data. An inspection of the
data was undertaken on the second derivative with a smoothing degree of 13 points in the
same ranges described above. The algorithm used to calculate the derivatives was the
Savitsky-Golay method with a polynomial order of 3. Derivative curves are extensively
used to reduce, in general, the signal-to-noise ratio. The second derivative in particular,
improves the detection of small spectral differences such as broadening of the bands,
unresolved peaks and peak shifts (Jollife, 1986; Robert et al., 2005; Mark and Workman,
2007; Rinan et al., 2009). The Savitzky-Golay method is used to calculate the derivative of
a smooth curve constructed through the original data points of the original spectrum, using
a number of neighbouring data points to estimate the curve (Mark and Workman, 2007;
Rannin et al., 2009).
The analysis was performed using two software packages: Spectrum Two software
supplied with the instrument (Perkin Elmer, Inc.) and The Unscrambler software (CAMO,
Norway). PCA and PLS were computed on pure ground wheat botanical tissue spectra
based on the whole spectral range (550-4000 cm–1
) and the fingerprint region (810-1800
cm–1
), using The Unscrambler software package. PCA analysis of the spectral data of
milled fractions was performed only in the fingerprint region and using mathematical pre-
treatment (baseline correction and second derivative). Only the most representative PCA
plots are shown to illustrate the results obtained.
Chapter 5
118
5.5 Manual isolation of wheat kernel tissues
The four major botanical components of the wheat kernel, endosperm, aleurone, pericarp
and germ, were obtained using dissection as described in Section 5.3.1.
Visual inspection under both stereomicroscope and microscope was used to confirm the
identity and purity of the isolated components. Figure 5.5 shows example images of the
four components obtained from the Mallacca wheat.
Figure 37 The four major wheat components hand dissected from Mallacca wheat.
A) Pericarp, B) Aleurone, C) Endosperm, D) Germ
The moisture content was determined according to Section 5.3.2 for both wheat types
before and after soaking, and for the botanical tissues recovered after hand dissection
(before grinding). These results are presented in Table 5.2.
0.1 mm
B A
D C
1 mm
0.1 mm
1 mm
Figure 5.5 The four major wheat components hand dissected from Mallacca wheat.
A) Pericarp, B) Aleurone, C) Endosperm, D) Germ
Chapter 5
119
Table 38 Moisture content of both wheat types before and after soaking, and the botanical
components.
Sample Moisture
content
(%)
Malacca (whole),
before soaking
13.28
Malacca (whole),
after soaking (7 h)
43.66
Pericarp Malacca 15.43
Endosperm
Malacca
13.34
Germ Malacca 16.92
Aleurone Malacca 13.00
Consort (whole),
before soaking
13.41
Consort (whole),
after soaking (7 h)
45.00
Pericarp Consort 14.75
Endosperm
Consort
13.68
Germ Consort 18.03
Aleurone Consort 13.67
Figure 5.5(A) shows the removed pericarp, which is the first layer separated from the
wheat kernel; it shows the “golden brown colour” that is the common colour observed in
the whole wheat kernel. The lignified cell walls are evident in the pericarp layer under the
microscope view.
After pericarp, aleurone is the second layer separated from the wheat grain because it is
situated in the interface between the pericarp and the starchy endosperm. Figure 5.5(B)
shows the translucent creamy white colour of the aleurone layer, visible under the
microscope. Polygonal shapes featuring diverse sizes and shapes are also observed.
The starchy endosperm is the third layer separated from the wheat grain, which is easily
obtained once the bran layer (pericarp and aleurone) has been removed. Figure 5.5(C)
shows the characteristic white colour and starchy texture of the pure endosperm,
confirming its purity.
Germ was isolated from the wheat kernel as described in Section 5.3.1; the sample
illustrated in Figure 5.5(D) shows the different characteristics of the two main germ
Table 5.2 Moisture content of both wheat types before and after soaking, and the botanical
components.
Chapter 5
120
components; the embryonic axis (inner part of the germ) and the scutellum (outer part of
the germ). The scutellum features a more brownish colour (similar to the pericarp colour)
than the embryionic axis, which is more golden-yellowish colour.
Fulcher et al. (1972), Barron et al. (2005, 2007), Barron and Rouau (2008), Choomjaihan
(2008), Hemery et al. (2009) and Barron (2011) described the same features observed in
the pictures shown above, indicating that the technique used for the botanical layers
dissection was appropriate, enabling their FTIR analysis.
5.6 FTIR-ATR analysis of ground botanical tissues
Figure 5.6 shows an overlay of the averaged spectra for pericarp, aleurone, endosperm and
germ for milled and dissected components from Mallacca wheat.
Figure 39 FTIR-ATR spectra of milled Mallacca dissected wheat grain.
Pericarp, Aleurone, Endosperm and Germ.
Figure 5.6 FTIR-ATR spectra of milled Mallacca dissected wheat grain.
Pericarp, Aleurone, Endosperm and Germ
Protein peaks, from left to right carbonyl, amide I and amide II
Carbohydrate peaks. Starch at 1000 and 1022, arabinoxylan at 1041 and 1075 cm
–1
CH stretching, mainly lipid
Water
Chapter 5
121
The main carbohydrate, protein and lipid regions of the spectra can be clearly observed,
and the peak intensities reveal the known make up of the different tissues in wheat.
Endosperm (in pink) has very strong starch absorption at 1000 and 1022 cm–1
, and little
else (the amide I peak overlaps with a peak obtained from purified starch (and other
carbohydrates e.g. pectin) at 1630 cm-1
, which is why pericarp and endosperm tissue
appear to have an amide I peak, but no carbonyl or amide II peak), showing that
endosperm tissue is rich in starch and contains protein (gliadins and glutenins) (Barron,
2011; Warren et al., 2011). The aleurone layer (in red) has both starch and arabinoxylan
absorption peaks in the carbohydrate region of its spectra which are difficult to resolve
without further analysis, but unlike endosperm, aleurone has considerable adsorption of
protein and some lipid. The germ (in green) has only a small shoulder for starch, and its
carbohydrate region is dominated by arabinoxylan (absorbance at 1041 with a shoulder at
1075 cm–1
). Germ is rich in lipid and protein. The pericarp (in turquoise) carbohydrate
region is dominated by cell wall polysaccharides, predominantly arabinoxylan, β-glucans
and cellulose (Kacurakova and Wilson, 2001; Jamme et al., 2008; Barron, 2011). The
pericarp is essentially free of protein, starch and lipid, and has noticeably lower water
content than the other tissues from the wheat grain.
There are, therefore, significant differences that can be exploited to identify the different
fractions spectrally. The spectra shown in Figure 5.6 are very similar to those obtained by
Barron (2011) over a wider range of wheat cultivars, showing that these spectral signals
are consistent between wheats.
5.7 Principal Component Analysis (PCA) on ground isolated tissues
PCA is a powerful data analysis technique able to reduce the dimensionality of data
collected by computing PC (principal components). PCs are linear combinations of the
variables of the system; they do not have any physical meaning, however they contain
most of the information from the original data, hence, PCs extract the useful information
and reduce the dimensionality (and the noise) of the data. The PCA model is constructed
from a matrix that contains in each row a sample and in each column a variable (Smith,
2002; Blanco-Romía and Alcalá-Bernández, 2009).
A score plot is the result of the original data projected in a new coordinated system of PCs.
Some characteristics about the relationship or differences of the samples to each other can
Chapter 5
122
be better visualized and understood based on the location of the samples projected in the
score plot. This means that in a score plot, all the samples close to each other are similar
and those samples that are separated from each other are different. Indeed, within a score
space, various groups of samples can be formed, such, that related samples are clustered
together (Smith, 2002; Hair et al., 2006).
The relative contributions of the original variables to the principal components are
described by the loading; it shows how much each of these original variables contributes
and explains the variance. In PCA, in general two or three PCs are enough to relate
variables (Smith, 2002; Hair et al., 2006).
Figure 5.7 shows the PCA score of the four botanical tissues isolated from both wheat
types. PCA analysis was performed first on the whole spectra recorded (550-4000 cm–1
)
without any spectral pre-treatment to see, at first, if the four botanical tissues isolated could
be separated into four distinctive clusters, and to see if it was feasible to identify specific
spectral signatures from these ground tissues.
Chapter 5
123
Figure 40 PCA scores of the four botanical tissues (pericarp, aleurone, endosperm and germ,
labelled P, A, E and G, respectively) of each wheat type (Consort and Mallacca, subscripted c and
m, respectively).
Figure 5.7 PCA scores of the four botanical tissues (pericarp, aleurone, endosperm and germ,
labelled P, A, E and G, respectively) of each wheat type (Consort and Mallacca, subscripted c and
m, respectively).
Chapter 5
124
Figure 41 Loading vectors associated to (A) PC1 and (B) PC2
Figures 5.7 and 5.8 show a map built up from the first and second principal components,
accounting for 91% of the variability, and the loading vectors for PC1 and PC2. The
largest variance of the data is along PC1 (54%), which separates, mainly the endosperm
(positive side of PC1), from the rest of the tissues. Figure 5.8(A), which is the Loading
associated with PC1, shows the strongest absorbance at 1000 cm–1
, which is related to the
starch content in the starchy endosperm (Barron, 2011; Warren et al., 2011).
Along PC2 (37%), from top to bottom, germ tissue forms a well defined cluster in the
positive part of PC2, compared with the rest of the tissues. In Figure 5.8(B), with is the
loading associated with PC2, shows the strongest absorbance at 2924 and 2854 cm–1
,
which are associated with lipids (Barron et al., 2005, Barron, 2011); this makes sense, as
lipids are mostly present in the germ fraction of the wheat kernel. Aleurone and pericarp
A
B
2924 cm-1
2854 cm-1
1540 cm-1
1740 cm-1
1648 cm-1
1000 cm-1
Figure 5.8 Loading vector associated to (A) PC1 and (B) PC2.
Chapter 5
125
tissues clustered together in the negative part of PC2. The Loading associated with PC2
(Figure 5.8(B)) shows higher absorbance at 1740 cm–1
, which is associated with the lipid
part, and at 1648 and 1540 cm–1
, which are associated with the protein region present in the
aleurone layer (Pomeranz, 1988, Barron, 2011).
The PCA on the raw spectral data did not distinguish well between aleurone and pericarp.
The grouping effect of aleurone with pericarp may be caused by: (i) spectral data was not
good enough and/or (ii) the isolated tissues were not absolutely pure (some aleurone tissue
could have remained attached to pericarp).
To tackle the first point, the spectral data perhaps needs to be improved (use of
mathematical tools such as normalization, derivatives, smoothing, etc), hence, in order to
improve the clusters obtained in the PCA, the mathematical pre-treatment described in
Section 5.3.5 was applied to the data set of the four ground botanical tissues, first in the
whole spectra, and, second, in the polysaccharide fingerprint region (1800-810 cm–1
). PCA
was performed afterwards, in both cases.
Only minima exhibited within the second derivative spectra were analysed as this is where
the biochemical information is available after pre-processing the data (Mark and
Workman, 2007). Inspection of the second derivative spectra confirmed the observations
described above, revealing some differences among the botanical tissues, enabling their
identification. Some of these characteristics found are described below and shown in
Figures 5.9 and 5.1.
Chapter 5
126
Figure 42 (A) Baseline corrected and normalised FTIR-ATR spectra and (B) second derivative
computed for each botanical tissue of milled Consort dissected wheat grain. Whole spectra (4000
to 550 cm-1
). Pericarp, Aleurone, Endosperm and Germ.
A
B
Figure 5.9 (A) Baseline corrected and normalised FTIR-ATR spectra and (B) second derivative
computed for each botanical tissue of milled Consort dissected wheat grain. Whole spectra (4000
to 550 cm–1
. Pericarp, Aleurone, Endosperm and Germ.
Chapter 5
127
Figure 43 (A) Baseline corrected and normalised FTIR-ATR spectra and (B) second derivative
computed for each botanical tissue of milled Consort dissected wheat grain. Spectral range from
1800 to 810 cm-1
. Pericarp, Aleurone, Endosperm and Germ.
Carbohydrate region (900-1400 cm–1
). Two higher absorbance features were
observed in this region: at 1022 cm–1
, with shoulders at 930, 1075 and 1151
cm–1
, and at 1041 cm–1
with shoulders at 1158 and 997 cm–1
(Figures 5.9 and
5.10). At 1022 cm–1
, the vibration of -CH2OH and C-OH groups of
carbohydrates (sucrose, glucose), is found (Dukor, 2002; Huleihel et al., 2002),
mainly associated with starch (Smits et al., 1998; Kacurakova and Wilson.,
2001), which is the main component of endosperm tissue. The starchy
endosperm is the major source of energy of the grain. Conversely, PO2–
stretching and C-OH stretching of oligosaccharides are found at 1158 cm–1
(Zanyar et al., 2008); at 985 cm–1
the vibration of OCH3 groups (mainly from
polysaccharides) is found (Robert et al., 2005); all of them primarily associated
with cell wall polymers such as arabinoxylan, β-glucans, cellulose and lignin
(Himmelsbach et al., 1998; Jamme et al., 2008; Barron, 2011), which are highly
present in the pericarp tissue because this is the protective layer of the grain.
Protein region (1721-1400 cm–1
). In this region, higher absorbance was
observed at 1740, 1648 and 1540 cm–1
(Figures 5.9 and 5.10), mainly in the
aleurone layer and some in endosperm and germ tissues. According to Toone
(1994) and Zanyar et al. (2008), the stretching vibration mode of the ester
group (C=O), mainly associated with phospholipids and triglycerides, is found
A B
cm–1 cm–1
Figure 5.10 (A) Baseline corrected and normalised FTIR-ATR spectra and (B) second derivative
computed for each botanical tissue of milled Consort dissected wheat grain. Spectral range from
1800 to 810 cm–1
. Pericarp, Aleurone, Endosperm and Germ.
Chapter 5
128
at 1740 cm–1
, which may be related to lipid transfer proteins present in the germ
tissue. At 1648 cm–1
are found unordered random coils and turns of amide I,
while at 1540 cm–1
, the protein amide II absorption is observed, predominately
related with the β-sheet of amide II (Bonnin et al., 1999). Goormaghtigh et al.
(1994) reported that some amino acid side chain groups absorb in amide I and II
regions; for example, the data at 1585 and 1567 cm–1
might correspond to
asparagine (Asp) and glutamine (Glu) COO- residues associated with wheat
germ agglutinin in germ. The starchy endosperm is characterized by high
contents of Glu and proline (Pro), mainly associated with gluten proteins
(gliadins and glutenins) and low levels of basic aminoacids, while the aleurone
and germ contain significantly less Pro and Glu, with high levels of arginine
(Arg) (associated with storage globulins and albumins mainly present in the
aleurone layer) and Asp (associated with wheat germ agglutinin in germ)
(Nagata and Burger, 1974). Although the protein concentration in the
endosperm is a third of that found in the germ and less than half of that found in
the aleurone layer, endosperm proteins represent nearly three quarters of the
total grain protein content (Pomeranz, 1988).
Lipid region (2700-3100 cm–1
). In this region, higher absorbance was obtained
at 3009, 2924 and 2854 cm–1
(Figure 5.6), mainly in the germ tissue. Zanyar et
al. (2008) described that at 3009 cm–1
, the vibration of the =C-H groups are
related to lipids and fatty acids. At 2924 cm–1
is found the asymmetric
stretching vibration of CH2 of acyl chains (lipids) (Fabian et al., 1995).
Although lipids are unevenly distributed in the wheat grain, the higher
proportion is contained in the germ tissue (MacMasters et al., 1971).
All these observations are in agreement with the work of Barron and co-workers (Barron et
al., 2005, 2007; Barron and Rouau, 2008; Barron, 2011), who have described the same
contents at similar wavenumbers for different pure hand-isolated wheat tissues, such as
endosperm, aleurone layer, intermediate layer, outer pericarp and whole outer layers, over
a wider range of wheat cultivars, showing that these spectral results are very reproducible
and the technique applied for the hand-dissection was appropriate, although it may be
improved for a better separation of aleurone from endosperm, as shown in Figures 5.11 and
5.12 and described in the next part.
Chapter 5
129
5.7.2 Improving the separation of the clusters by using
mathematical pre-treatment tools
The PCA map constructed (from 24 spectra) from the ground botanical tissues after the
mathematical pre-treatment described in Section 5.3.5 is shown in Figures 5.11 and 5.12.
Figure 44 (A) PCA scores of the four botanical tissues (pericarp, aleurone, endosperm and germ) of
each wheat type (Consort and Mallacca) and (B) their associated loadings for PC1 and PC2 (shown
as Bi-plot of scores (dots) and loadings (wavenumbers)), after mathematical pre-treatment. Whole
spectra (4000 to 550 cm–1
).
A
B
Figure 5.11 (A) PCA scores of the four botanical tissues (pericarp, aleurone, endosperm and germ)
of each wheat type (Consort and Mallacca) and (B) their associated loadings for PC1 and PC2
(shown as Bi-plot of scores (dots) and loadings (wavenumbers)), after mathematical pre-treatment.
Whole spectra (4000 to 550 cm–1
).
Chapter 5
130
Figure 45.12 PCA (A) scores of the four botanical tissues (pericarp, aleurone, endosperm and
germ) of each wheat type (Consort and Mallacca) and (B) their associated loadings for PC1 and
PC2 (shown as Bi-plot of scores (dots) and loadings (wavenumbers)), after mathematical pre-
treatment. Spectral range from 810 to 1800 cm–1
A
B
5.12 Figure 5.12 (A) PCA scores of the four botanical tissues (pericarp, aleurone, endosperm and germ)
of each wheat type (Consort and Mallacca) and (B) their associated loadings for PC1 and PC2
(shown as Bi-plot of scores (dots) and loadings (wavenumbers)), after mathematical pre-treatment.
Spectral range from 810 to 550 cm–1
.
Chapter 5
131
As shown in Figures 5.11 and 5.12, germ tissue was easily distinguished from pericarp
tissue, whichever the spectral range selected (whole spectra or from 810-1800 cm–1
region). Endosperm and aleurone showed close coordinates in principal components 1 and
2 that account for the 93% of the variability. This could be explained by limitations in hand
dissection. Related to the loadings observed in the Bi-plot (Figures 5.11 and 5.12), the
wavenumbers located far away from the centre of the coordinates (zero), are the ones that
highly contribute to the variance in the data set. The wavenumbers related to the
carbohydrate fingerprint region (900-1400 cm–1
), which are mainly associated with the
starch content in the endosperm and with cell wall polymers present in the pericarp; the
wavenumbers of the protein region (1721-1475 cm–1
), mostly related to the secondary
structure of the proteins present in the aleurone and endosperm, and some in germ tissue;
and the wavenumbers of the lipidic region (3100-2800 cm-1
), mainly associated to the fatty
acids present in the germ, are the wavenumbers that revealed the major contribution to the
PC1, which is the principal component that accounts for the 65-68% of the variability
(Figures 5.11 and 5.12).
The best separation of the four ground wheat botanical tissues in PCA was observed after
computing the mathematical pre-treatment (base line correction and second derivative).
Furthermore, the Bi-plot of scores and loadings (Figures 5.11 and 5.12) clearly show the
wavenumbers (the ones that are far away from the centre in the Bi-plot) that highly
contribute to explain the chemical structures present in the samples, enhancing the
qualitative analysis.
Results of the botanical tissues analysis suggested that some regions in the spectra were
more intense for specific wheat component, indicated by specific wavenumbers. This
could enable the identification and quantification of the different tissues in samples of
unknown composition (i.e milled fractions), which is the aim of the current work.
Chapter 5
132
5.7.3 PCA analysis of milled samples
Figure 5.13 shows the spectra of different milled fractions from Mallacca wheat. Principal
Component Analysis was performed in the milled samples as described in Section 5.3.5.,
and only in the region from 810-1800 cm–1
, which is the fingerprint region. The analysis
was performed only for Mallacca samples milled under D-D disposition with a Roll Gap of
0.5 mm. Based on the botanical tissues analysis the spectra of milled samples were base
line-corrected. Second derivative using the Savitsky-Golay method was performed in both
the whole spectra (data not shown) and in the fingerprint region of the spectra (810-1800
cm–1
). The lipid region did not show high vibration, thus, the only section used for the
PCA analysis was the fingerprint region (810-1800 cm–1
), as described above. Equally,
only the second derivative plots and their original spectra in this region are shown in
Figure 5.14.
Figure 46 Spectra of different milled fractions. Mallacca wheat milled under D-D disposition with
a Roll Gap of 0.5 mm. <212µm, 212µm, 500µm, 850µm, 1180µm,
1400µm, 1700µm, 2000µm.
Figure 5.13 Spectra of different milled fractions. Mallacca wheat milled under D-D disposition
with a roll gap of 0.5 mm. <212 µm, 212 µm, 500 µm, 850 µm, 1180 µm,
1400 µm, 1700 µm, 2000 µm.
Chapter 5
133
As can be seen from Figure 5.13, the same peaks that were found in the botanical tissues
were observed here as well, in all the samples, in different proportions, thus, enabling the
potential quantification of the wheat components in the milled samples.
Figure 5.147 (A) Base line corrected and normalised spectral data and (B) Second derivative
computed. Range (1800-810 cm-1
). Mallacca wheat milled under D-D disposition with a Roll Gap
of 0.5 mm.
<212µm, 212µm, 500µm, 850µm, 1180µm, 1400µm, 1700µm,
2000µm.
Figure 5.14 shows the spectra from all the milled samples split along with their second
derivative plots. Interestingly, the strongest vibration was observed in the carbohydrate
region, from 900-1200 cm–1
, with the highest peak centred at 1075cm–1
, which is related to
cell wall polymers.
After computing the PCA analysis only in the fingerprint region (810-1800 cm–1
) with the
milled fractions (duplicates), quite defined clusters were obtained, as shown in Figure 5.15.
A B
Figure 5.14 (A) Base line corrected and normalised spectral data and (B) Second derivative
computed. Range (1800-810 cm–1
). Mallacca wheat milled under D-D disposition with roll gap of
0.5 mm. <212 µm, 212 µm, 500 µm, 850 µm, 1180 µm, 1400 µm,
1700 µm, 2000 µm.
Chapter 5
134
Figure 48 PCA scores of wheat milled samples, for size fractions from Mallacca wheat milled
under D-D disposition with a Roll Gap of 0.5 mm.
Figure 5.15 shows a map built up from the first and second principal components
accounting for 96% of the variability. The largest variance of the data is along PC1 (92%),
which separates the fine particles (<212, 212 and 500 µm, negative side of PC1) from the
coarse particles (850, 1180, 1400, 1700, 2000 µm, positive side of PC1). Note that this is
not a particle size artefact, as all fractions were milled to a consistent size for analysis.
Figure 5.16, which is the Loading associated with PC1, shows the strongest absorbance in
the carbohydrate region (quite similar to the results obtained in the second derivative, in
Figure 5.14(B)), mainly at 1065 cm–1
and 1075 cm–1
, which is related to cell wall
polysaccharides associated with the pericarp tissue (Jamme et al., 2008; Barron, 2011).
Interestingly, the milled samples 850, 1400 and 1700 µm are closely clustered, suggesting
that these samples were similar in composition; however there is no evidence of a smooth
transition in composition with particle size, i.e of the clusters forming a smooth curve as
they move from <212 µm to >2000 µm.
.
Figure 5.15 PCA scores of wheat milled samples, for size fractions from Mallacca wheat milled
under D-D disposition with a roll gap of 0.5 mm.
Chapter 5
135
Figure 49 Loading vectors associated with PC1.
5.8 Calculating the proportion of botanical tissues throughout the milled
samples
Some trials were carried out, to try to quantify the content of each botanical tissue
(pericarp, aleurone, endosperm and germ) within a milled fraction. For example, bran
tends to stay intact as relatively large particles during first break roller milling, such that
the large particles are expected to contain higher proportions of pericarp and aleurone,
while small particles would be expected to be higher in endosperm material. As a first
attempt, from the whole spectra of botanical tissues, the eight most representative peaks
were chosen. The eight peaks were found in the baseline corrected and normalised spectra
of the four botanical tissues but in different proportions (Figure 5.6). The absorbance at
each wavenumber from these peaks was selected and their heights were measured with the
Spectrum software (Perkin Elmer, USA). These data were introduced into an Excel
spreadsheet. Two samples were tested, Mallacca wheat milled under D-D disposition with
a roll gap of 0.5 mm, 2000 µm and <212 µm particle size (although both samples were
ground as mentioned in Section 5.3.3 in order to have consistent samples). The heights at
the wavenumbers measured for the botanical tissues were collected equally for the
samples. The theoretical height was calculated as follows:
Figure 5.16 Loading vectors associated with PC1.
Chapter 5
136
)()()()( titititih GGEEAAPPC (5.2)
where:
Ch = height calculated for each peak
Pi = height of peak i in the pericarp spectrum
Pt = pericarp fraction
Ai = height of peak i in the aleurone spectrum
At = aleurone fraction
Ei = height of peak i in the endosperm spectrum
Et = endosperm fraction
Gi = height of peak i in the germ spectrum
Gt = germ fraction
This can be expressed as a matrix equation in short form as:
bxA (5.3)
where A describes a linear mapping from the vector space x to the vector space b.
The matrices A, x and b were composed as follows:
=
(5.4)
Chapter 5
137
where:
(1) Am,n is the height of peak n in m botanical component
(2) xm,i is the proportion of the botanical component m in i particle size
(3) bn,i is the height of peak n in i size fraction
(4) m refers to the four botanical components, P, A, E, and G (pericarp, aleurone,
endosperm and germ)
(5) n refers to the eight peaks chosen at a particular wavenumber (see first column in
table 5.3).
(6) i refers to the milled size fraction (2000, 1700, 1400, 1180, 850, 500, 212 and
<212µm)
Table 5.3 shows five cells highlighted, four at the top, corresponding to the fraction of each
botanical tissue, and one at the bottom, which is the total error.
Table 50 Example of height data introduced in the spreadsheet to quantify the relative proportion of
each botanical component in the unknown sample.
MALLACCA Germ Aleurone Pericarp Endosperm
0.19 0.00 0.00 0.81
Botanical Tissues
Experimental Calculated Error^2
Germ Aleurone Pericarp EndospermMallacca D-D 0.5 mm
2000 microms
Peaks (wavenumbers) height (A) height (A) height (A) height (A) height (A) height (A)
A (3300.48 cm-1) 0.044134 0.045645 0.024834 0.038762 0.046 0.040 3.82E-05
B (2928.98 cm-1) 0.05918 0.01946 0.0051657 0.00052528 0.023 0.012 1.37E-04
C (2857.84 cm-1) 0.03396 0.0055848 0 0 0.015 0.006 8.21E-05
D (1743.35 cm-1) 0.050975 0.0063914 0 0 0.008 0.010 2.44E-06
E (1644.55 cm-1) 0.13769 0.094375 0.041822 0.033213 0.047 0.053 2.92E-05
F (1541.80 cm-1) 0.088039 0.051423 0.019533 0.012456 0.029 0.027 6.98E-06
G (1249.34 cm-1) 0.078436 0.05103 0.040675 0.024167 0.032 0.034 5.41E-06
H (1035.93 cm-1) 0.18865 0.2288 0.19292 0.1812 0.121 0.183 3.75E-03
4.05E-03
5.3 Table 5.3 Example of height data introduced in the spreadsheet to quantify the relative proportion
of each botanical component in the unknown sample.
Chapter 5
138
The sample tested here was the 2000 µm samples from Mallacca wheat milled under D-D
disposition with a roll gap of 0.5 mm. The result obtained indicates that this sample is only
81% endosperm and 19% germ, no contribution of aleurone and pericarp was obtained,
which seems to be an infeasible result since this sample, as was described above, should be
rich in pericarp and aleurone.
The other tested sample (data not shown) was the <212 µm sample from Mallacca wheat
milled under D-D disposition with a roll gap of 0.5 mm, which should be rich in
endosperm. The resulting calculation indicated that this sample was made of 43% germ
and 57% endosperm, and no contribution of aleurone and pericarp was found. Although
this result may be slightly feasible, the first one was not, thus, the calculation was not
considered plausible at this stage.
More trials were done with other samples and using the whole spectra instead of peaks,
however, similar unsatisfactory results were obtained.
Based on the work of Barron (2011), Partial Least Square (PLS) analysis was performed to
quantify the proportion of botanical tissues within milled fractions. However, due to the
insufficient quantity of hand dissected material required to build up a representative
calibration curve, along with the absence of facilities to perform wet chemistry analysis (i.e
biochemical markers for each wheat tissue) that were able to back up the FTIR analysis,
the PLS performed was insufficient to quantify accurately the proportion of the four major
wheat components dissected in the milled fractions. More information about PLS, its
meaning and importance, are further discussed in the following chapter. Similarly, the
bases of PLS are found in the Appendix 5.
Although FTIR and PCA analysis were useful to corroborate some insights about wheat
botanical tissues obtained with the hand dissection technique, the use of the PLS technique
in particular was not successful to quantify the botanical components for the issues
described earlier.
Another approach was made based on sugar profiles and HPLC analysis. For this, only two
wheat component were selected for analysis: endosperm and bran (in this work, pericarp is
treated as the bran layer as it is easier to isolate and analyse, although it is known, as
describe in Chapter 2, that bran is composed of several layers besides pericarp). These
components were considered to contain the most representative and easiest sugars to be
Chapter 5
139
quantified by HPLC, glucose (for endosperm) and arabinose and xylose (for bran). The
following section describes the preparation of the standard solutions for the calibration
curve, the hydrolysis method followed and the HPLC analysis performed, and presents the
results of this alternative profiling approach.
5.9 Sugar quantification by High Performance Liquid Chromatography
High Performance Liquid Chromatography (HPLC) is a powerful technique widely used to
separate, identify and quantify the components in a mixture. This technique involves
pumping pressurized liquid solvent (mobile phase) through a column packed with solid
material. The solvent contains the sample of interest. Each compound of the sample has a
particular interaction with the stationary phase depending on its nature; this results in
different flow rates for each component and hence separation as they flow out the column.
Retention time is defined as the time taken by the solute to pass through the column,
measured from the moment the sample is injected to the mobile phase until the moment the
peak maximum is displayed for a particular compound (www.chem.agilent.com; Fanali et
al., 2013).
Hand-dissected endosperm and pericarp tissues and two milled fractions (<212 and 2000
µm from Mallacca wheat milled under D-D disposition with a roll gap of 0.5mm) were
hydrolyzed and analyzed with HPLC according to the procedures reported by Du et al.
(2009) and Salmeron-Ochoa (2010).
5.9.1 Stock and standard solutions
Pure glucose, arabinose and xylose from a Carbohydrates kit from Sigma-Aldrich (Dorset,
UK) were weighed out individually in 25 mL volumetric flasks and diluted with ultrapure
water to produce concentrations of 20.0, 10.0 and 30.0 g/L respectively. Five additional 25
mL volumetric flasks were used to prepare the diluted solutions used in the calibration
curve. Stock solutions were prepared in duplicate.
Chapter 5
140
5.9.2 Sample preparation for chemical hydrolysis.
1. Approximately 100 mg of ground sample (botanical tissues and milled fractions)
were weighed out and placed in labelled glass tubes.
2. 2 mL of H2SO4 (1 M) was added to the samples; tubes were closed and placed in an
oven set at 121oC for 1 hour.
3. After 1 hour, the glass tubes were removed from the oven and left to cool down to
room temperature.
4. Samples were neutralized (pH 7) with NaOH (1 M). 3.5 ml NaOH (1 M) was added
to each tube and the pH tested. In general, all the samples initially showed a very
acid pH (around 2), thus, NaOH (1 M) was further added drop by drop until the pH
neutralized. Depending on the sample the quantity of drops varied. The total
amount of NaOH was recorded; as a reference, 1 drop = 0.05 mL.
5. Neutralized samples were filtered through Whatman filter paper Grade 4 (Sigma-
Aldrich) into clean glass tubes.
6. Approximately 2 mL of filtered sample was filtered through Sartorius Minisart 0.45
µm membrane filter (Fisher, UK) into HPLC vials.
7. Vials were capped, labelled and stored in a fridge at 4oC until needed. The
following day, samples were run through the HPLC.
5.9.3 HPLC analysis
Figure 5.17 shows the HPLC equipment used for the sugar quantification. The analysis
was performed using a Varian ProStar HPLC system (Varian Inc., UK) with a Hi-Plex Ca
8 µm column (300 x 7.7 mm, Agilent Technologies, UK) preceded by a guard column PL
Hi-Plex Ca (5 x 3 mm, Agilent Technologies, UK). The column was maintained at 85oC in
an oven (Shimadzu CTO-6A). The mobile phase used was ultrapure water (filtered through
a 0.45 µm acetate membrane filter) operated at a flow rate of 0.6 mL/min. Sugars were
detected by an evaporative light scattering detector (ELSD) (PL-ELS 2100), (Polymer
Laboratories, UK) with the following settings: evaporator 90oC, nebulizer 35
oC and gas
flow rate (N2) 1.6 mL/min. The program used to process and calculate the peak areas was
Interactive Graphics (Version 6.20, Varian Inc., UK). The three sugars were identified on
the basis of their retention times, and their concentrations were determined by comparison
with their corresponding calibration curve.
Chapter 5
141
Figure 517 HPLC equipment used.
Figure 5.18 shows the HPLC chromatograms of pericarp (top) and endosperm (bottom),
indicating their corresponding sugar peaks detected. As expected, the pericarp layer
contains the major non-cellulosic polysaccharides, arabinoxylans; thus after hydrolysis,
xylose (retention time around 12 minutes) and arabinose (retention time around 14
minutes) were released. Conversely, endosperm, although it also contains arabinoxylans
(Barron et al., 2007), is richer in starch, and after the hydrolysis procedure, glucose
(retention time around 11 minutes) is mainly liberated. In all the HPLC chromatograms
obtained it was observed a wide peak that came out at first, which may correspond to a
mixture of the salt formed (NaCl) after the reaction of HCl (used for the chemical
hydrolysis) and NaOH (used for neutralize the hydrolyzed samples), or maybe ferulic acid,
and other unidentified components.
Column oven
ELS
D
Autosampler
Reservoir
of mobile
phase
UV
detector
Solvent
pumps
Figure 5.17 HPLC equipment used.
Chapter 5
142
Figure 52 HPLC chromatograms of pericarp (A) and endosperm (B) after being hydrolyzed with
H2SO4 and neutralized with NaOH.
Figure 5.19 shows the HPLC chromatograms of the two milled samples: (A) Mallacca,
<212 µm milled with a roll gap of 0.5 mm and under D-D disposition, and (B) Mallacca,
2000 µm milled with a roll gap of 0.5 mm and under D-D disposition, and their
corresponding sugars identified in both samples. For the sample corresponding to <212 µm
(Figure 5.19, top), glucose dominates over arabinose and xylose, which is expected
considering that fine particles are rich in endosperm. Although it is known that
arabinoxylans are not exclusively present in peripheral layers of the wheat grain, these are
also present in cell wall of endosperm (Ordaz-Ortiz and Saulnier, 2005; Barron et al.,
2005; Barron et al., 2007), the amount contained in the endosperm cell-wall is much lower
compared with the higher amount of starch present in the endosperm (Barron et al., 2007);
thus, in the <212 µm sample, which is rich in endosperm particles, the content of arabinose
and xylose is diluted by the amount of glucose contained in the sample.
By contrast, as observed in Table 5.4, the sample corresponding to 2000 µm shows a
higher contribution of arabinose and xylose than the other milled fraction; however glucose
is still the predominant sugar. Coarse particles are rich in large bran particles and some mid
range endosperm particles thus, the contribution of the arabinoxylans given by the bran
particles is more evident in this sample, which is reflected in the amount of arabinose and
xylose quantified.
Endosperm
Xyl Ara
Glc
Pericarp A
B
5.18 Figure 5.18 HPLC chromatograms of pericarp (A) and endosperm (B) after being hydrolyzed with
H2SO4 and neutralized with NaOH.
Chapter 5
143
Figure 53 HPLC chromatograms of the sugars present in the two samples of Mallacca wheat milled
under D-D disposition with a roll gap of 0.5mm, after being hydrolyzed with H2SO4 and
neutralized with NaOH.
Table 54 Sugars composition (% in 100 mg of sample) of two wheat grain tissues, two milled
fractions from Mallacca wheat and the results reported by Barron et al. (2007) for starchy
endosperm and outer pericarp from Caphorn and Crousty wheat varieties.
Sample Wheat
variety
Arabinose
(%)
Xylose
(%)
AX
(A+X)(0.88)
(%)
Glucose
(%)
Total of
the three
sugars
released
(%)
Endosperm Mallacca 0 0 0 70.8 70.8
Pericarp Mallacca 12.8 15.3 24.7 2.8 27.5
Starchy
endosperm
(Barron et
al., 2007)
Caphorn 0.8
0.7
1.0
0.8
1.7
1.5
74.4
78.1
76.1
79.6 Crousty
2000 µm
D-D 0.5mm
Mallacca 1.6 1.4 2.6 55.3 57.9
<212 µm
D-D 0.5mm
Mallacca 0.7 0.2 0.8 66.9 67.7
Outer
pericarp
(Barron et
al., 2007)
Caphorn 25.4
24.1
21.5
21.3
46.9
45.4
1.6
1.4
48.5
46.8
Crousty
Mallacca, <212 µm D-D 0.5mm
Mallacca, 2000 µm D-D 0.5mm
Glc
Glc
Ara Xyl
Ara Xyl
A
B
Figure 5.19 HPLC chromatograms of the sugars present in the two samples of Mallacca wheat
milled under D-D disposition with a roll gap of 0.5 mm, after being hydrolyzed with H2SO4 and
neutralized with NaOH.
Table 5.4 Sugar composition (% in 100 mg of sample) of two wheat grain tissues, two milled
fractions from Mallacca wheat and the results reported by Barron et al. (2007) for starchy
endosperm and outer pericarp from Caphorn and Crousty wheat varieties.
Chapter 5
144
Table 5.4 shows the percentage of Glucose, Arabinose and Xylose contained in wheat
tissues and milled fractions. The results of the current work are compared with the results
of Barron et al. (2007). In that work, hard and soft wheat varieties (Caphorn and Crousty)
were hydrolyzed with H2SO4 (1M, 2h, 100 oC) and the amount of neutral sugars was
quantified by determining the anhydro-sugars by gas-liquid chromatography of their alditol
acetates. It is important to be cautious when comparing results from different wheat
varieties, hydrolysed under different conditions and analyzed with a different analytical
technique; however, the results presented by Barron et al. (2007) represent the most
comparable reference available.
The amount of glucose in the endosperm is slightly lower compared with the values
reported by Barron et al. (2007). By contrast, in their results, the amount of arabinose and
xylose was quantified, and for the current work, as was described earlier, it was not
possible (fine particles are rich in endosperm; hence glucose dominates over arabinose and
xylose). The amount of arabinoxylans (AX) released by the pericarp, 28%, was lower
compared with the results of Barron et al. (2007), who reported that the outer pericarp
contains around 45% of AX as noted above. The amount of sugars released by the milled
fractions seems to be plausible although there are not direct sources to compare them with.
Having a look again at the low amount of AX released in the pericarp layer dissected in the
current work in comparison with only the outer pericarp reported by Barron et al. (2007), it
was considered that perhaps the hydrolysis method performed in the current work was not
releasing all the AX content in the pericarp. Therefore, a hydrolysis trial was performed
using commercial wheat bran (Jordans ®
) and pure AX standards (high, medium and low
viscosities) (Megazyme, Ireland). Wheat bran and both sugars were hydrolysed according
to the procedure described in Section 5.9.2. For wheat bran 100 mg was hydrolysed, but
for both sugars instead of 100 mg, 20 mg samples were used (due to the purity of the
standards). Samples were hydrolysed at eight different times (15, 30, 45, 60, 75, 90, 105
and 120 mins). All analyses were performed in triplicate. The results obtained in the
hydrolysis trial were performed in collaboration with Ruth Bell (a fellow PhD student from
the Satake Centre); hence the results are now explained here. Briefly under the conditions
described in section 5.9.2 and varying the hydrolysis time, a dramatic decay in the
Arabinose and Xylose content after 60 mins was observed in the wheat bran, thus,
suggesting that the highest amount of sugars were released after this time, and only very
little amount of Arabinose and Xylose was released after 60 mins. Related to the pure AX
Chapter 5
145
standards, the maximum yield of Arabinose and Xylose released was obtained after only
15 mins of hydrolysis. The suggested explanation here is that for pure samples (like AX
standards) the hydrolysis quickly releases the sugars available because there is nothing else
that can block or delay the release of the sugars. The opposite effect is found in complex
samples (like wheat bran) where there can find a different type of components that are
cross-linked such as proteins, phospholipids, lignin, cellulose, etc., that can block and/or
delay the complete release of the sugars (Stone and Morell., 2009). Based on the results
obtained after the hydrolysis trial, it was decided that 1 h of hydrolysis (the actual time
preformed) was enough to release the maximum amount of sugars in the botanical tissues
dissected and milled fractions.
Although it was shown that the hydrolysis conditions performed in the current work were
sufficient to quantify the amount of sugars in the samples that could enable building up a
sugar profile for each sample, only the samples shown in Table 5.4 were analyzed due to
the complexity of the process and the time required for the analysis. In principle it could be
applied but unfortunately the time required for each sample is considerable, and the
amount of sugars quantified is limited by the detection limit of the HPLC, in particular for
arabinose and xylose contained in samples particularly rich in endosperm particles; thus, it
was decided to explore a more feasible and faster option.
As was described in Chapter 3, Barron (2011) developed a technique to predict the relative
tissue proportions in wheat mill streams based on FTIR spectroscopy and PLS models. Due
to the good quality of prediction obtained in the samples studied, it was decided to contact
this research group based on INRA, UMR, Montpellier, France to work together to finally
extend the breakage equation to include botanical tissues. The work performed and the
results obtained are discussed in the following chapter.
Chapter 5
146
5.10 Summary
The four major botanical components of the wheat kernel, pericarp, aleurone, endosperm
and germ, were hand isolated from representative hard and soft wheat varieties and dried to
<16% moisture content. The dissected tissues were checked under the microscope and via
FTIR analysis, which showed effective separation of the tissues, but with the aleurone
layer appearing to be slightly contaminated with endosperm. The main lipid, protein and
carbohydrate regions of the spectra were clearly distinguished, and the peak intensities
reflected the known compositions of the different tissues in wheat. Endosperm had very
strong starch adsorption at 1000 and 1022 cm–1
, and in the protein region, showing that
Endosperm is rich in starch and proteins. The aleurone layer had both starch and
arabinoxylan absorption peaks in the carbohydrate region of its spectra, but unlike
endosperm aleurone has considerable absorption of protein and some lipid. The germ had
only a small shoulder for starch, and its carbohydrate region was dominated by
arabinoxylan (absorbance at 1041 with a shoulder at 1075cm–1
). Germ is rich in lipid and
protein. The pericarp carbohydrate region is dominated by cell wall polysaccharides,
predominantly arabinoxylan, β-glucans and cellulose. The pericarp is essentially free of
protein and lipid.
Principal component analysis was applied to the spectra of the botanical components, using
different mathematical pre-treatments before performing PCA. The best separation of the
four botanical tissues in PCA was observed after base line correction and second
derivative. The Bi-plot of scores and loadings showed the wavenumbers that highly
contribute to the chemical structures present in the samples, enhancing the qualitative
analysis. Results of the botanical tissues analysis suggested that some regions in the spectra
were more intense for each wheat component, identifying specific wavenumbers that could
be used for identification and quantification of botanical components in milled fractions.
In milled samples PCA was also performed. The largest variance of the data observed
separated the fine particles (<212, 212 and 500 µm) from the coarse particles (850, 1180,
1400, 1700, 2000 µm) along PC1. The loading associated with PC1, showed the strongest
absorbance in the carbohydrate region, mainly at 1065 cm–1
and 1075 cm–1
, which is
related to cell wall polysaccharides associated with the pericarp tissue. Interestingly, the
milled samples 850, 1400 and 1700 µm were closely clustered, suggesting that these
samples are similar in composition.
Chapter 5
147
An attempt to quantify the botanical tissues in milled fractions was carried out by selecting
eight specific peaks observed in the spectrum of all the samples (botanical tissues and
milled fractions), but in different proportions, and measuring the heights of each peak in
each sample. The relative contribution of each wheat component in milled fraction samples
was calculated. Only few samples were tested, however, implausible results were obtained.
More trials were performed using the whole spectra instead of specific peaks, but similarly
unsatisfactory results were obtained. Although there was an attempt to apply the PLS
analysis, the required quantity of samples for building up a proper calibration curve was
not enough, thus, this statistical approach was abandoned.
An alternative approach was conducted with HPLC and sugar analysis. Good results were
obtained, but the technique was to complex and limited by the detection limit of HPLC, in
particular for arabinose and xylose content in samples rich in endosperm particles.
In order to obtain higher quality composition data, a collaboration with INRA, UMR,
Montpellier, France was instigated. From Barron (2011) it was known that this research
group investigated a technique to predict the relative tissue proportions in wheat mill
fractions based on FTIR spectroscopy and PLS models. Good prediction was obtained in
their study; thus it was decided to contact this research group to work together to finally
extend the breakage equation to include botanical tissues. The work performed and the
results obtained are presented in the following chapter.
Chapter 6
148
CHAPTER 6
DEVELOPING A COMPOSITIONAL BREAKAGE EQUATION
FOR WHEAT MILLING
6.1 Introduction
The complete understanding of wheat milling requires the prediction of the size
distribution of broken particles and also the composition of particles of different sizes. In
the present chapter, the compositional breakage equation is derived, and the forms of the
compositional breakage functions are determined using experimental data for pericarp,
intermediate layer, aleurone and starchy endosperm generated from spectroscopic analysis
of milled fractions. As was described in Chapter 3, one of the objectives of the current
work is to represent the continuous equivalent of the discrete compositional breakage
matrices introduced by Fistes and Tanovic (2006), since continuous functions are more
generally applicable and more readily interpretable, thus yielding greater predictive power
and greater mechanistic insights regarding wheat breakage.
6.2 Experiment procedures
Whole wheat kernels and milled fractions were sent to Dr Cécile Barron at INRA,
Montpellier, France for spectroscopic analysis. The botanical tissues used were isolated
according to the method described by Hemery et al. (2009). For the purpose of
demonstrating the approach (as Dr Barron’s schedule allowed only a limited number of
samples to be analysed), it was decided to analyse the milled fractions from Mallacca
wheat (SKCS hardness = 52.52, diameter = 3.25 mm) and Consort wheat (SKCS hardness
= 33.87, diameter = 2.89 mm) milled under Sharp-to-Sharp (S-S) and Dull-to-Dull (D-D)
dispositions at roll gap of 0.5 mm. From the batch previously described in Section 5.3.3,
Table 5.1, eight size fractions were obtained: 2000, 1700, 1400, 1180, 850, 500, 212 and
212 µm. In total 34 samples were analyzed: two wheat types × two dispositions × one roll
gap × eight fractions = 32, plus the two whole wheats = 34.
Chapter 6
149
6.2.1 Sample preparation for FTIR analysis
The protocol performed at INRA was based on the method described by Hemery et al.
(2009) and Barron (2011): milled fractions were first ground in liquid nitrogen with a Spex
CertiPrep 6750 laboratory impact grinder to have a homogenous size. Spectra were
recorded in the MIR region using a Nicolet Nexus 6700 (ThermoScientific, Courtaboeuf,
France) spectrometer equipped with an ATR Smart DuraSampleIR accessory
(ThermoScientific, U.K.) and a Mercury Cadmium-Telluride-High D detector. Spectra
were recorded between 800 and 4000 cm-1
. Samples were pressed onto the diamond ATR
area. For each sample five spectra were collected. An air-background scan was recorded
every three spectra. Partial Least Square (PLS) quantification was applied using models
developed by Barron (2011). (Appendix 5 presents a brief description of the basic concepts
of PLS.) Similar spectral pre-treatments were then applied to predict each tissue
proportion. Pericarp, intermediate layer, aleurone and starchy endosperm were predicted in
each milled fraction, and the results interpreted through the compositional breakage
equation.
The following section is adapted from the mathematical modelling proposed by
Choomjaihan (2008) for extending the breakage equation to include composition.
6.3 Derivation of a compositional breakage equation
As described in Chapter 3, the cumulative form of the breakage equation for roller miller
is:
dDDDxBxP
D
D
)(),( 1
0
2 (6.1)
This equation describes the correlation between the input and output particle distributions,
where D is the input particle size, x is the output particle size, P2(x) is the proportion by
mass of output material smaller than size x, B(x, D) is the breakage function and ρ1(D) is
the probability density function describing the input particle size distribution. The logic of
the breakage equation is that the total mass of particles smaller than a given size x arises
from contributions from all the inlet particles. The contribution from inlet particles initially
of size D depends on how many of those particles there are (which is quantified by ρ1(D))
Chapter 6
150
and on how those particles break (which is quantified by the breakage function, B(x,D).
The total mass smaller than x is found by integrating all of these contributions over the
range of inlet particle sizes.
Applying equivalent logic, the composition of particles can also be described and related to
the particle size distribution. As described in Chapter 3, Choomjaihan (2008) took an
approach to quantify the botanical tissues in milled fractions by relating the characteristic
mineral profiles of the individual grain components. The initial proposal made by
Choomjaihan (2008) was that the whole wheat kernel and its milled fractions are made up
of four main components: pericarp (including testa and nucellar tissues), aleurone,
endosperm and germ. The mass proportions of these components are represented by Xpe,
Xal, Xen and Xge respectively:
masstotal
component botanical theofmass total iiX (6.2)
where i represents the different botanical components (pericarp, aleurone, endosperm, and
germ).
Note the sum of the four components is equal to one:
1 geenalpe
i
i XXXXX (6.3)
Typical values of these concentrations are Xpe = 0.08, Xal = 0.07, Xen = 0.82, and Xge = 0.03
(Pomeranz, 1988).
On breakage, the particles released contain different proportions of pericarp, aleurone,
endosperm and germ. The particles in a size range 212-500 µm for example, have a
proportion of each component that will be different from particles from a different size
range, say 1750-2000 µm; smaller particles may be richer in endosperm tissue than larger
particles; conversely, larger particles may contain more bran material (i.e pericarp and
aleurone tissues) than endosperm.
As described above, P2(x) is the total proportion of the output particles that are smaller
than size x. Such particles, as a whole, are made up of pericarp, aleurone, endosperm and
germ. P2(x) may be calculated as shown in Equation 6.4.
Chapter 6
151
)(·)(·)(·)(·
)(·
)(masstotal
ansmaller th particles ofmass total2
xYXxYXxYXxYX
xYX
xP
gegeenenalalpepe
i
ii
x
(6.4)
where Yi(x) is the proportion (by mass) of the corresponding component that is in particles
smaller than size x.
i
xixYi
component botanical theof masstotal
ansmaller th particles in component botanical theofmass)( (6.5)
Figure 6.1 exemplifies how the distributions of the four components sum to give the total
PSD, while Figure 6.2 shows the distributions in their non-cumulative forms (where i(x)
is the differential of Yi(x), as defined later). Note that both are contrived examples to
illustrate the relationships, not realistic representations of the proportion of the four
components in real wheat particles.
Figure 55 Contrived example that shows how the cumulative PSD is comprised of the cumulative
distributions of the four botanical components in particles of different sizes. Adapted from
Choomjaihan, 2008.
0
10
20
30
40
50
60
70
80
90
100
0 1000 2000 3000 4000
Xi Y
i (x
)
Particle size x (m)
Pericarp
Aleurone
Endosperm
Germ
Total
B
A C
Xen
Xpe
Xal
Xge
6.1 Figure 6.1 Contrived example that shows how the cumulative PSD is comprised of the cumulative
distributions of the four botanical components in particles of different sizes. Adapted from
Choomjaihan, 2008.
Chapter 6
152
Figure 56.2 Non-cumulative form of the contrived example of Figure 6.1, displaying how particles
of different size are made up of different compositions. Adapted from Choomjaihan, 2008.
On a second contrived but more realistic example, not related to the plots above, let’s
consider the typical values of the concentrations reported by Pomeranz (1988), Xpe = 0.08,
Xal = 0.07, Xen = 0.82, Xge = 0.03; when the wheat is milled, the resulting particles are
distributed in size from 0 up to 4000 µm, with most of the particles at the smaller end of
the range. Let´s consider only the particles that are smaller than 500 µm; let´s imagine that
after breaking up the wheat kernels, 40% of the total pericarp has ended up in the particles
smaller than 500 µm; the other 60% is in particles that have remained larger than 500 µm.
However, the aleurone has not broken so easily, thus only 30% of the total aleurone has
ended up in the particles smaller than 500 µm; meanwhile 70% of the aleurone has stayed
in the larger particles. The endosperm has broken readily; hence 80% of the endosperm is
now in small particles, with only 20% in large particles. Meanwhile the germ is equally
split; thus half of the germ material is in particles that are smaller than 500 µm.
Hence:
50.0)500(,80.0)500(,30.0)500(,40.0)500( geenalpe YYYY (6.6)
Thus, the total proportion of particles smaller than 500 µm is calculated as:
724.0
)015.0()656.0()021.0()032.0(
)50.0·(03.0)80.0·(82.0)30.0·(07.0)40.0·(08.0)(2
xP
0.0000
0.0001
0.0002
0.0003
0.0004
0.0005
0.0006
0.0007
0.0008
0 1000 2000 3000 4000
i (
x)
Particle size x (m)
Pericarp
Aleurone
Endosperm
Germ
Total
dx
6.2 Figure 6.2 Non-cumulative form of the contrived example of Figure 6.1, displaying how particles
of different size are made up of different compositions. Adapted from Choomhaijan (2008).
Chapter 6
153
i.e about 72% of particles are smaller than 500 µm. Considering these particles as a whole,
their composition is (0.032/0.724)·100=4.4% pericarp, 3.0% aleurone, 90.6% endosperm
and 2.0% germ, i.e they are rich in endosperm, and depleted in the other components,
compared with the material as a whole.
Although this is a contrived example in order to exemplify the mathematics, it reflects the
well-known behaviour of wheat during roller mill breakage, where the bran (pericarp and
aleurone) is likely to form large particles, while endosperm tends to break more easily,
producing small particles. This is the basis for producing relatively pure white flour, the
separation of bran from endosperm based on their different sizes using repeated milling
and sifting operations. As in the first contrived example in Figures 6.1 and 6.2, it would be
expected to get smaller particles rich in endosperm material, compared with the endosperm
content of the whole wheat.
For the botanical component i, Y*i(x) is the concentration of component i in particles
smaller than x, which is equal to the ratio between the total mass of i in particles smaller
than size x, and the total mass of particles smaller than size x. The latter is the sum of the
individual components, thus:
)(·)(·)(·)(·
)(·
)(
)(·
an smaller th particlesin masstotal
an smaller th particlesin componentofmass)(
2
*
xYXxYXxYXxYX
xYX
xP
xYX
x
xixY
gegeenenalalpepe
ii
ii
i
(6.7)
and equally for the concentrations of the other components, defined as Y*pe(x), Y*al(x) and
Y*ge(x). (Note that this is different to Yi(x) defined in Eqn. 6.5, where the denominator is
the total mass of component i in all the particles, not the total mass of all components just
in particles smaller than x.) Similarly to Xi, the sum of all Y*i concentration gives 1:
1)()()()()( ***** xYxYxYxYxY geenalpe
i
i (6.8)
As an example, referring to Figure 6.1, Ype(x) is defined by the point A divided by the point
C (the mass of pericarp in particles smaller than x divided by the total mass of pericarp),
while Ype*(x) is defined by the point A divided by the point B (the mass of pericarp in
Chapter 6
154
particles smaller than x divided by the total mass of particles smaller than x, i.e the average
concentration of pericarp in particles smaller than x).
As was described above, Yi(x) is the proportion of component i in particles smaller than
size x, while the proportion of component i in particles of size x is represented by the
probability density function (ρi (x)), which is the non-cumulative form or differential of
Yi(x):
dx
xdYx i
i
)()( (6.9)
Hence, the ratio between the mass of the botanical component i in particles in the size
range x and x + dx, and the total mass of the component i is equal to:
dxxii
dxxxi)·(
of masstotal
,rangesizetheincomponentofmass
(6.10)
while the concentration of the botanical component i in the size range x and x + dx, in the
total mass, is equal to:
)(·)·()·(· 2masstotal
,rangesizetheincomponentofmassxydxxdxxX iii
dxxxi
(6.11)
where (x) is the derivative of the total proportion of particles smaller than x (or the
probability density function that describes the outlet PSD), and yi (x) is the ratio between
the mass of the botanical component i in particles in the size range x and x + dx and the
total mass in particles in the same range (i.e the concentration of component i); then:
dx
xdPx
)()( 2
2 (6.12)
dxxx
dxxxixyi
,rangesizethein masstotal
,rangesizetheincomponentofmass)( (6.13)
Similarly to Yi*, the sum of all yi concentrations gives 1:
1)()()()()( xyxyxyxyxy geenalpe
i
i (6.14)
Chapter 6
155
Equation 6.13 calculates the concentration of component i in particles of size x (note, this
is not the same as the average concentration in particles smaller than x, as defined by
Y*i(x) in Eqn. 6.7). Adding the four botanical components gives:
dxxxydxxdxxXi
i
i
ii )·()(·)·()·(· 22 (6.15)
Dividing by dx, gives:
)()(· 2 xxXi
ii (6.16)
Yi (x) can be calculated as the integral between 0 and x of the probability density function
as shown in Equation 6.17:
x
ii dxxxY0
)·()( (6.17)
The breakage equation is described by Equation 6.1. If the value of D does not vary
dramatically in wheat kernel size, then the breakage is described by P2(x) = B(x,D) or,
more generally, by B(x,G/D), which is defined as the proportion of particles smaller than
size x arising from breakage of wheat at a particular milling ratio G/D, where G is the roll
gap. The functions ype(x), yal(x), yen(x) and yge(x) equally become ype(x,G/D), yal(x,G/D),
yen(x,G/D) and yge(x,G/D). The yi(x,G/D) is defined as the proportion of botanical
component i in particles of size x produced from milling wheat at a milling ratio G/D. If
yi(x,G/D) is known, then both the size distribution of particles following breakage and their
composition can be predicted. Thus the compositional breakage equation is:
i
x
i
i i
x
iiii
dxDGxyDGx
dxDGxXDGxYXDGxP
0
2
0
2
)·/,()·/,(
)·/,(·)/,(·)/,(
(6.18)
Chapter 6
156
and in its non-cumulative form:
i
i
i
ii
DGxyDGx
DGxXDGx
)/,()·/,(
)/,(·)/,(
2
2
(6.19)
Equations 6.18 and 6.19 enable the calculation from a single equation of both the PSD, and
the composition of each size fraction. This simplifies the problem to establishing
“concentration functions” to describe ype(x,G/D), yal(x,G/D), yen(x,G/D) and yge(x,G/D),
leading to “compositional breakage functions” that describe pe(x,G/D), al(x,G/D),
en(x,G/D) and ge(x,G/D). In principle, this could be done by roller milling wheat using
different roll gaps, then sifting the particles obtained to get different size fractions and
finally measuring the composition of those size fractions, i.e the concentrations of pericarp,
aleurone, endosperm and germ in each milled fraction. If it is known how these
concentrations vary within size fractions, then in theory curves could be fitted to the
experimental data to describe these variations as functions of x and G/D. Eventually, with a
very large experimental programme, these compositional breakage functions could be
extended to include hardness, as Campbell et al. (2007) did for the size-based breakage
function. These aims were beyond the scope of the current work.
Equations 6.18 and 6.19 represent the continuous equivalent of the discrete compositional
breakage matrices introduced by Fistes and Tanovic (2006). The equations presented here
are continuous functions that are more generally applicable and more readily interpretable.
6.4 Compositional breakage functions
Having derived the compositional breakage equation above, the first objective of the
current work, the second objective is to begin to understand the form of the compositional
breakage functions by generating experimental data. In principle, this is as simple as
measuring the concentrations of pericarp, aleurone, endosperm and germ in size fractions
following milling. However, there are two difficulties with this. Firstly, these
concentration functions are not probability density functions and hence don’t have the well
defined constraints of probability density functions that allow easy fitting. Secondly,
measuring the proportions of these materials in milled wheat samples is not
straightforward.
Chapter 6
157
Addressing the first issue, Equation (6.11) can be rearranged to give:
)(
)()(
2 x
xXxy ii
i
(6.20)
where
dx
xdPx
)()( 2
2 (6.21)
and ρi(x) is similarly the derivative of Yi(x) as previously defined in Equation 6.17. As
described in Chapter 3, Campbell et al. (2012) introduced the Double Normalised
Kumaraswamy Breakage function (DNKBF) as a flexible probability density function to
describe the PSD produced from roller milling of wheat; it has a cumulative form that is
simple to fit and is differentiable. Considering that the DNKBF could have the flexibility
to describe Yi(x) too, from which ρi(x) could be obtained by differentiation, then Equation
6.20 enables yi(x), the concentration of component i in particles of size x, to be calculated
as the ratio of these two probability density functions. This approach, that considers the
fitting of a simple function to the cumulative data, is likely to deal better with inaccuracies
in the experimental data, than trying to fit the concentration data directly. Another
advantage of this approach is that may yield more meaningful descriptions of the
compositional breakage functions.
For the second issue, the measuring of the botanical components can be done in principle
with suitable biochemical markers that are specific for each botanical tissue (Barron et al.,
2007; Barron and Rouau, 2008; Hemery et al., 2009; Barron et al., 2011). However, as
described in Chapter 3, Barron (2011) predicted the relative tissue proportion in wheat mill
streams by FTIR spectroscopy and PLS analysis. As a reminder, in the study of Barron
(2011), aleurone layer, intermediate layer, outer pericarp and starchy endosperm were
isolated as in previous works from the same author from various common wheat cultivars.
Different milled streams, ranging from debranning, conventional milling and bran
fractionation were produced from two French wheat varieties. The spectra of botanical
Chapter 6
158
tissues and milled fractions were collected with a FTIR coupled with an ATR device. The
biochemical markers technique studied by the same author was used as the reference
method (Barron et al., 2007; Hemery et al., 2009; Barron et al., 2011). PLS models were
developed to predict the proportion of the botanical tissues in the milled streams. The
predictions obtained were good despite the complex natures and compositions of botanical
tissues.
As was described in Sections 6.2 and 6.3, based on the spectroscopy-based models
developed by Barron (2011), for the eight milled fractions (2000, 1700, 1400, 1180, 850,
500, 212 and <212 µm) of Mallacca and Consort wheats milled under S-S and D-D
dispositions at 0.5 mm roll gap, the proportions of pericarp, intermediate layer, aleurone
and starchy endosperm in each fraction were estimated.
Figure 6.3 shows a schematic representation of a broken wheat particle including the four
layers described by Barron (2011) and used in the current study: pericarp, intermediate
layer, aleurone and starchy endosperm. Barron concluded that on average the ratio of
aleurone to intermediate layer to pericarp was roughly 50:25:25, and that together these
layers made up around 16% of the total kernel (Barron, 2011; Barron et al., 2011); Figure
6.3 illustrates these relative proportions (with the proportion of starchy endosperm not
shown to this scale in the figure, as the starch is unlikely to be present in this natural
proportion in this type of broken particle, as other particles are pure endosperm). The
horizontal scale is also not in proportion; a flattened bran particle from which endosperm is
scraped, as described in Chapter 4, would be much longer relatively to its thickness than
suggested by the figure.
Chapter 6
159
Figure 6.3 Broken wheat particle. Note: Intermediate layer is composed of three layers: hyaline
layer, testa and inner pericarp (Barron et al., 2007; Barron, 2011).
At this point, it is necessary to be cautious with certain arguments. The correlations used in
the models are based on French wheats, such that the absolute results generated for these
UK samples are unlikely to be accurate. However, the relative values obtained are probably
sufficiently good to allow this approach to be demonstrated and to give valid insights.
Meanwhile, the models of Barron (2011) do not quantify Germ, and they differentiate
between the outer pericarp and the intermediate layer; the derivation above lumped these
layers together as pericarp. The information that the models provide is consequently not
exactly in the form of the derivations described above, in particular, not intending to
provide exclusive proportions of components that sum to one (Barron, 2011). The values
obtained for pericarp, for example, should be considered to show how the pericarp
concentration changes with particle size, but these values and the corresponding values for
intermediate layer, aleurone and endosperm are not expected to sum to unity. Hence, the
data can be used along with Equation 6.20 to find the form of the compositional breakage
functions, but not their absolute values; these functions could be used to inform Equation
6.20, the compositional breakage equation, but with some loss of accuracy. Knowing the
general forms would, however, simplify an experimental programme aimed at quantifying
these functions more extensively. However, acknowledging the slight inaccuracies, the
rest of this chapter describes and interprets the compositional data in terms of the
compositional breakage functions defined above.
Starchy Endosperm
Aleurone
Pericarp
Intermediate Layer
84%
8% 4% 4%
Figure 6.3 Broken wheat particle. Note: Intermediate layer is composed of three layers: hyaline
layer, testa and inner pericarp (Barron et al., 2007; Barron, 2011).
Chapter 6
160
The next section presents and explains the results obtained from the PLS modelling and the
fitting of the DNKBF to the experimental data.
6.5 Finding the concentration functions using the Double Kumaraswamy
Functions fitted to the PSD and to the compositional distributions
Table 6.1 shows the proportion of milled material on each sieve size after milling Mallacca
and Consort wheats under S-S and D-D dispositions, and the percentages of pericarp,
intermediate layer, aleurone and starchy endosperm in each fraction as predicted by the
model of Barron (2011), along with the predictions for each component in whole wheat
samples. Note that the independent raw data for each component did not sum to unity, due
to inherent errors in the predictions and in their application to UK wheats; on average the
total material was overestimated by 8.3% for the Mallacca samples and 4.9% for Consort,
possibly suggesting that the French wheats used to generate the models were more similar
to the soft Consort wheat. The data reported in Table 6.1 have been normalised to unity, as
a reasonable approximation to the composition of particles in each size range, and to fit the
assumptions underlying the formulation of the compositional breakage equation.
Chapter 6
161
Table 6.1 Particle size distributions and composition of size fractions following milling of Mallacca
and Consort wheats under Sharp-to-Sharp and Dull-to-Dull dispositions.
Sieve Size
(µm)
X
Percentage
on sieve
(%)
ρ2(x)
Pericarp
concentration
(%)
ype
Intermediate
layer
concentration
(%)
yInlay
Aleurone
concentration
(%)
yal
Starchy
endosperm
concentration
(%)
yen
MALLACCA
Sharp-to-Sharp (S-S)
2000 7.92 12.6 5.5 6.6 75.4
1700 10.78 11.4 2.0 11.4 75.3
1400 19.49 11.7 1.6 6.1 80.6
1180 12.87 13.9 2.4 8.9 74.8
850 14.88 12.7 1.1 5.5 80.7
500 14.09 6.5 2.0 2.4 89.2
212 10.88 3.9 0.7 7.0 88.4
<212 9.10 9.2 1.9 9.7 79.2
Average 10.4 2.0 6.9 80.8
Dull-to-Dull (D-D)
2000 35.74 8.9 3.6 5.2 82.3
1700 11.66 15.2 3.0 7.1 74.7
1400 10.35 14.2 0.9 8.5 76.4
1180 5.14 13.3 2.7 3.6 80.4
850 6.47 8.9 2.5 2.1 86.4
500 10.75 5.7 1.7 5.1 87.5
212 11.06 7.8 0.0 4.5 87.7
<212 8.83 2.1 4.1 7.3 86.5
Average 9.3 2.6 5.6 82.5
Whole grain
(Xi)
8.3 1.2 6.0 84.4
CONSORT
Sharp-to-Sharp (S-S)
2000 17.93 3.8 3.5 11.0 81.8
1700 10.35 5.6 2.3 13.0 79.1
1400 14.37 7.2 2.8 11.7 78.3
1180 10.39 9.8 0.0 8.2 82.0
850 9.94 7.3 1.7 7.4 83.6
500 15.0 3.6 3.0 6.5 86.9
212 11.79 0.1 3.1 4.0 92.8
<212 10.23 0.9 3.8 2.8 92.5
Average 4.7 2.6 8.3 84.4
Dull-to-Dull (D-D)
2000 37.95 6.5 3.8 15.1 74.6
1700 8.86 8.3 1.4 11.8 78.5
1400 6.91 7.0 1.4 13.2 78.4
1180 4.78 9.5 1.1 12.9 76.5
850 6.31 4.7 1.9 9.1 84.3
500 12.09 0.9 4.1 5.6 89.4
212 12.16 0.0 4.5 7.0 88.6
<212 10.95 0.0 3.6 10.3 86.1
Average 4.5 3.2 11.5 80.7
Whole grain
(Xi)
2.3 2.9 5.8 88.9
6.1 Table 6.1 Particle size distributions and composition of size fractions following milling of
Mallacca and Consort wheats under Sharp-to-Sharp and Dull-to-Dull dispositions.
Chapter 6
162
The total percentage of each component in the whole Mallacca wheat was Xpe = 8.3%,
XInlay = 1.2%, Xal = 6.0% and Xen = 84.4%; and in the whole Consort wheat was Xpe =
2.3%, Xal = 5.8%, Xen = 88.9% and XInlay = 2.9%. Multiplying the amount of material on
each sieve by the concentration of a given component, and summing these, allows the
cumulative compositional distributions, Ype(x), Yal(x), Yen(x) and YInlay(x) (the proportion by
mass of the total botanical component that is in particles smaller than x) to be calculated.
The total is reported as the average for each component in Table 6.1, for each wheat type
under each milling disposition. Ideally, these averages would be the same under both
dispositions, and identical with the predicted compositions of the whole grains. Inspection
of Table 6.1 shows that there are some significant discrepancies, which underline again the
inherent errors in the prediction method and in its application to UK wheats. Nevertheless,
the data allow the compositional breakage function approach to be demonstrated, with
appropriate caution, and using the averages rather than the data for whole wheat in order to
ensure internal consistency in the analysis. The justification for this is that the average
values are averaged from eight measurements, compared with just one for the whole wheat
samples, and that in any case the PLS models were developed for milled stocks rather than
for whole wheats (Barron, 2011), so the results for the milled fractions might be expected
to be more accurate than those for the whole wheats.
Figure 6.4 shows the cumulative distributions for the PSD and for the four component
distributions, for the Mallacca wheat milled under S-S distribution. Figure 6.5 presents the
experimental data and the fitted size distributions in their non-cumulative forms for
Mallacca wheat. Table 6.2 reports the fitted DNKBF parameters for both wheat types.
Chapter 6
163
(a)
(b) (c)
(d) (e)
Figure 57 Cumulative particle size and component distributions, for Mallacca wheat milled under
a Sharp-to-Sharp distribution.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
P2(z
)
z
PSD
PSDType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Yp
e(z
)
z
Pericarp
PericarpType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
YIn
lay(
z)
z
Intermediate layer
Intermediate layerType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Ya
l(z)
z
Aleurone
AleuroneType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Ye
n(z
)
z
Starchy endospermStarchy endospermType 1Type 2DNKBF
6.4 Figure 6.4 Cumulative particle size and component distributions, for Mallacca wheat milled under
a Sharp-to-Sharp disposition.
Chapter 6
164
(a)
(b) (c)
(d) (e)
Figure 6.58 Non-cumulative particle size and component distributions, for Mallacca wheat milled
under a Sharp-to-Sharp distribution.
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1ρ
2(z
)
z
PSD
PSDType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρp
e(z
)
z
Pericarp
PericarpType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρIn
lay(
z)
z
Intermediate layer
Intermediate layerType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρal
(z)
z
Aleurone
AleuroneType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρe
n(z
)
z
Starchy endospermStarchy endospermType 1Type 2DNKBF
Figure 6.5 Non-cumulative particle size and component distributions, for Mallacca wheat milled
under a Sharp-to-Sharp disposition.
Chapter 6
165
Table 59 Fitted DNKBF parameters.
MALLACCA
Sharp-to-Sharp (S-S)
m1 n1 m2 n2
PSD 0.358 5.536 178.1 1.08 3.44
Pericarp 0.733 4.048 53.90 0.38 0.91
Intermediate
layer
0.374 4.814 99.98 0.79 1.26
Aleurone 0.558 5.177 99.97 0.63 2.13
Starchy
endosperm
0.293 6.294 342.8 1.18 3.98
Dull-to-Dull (D-D)
PSD 0.379 7.892 99.94 0.92 2.36
Pericarp 0.419 6.443 99.94 1.06 1.59
Intermediate
layer
0.263 7.037 99.93 0.41 0.47
Aleurone 0.455 6.999 99.93 0.61 1.44
Starchy
endosperm
0.395 8.155 99.92 0.97 2.91
CONSORT
Sharp-to-Sharp (S-S)
PSD 0.143 8.211 1526 0.99 2.24
Pericarp 0.790 4.024 53.90 0.75 0.63
Intermediate
layer
0.421 7.244 99.97 1.15 7.94
Aleurone 0.356 5.646 99.97 1.24 2.25
Starchy
endosperm
0.124 6.737 342.8 0.95 2.29
Dull-to-Dull (D-D)
PSD 0.432 8.672 99.94 0.98 3.79
Pericarp 0.228 4.355 99.69 6.13 24.25
Intermediate
layer
0.286 2.279 99.96 0.35 0.31
Aleurone 0.133 6.161 99.93 0.49 0.51
Starchy
endosperm
0.421 8.563 99.92 1.03 4.93
In order to fit the Double Kumaraswamy Function, the x-axis was normalised by dividing
particle size by 4000 µm, in order to yield Kumaraswamy shape parameters consistent with
previously reported work (although the current work only used 2000 µm for its largest
sieve, so the data beyond this size is not available). The DNKBF in its cumulative form is:
α
Table 6.2 Fitted DNKBF parameters.
Chapter 6
166
Breakage2 Type Breakage1 Type
2
22
11 11111
nmnmzzzP (6.22)
where z is the normalised size, P(z) is the percentage smaller than z, α is the proportion of
the PSD that can be described as Type 1 breakage, and m1 and n1 are parameters
corresponding to Type 1 breakage. The quantity (1– α) gives the proportion of Type 2
breakage, while m2 and n2 are the parameters that describe the form of Type 2 breakage.
Differentiating Equation 6.21 gives the non-cumulative form of the DNKBF:
Breakage2 Type
1
22
Breakage1 Type
1
112
222
111 11)(11
nmmnmmzznmzznmz
(6.23)
Considering the particle size distribution in Figure 6.4(a) and Figure 6.5(a), the DNKBF
describes the data well, yielding values of α = 0.36, m1 = 5.54, n1 = 178.10, m2 = 1.08 and
n2 = 3.44; these values are broadly consistent with previous work for a wheat of hardness
around 50, milled under S-S (Campbell et al., 2012).
Figures 6.4(a) and 6.5(a) also show the Type 1 and Type 2 functions that combine to give
the DNKBF. The values of m1 and n1 describe a narrow peak of mid-range particles, while
those for m2 and n2 describe a broad distribution of mostly small particles but extending to
include the very large particles. Chapter 4 described a mechanism for Type 2 breakage that
explains the co-production of the very large bran particles and the small endosperm
particles, and hence why they are described by the same Type 2 breakage function
(Galindez-Najera and Campbell, 2014).
Considering now the cumulative distribution shown for the pericarp in Figures 6.4(b) and
6.5(b), again the DNKBF describes the data well. Comparing Figures 6.4(a) and 6.5(b), it
appears that the pericarp material is noticeably concentrated in the mid-range particles. The
DNKBF shape parameters are m1 = 4.05, n1 = 53.9, m2 = 0.38 and n2 = 0.91, with the
proportion of Type 1 breakage, α = 0.733. The decrease in the Type 1 parameters, in
general, makes the Type 1 component of the distribution narrower, while the proportion of
Type 1, α, has increased to 0.733. Thus, pericarp is predominantly found in the mid-range
Type 1 particles resulting from breakage. This is a new insight into wheat breakage.
Chapter 6
167
The Type 2 parameters have both decreased to below 1, giving a very steep peak for the
very small particles, matching the experimental data at that point. This suggests that there
is a significant amount of pericarp in the very small particles. This can be understood as
pericarp “dust” that is produced during breakage. Although bran material (pericarp and
aleurone) tends to stay as large particles during roller milling, inevitably some small
particles of bran are produced, and this is evident here in the experimental data and in the
modelling of it. Again, this is a new insight that is consistent with the accepted physical
understanding of the nature of wheat breakage, but here has for the first time been
identified and described quantitatively.
Regarding the results for the aleurone layer, Figures 6.4(d) and 6.5(d) show very similar
results to those for pericarp; this makes sense, as the pericarp and aleurone tend to fuse
during conditioning and break together. The fit is not quite as good as for the pericarp,
despite the spectroscopic model being in general more accurate for aleurone than for
pericarp (Barron, 2011). However, the same features are evident: a greater concentration of
aleurone material in mid-range Type 1 particles, and the same spike of very small particles
of aleurone-containing “dust”. The proportion of Type 1 in this case is lower at 0.56, while
m1 = 5.20, n1 = 100, m2 = 0.63 and n2 = 2.13, all larger than the corresponding values for
pericarp. In general the increase in the values of the Kumaraswamy shape parameters
moves the distribution slightly to the right. This may suggest the aleurone is more
prevalent in slightly larger particles following breakage; possibly pericarp, being on the
outside, is eliminated from these larger particles more easily than aleurone, although a
physical mechanism is not obvious and the data does not support excessive speculation at
this point. However the more general point here is that the compositional variation of
particles is very similar for both pericarp and aleurone, and information from these two
different components points to similar conclusions regarding the nature of mid-range
particles and the production of bran dust.
Figures 6.4(e) and 6.5(e), which are related to the starchy endosperm, show contrasting
behaviour to the pericarp and aleurone, being more predominant in the smaller particles,
but with the fitted curves featuring a dip at the very smallest particles, consistent with these
particles containing significant amounts of bran dust. The proportion of Type 1 is 0.293,
with m1 = 6.30, n1 = 343, m2 = 1.18 and n2 = 3.98. The increase of m2 to >1 introduces the
hump at the lower end of the Type 2 curve. There is still a significant Type 1 bump in the
Chapter 6
168
middle of the distribution, indicating that there is a lot of endosperm material in these mid-
range Type 1 particles. This is for the simple reason that there are a lot of these Type 1
particles. It needs to be remembered that these distributions combine the particle size
distribution and the composition of those particles, such that the shapes of these curves are
dominated by the shape of the overall particle size distribution. The fit to the data is good,
but this data does not show clearly the concentrations of components in these particles.
Figures 6.4(c) and 6.5(c) show the results for the intermediate layer. This data is predicted
by the spectroscopic model least accurately, such that there is significant scatter in the data,
but the results show a similar pattern to those for pericarp and aleurone, adding confidence
that the features observed in the graphs for these two components are genuine.
As described previously, the concentration functions can be found by inserting the Double
Kumaraswamy Functions fitted to the particle size distribution and to the compositional
distributions into Equation 6.24:
ondistributisizeparticle
nmmnmm
ondistributii
nmmnmm
i
iii
zznmzznm
zznmzznmX
x
xXxy
222
111
222
111
11)(11
11)(11
)(
)()(
1
22
1
11
1
22
1
11
2
(6.24)
Figure 6.6 shows the concentration functions resulting from dividing the fitted DNKB
functions using Equation 6.24, for all four components, compared with the original
experimental data for each component’s concentration. The agreement is good, as one
would hope as it is a circular relationship – the experimental data was used to generate the
compositional breakage functions, so the reverse analysis (which is what the ratio of the
composition and particle size DNKBFs is) would be expected more or less to recreate the
experimental data. Figure 6.6 simply reassures that the analysis does indeed reveal
genuine features, while allowing continuous functions to be formulated that could not
readily be formulated from the raw compositional data.
Chapter 6
169
(a) (b)
(c) (d)
Figure 60 Concentration functions for pericarp, intermediate layer, aleurone and starchy
endosperm, compared with experimental data, for Mallacca wheat milled under a Sharp-to-Sharp
disposition.
Based on these results, some observations need to be considered; for example, although
dividing one wiggly function by another wiggly function gives an even more wiggly
function for which not every wiggle is meaningful, the curves obtained do seem to agree
with the trends in the experimental data. The curves and data beyond 2000 µm (z = 0.5)
should be largely ignored, as there was only one data point covering this entire range. But
below 2000 µm (z = 0.5), the concentration of pericarp as shown by the curve is high
initially and drops suddenly, indicating fine pericarp dust present as very small particles;
the experimental data also shows this. The concentration then increases to a peak for the
mid-range particles and begins to decrease again, features that are again reflected in the
experimental data.
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
yp
e (z
)
z
Pericarp
PericarpCocentration function
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
yIn
lay
(z)
z
Intermediate layer
Intermediate layerConcentration function
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
ya
l (z
)
z
Aleurone
AleuroneCocentration function
0
10
20
30
40
50
60
70
80
90
100
0 0.2 0.4 0.6 0.8 1
y en
(z)
z
Starchy endosperm
Starchy endospermConcentration function
Figure 6.6 Concentration functions for pericarp, intermediate layer, aleurone and starchy
endosperm, compared with experimental data, for Mallacca wheat milled under Sharp-to-Sharp
disposition.
Chapter 6
170
The curves and experimental data for aleurone show the same general pattern, although
with more scatter. The curves and data for the starchy endosperm show an inverse trend
with lower concentrations in the finest and the mid-range particles. The trend is less
pronounced because the endosperm dominates the composition of all the particles.
Meanwhile, the overall trend is downwards, consistent with the expectation that larger
particles are less concentrated in endosperm than smaller particles. The intermediate layer
seems to show a slightly increasing trend of concentration with particle size.
The concentration functions are clearly very complex; it would be not be possible to define
a simple function likely to be capable of describing variations in component concentration
for a range of wheats milled under a range of conditions. The approach presented here,
allowing the particle size distribution and the component distributions to be described by
Double Kumaraswamy Functions, the ratios of which give the concentration functions, is a
practical way to describe, quantify and interpret the effects of breakage on component
distributions.
Figures 6.7-6.9 show the equivalent results for the samples milled under a D-D disposition.
The fitted DNKBF parameters are again reported in Table 6.2.
Chapter 6
171
(a)
(b) (c)
(d) (e)
Figure 61 Cumulative particle size and component distributions, for Mallacca wheat milled under a
Dull-to-Dull distribution.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
P2(z
)
z
PSD
PSDType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Yp
e(z
)
z
Pericarp
PericarpType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
YIn
lay(
z)
z
Intermediate layer
Intermediate layerType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Yal
(z)
z
Aleurone
AleuroneType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Ye
n(z
)
z
Starchy endosperm
Starchy endospermType 1Type 2DNKBF
6.7 Figure 6.7 Cumulative particle size and component distributions, for Mallacca wheat milled under
a Dull-to-Dull disposition.
Chapter 6
172
(a)
(b) (c)
(d) (e)
Figure 62 Non-cumulative particle size and component distributions, for Mallacca wheat milled
under a Dull-to-Dull distribution.
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρ2(z
)
z
PSD
PSDType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρp
e(z
)
z
Pericarp
PericarpType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρIn
lay(
z)
z
Intermediate layer
Intermediate layerType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρal
(z)
z
Aleurone
AleuroneType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρe
n(z
)
z
Starchy endospermStarchy endospermType 1Type 2DNKBF
6.8 Figure 6.8 Cumulative particle size and component distributions, for Mallacca wheat milled under
a Dull-to-Dull disposition.
Chapter 6
173
(a) (b)
(c) (d)
Figure 63 Concentration functions for pericarp, aleurone, endosperm and intermediate layer,
compared with experimental data, for Mallacca wheat milled under a Dull-to-Dull disposition.
Although this is the same wheat, in other respects, these results are independent of those
discussed above; the size fractions were generated and analysed independently of those
produced from milling under S-S. It is encouraging that many of the features seen in the S-
S data also appear here: the higher concentrations of pericarp and aleurone in mid-range
Type 1 particles, and higher concentration of endosperm in smaller particles. A notable
difference is the absence of evidence of pericarp in the very fine dust, although there is still
evidence of aleurone material in this fine dust, and also of intermediate layer, while there is
a high concentration of pericarp in the slightly larger small particles. This probably reflects
limitations in this small set of experimental data, but could conceivably reflect
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
y p e
(z)
z
Pericarp
PericarpConcentration function
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
y In
lay (z
)
z
Intermediate layer
Intermediate layerConcentration function
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
y al (
z)
z
Aleurone
AleuroneConcentration function
0
10
20
30
40
50
60
70
80
90
100
0 0.2 0.4 0.6 0.8 1
y en
(z)
z
Starchy endosperm
Starchy endospermConcentration function
6.9 Figure 6.9 Concentration functions for pericarp, aleurone, endosperm and intermediate layer,
compared with experimental data, for Mallacca wheat milled under a Dull-to-Dull disposition.
differences
Chapter 6
174
in the nature of breakage under Dull-to-Dull compared with Sharp-to-Sharp milling.
Galindez-Najera and Campbell (2014) (based on Chapter 4) described differences in the
scraping of bran particles formed from Dull-to-Dull milling compared with Sharp-to-
Sharp. Based on this description, it is plausible that D-D gives less production of bran dust
in the first place, but yields more effective scraping of endosperm from the inside of the
large bran particles, this scraping generating aleurone and intermediate layer material in the
finest particles, but not getting as far as pericarp.
Figure 6.10 shows the cumulative distributions for the PSD and for the four component
distributions, for the Consort wheat milled under S-S distribution. Figure 6.11 presents the
experimental data and the fitted size distributions in their non-cumulative forms for
Consort wheat. The fitted DNKBF parameters are again reported in Table 6.2.
Chapter 6
175
(a)
(b) (c)
(d) (e)
Figure 64 Cumulative particle size and component distributions, for Consort wheat milled under a
Sharp-to-Sharp distribution.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1P
2(z
)z
PSD
PSDType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Yp
e(z
)
z
Pericarp
PericarpType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
YIn
lay(z
)
z
Intermediate layer
Intermediate layerType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Yal
(z)
z
Aleurone
AleuroneType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Ye
n(z
)
z
Starchy endospermStarchy endospermType 1Type 2DNKBF
6.10 Figure 6.10 Cumulative particle size and component distributions, for Consort wheat milled under
a Sharp-to-Sharp disposition.
Chapter 6
176
(a)
(b) (c)
(d) (e)
Figure 65 Non-cumulative particle size and component distributions, for Consort wheat milled
under a Sharp-to-Sharp distribution.
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρ2(z
)
z
PSDPSDType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρp
e(z
)
z
Pericarp
PericarpType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρIn
lay(z
)
z
Intermediate layerIntermediate layerType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρal
(z)
z
Aleurone
AleuroneType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρe
n(z
)
z
Starchy endosperm
Starchy endospermType 1Type 2DNKBF
6.11 Figure 6.11 Cumulative particle size and component distributions, for Consort wheat milled under
a Sharp-to-Sharp disposition.
Chapter 6
177
Considering the particle size distribution in Figure 6.10(a) and Figure 6.11(a), the DNKBF
describes the data well, yielding values of α = 0.143, m1 = 8.21, n1 = 1526.82, m2 = 0.99
and n2 = 2.24; these values are broadly consistent with previous work for a wheat of
hardness around 30, milled under S-S (Campbell et al., 2012).
Figures 6.10(a) and 6.11(a) also show the Type 1 and Type 2 functions that combine to
give the DNKBF. As a reminder, the values of m1 and n1 describe a narrow peak of mid-
range particles, while those for m2 and n2 describe a broad distribution of mostly small
particles but extending to include the very large particles.
Considering now the cumulative distribution shown for the pericarp in Figures 6.10(b) and
6.11(b), again the DNKBF describes the data well. Comparing Figures 6.10(a) and 6.11(b),
it appears that the pericarp material is clearly concentrated in the mid-range particles. The
DNKBF shape parameters are m1 = 4.02, n1 = 53.9, m2 = 0.75 and n2 = 0.63, with the
proportion of Type 1 breakage, α = 0.79. The decrease in the Type 1 parameters, in
general, makes the Type 1 component of the distribution narrower, while the proportion of
Type 1, α, has increased to 0.79. Thus, pericarp is predominantly found in the mid-range
Type 1 particles resulting from breakage. These results are similar to the findings for
Mallacca wheat.
Similar to Mallacca wheat, the Type 2 parameters for Consort wheat have both decreased
to below 1, but unlike Mallacca, a very small steep peak for the very small particles is
observed for Consort, matching the experimental data at that point. This suggests a little
amount of pericarp “dust” in the very small particles that is produced during breakage.
Although bran material (pericarp and aleurone) tends to stay as large particles during roller
milling, inevitably some small particles of bran are produced. Although this new insight is
not as evident as it is for Mallacca, there is still evident in both the experimental data and
in the modelling for Consort. It is proposed cautiously at this point, recognising that this
work is only for two wheat types and so far only a single component and only the S-S data
have been considered. But it serves at this point to illustrate the nature of the compositional
breakage function interpretation and the insights that can result.
Regarding the results for the aleurone layer, Figures 6.10(d) and 6.11(d) show a similar
pattern to those for pericarp, although unlike pericarp for Mallacca wheat, there is not a
steep peak for the very small particles (less dust production). The fit is not quite as good as
Chapter 6
178
for the pericarp, despite the spectroscopic model being in general more accurate for
aleurone than for pericarp (Barron, 2011). Similar to pericarp, a greater concentration of
aleurone material in mid-range Type 1 particles is evident and tiny very small particles of
aleurone-containing “dust”. The proportion of Type 1 in this case is lower at 0.36, while m1
= 5.65, n1 = 100, m2 = 1.240 and n2 = 2.252, all larger than the corresponding values for
pericarp. In general the increase in the values of the Kumaraswamy shape parameters
moves the distribution slightly to the right. This may suggest once again the aleurone is
more prevalent in slightly larger particles following breakage; possibly pericarp, being on
the outside, is eliminated from these larger particles more easily than aleurone, or, perhaps
the production of aleurone is coming from inside, in other words, the starchy endosperm
has been scraped off, reaching the aleurone.
Figures 6.10(e) and 6.11(e), which are related to the starchy endosperm, shows contrasting
behaviour to the pericarp and aleurone, being more predominant in the smaller particles,
but with the fitted curves featuring a dip at the very smallest particles, consistent with these
particles containing significant amounts of bran dust. The proportion of Type 1 is 0.124,
with m1 = 6.74, n1 = 343, m2 = 0.951 and n2 = 2.29. Similar to Mallacca wheat, there is a
significant Type 1 bump in the middle of the distribution, indicating that there is a lot of
endosperm material in these mid-range Type 1 particles. Again, this is for the simple
reason that there are a lot of these Type 1 particles.
Figures 6.10(c) and 6.11(c) show the results for the intermediate layer. As noted earlier,
this data is predicted by the spectroscopic model least accurately, such that there is
significant scatter in the data. However, the intermediate layer shows an opposite
behaviour with respect to pericarp and aleurone; for example, the production of dust is
considerable higher and not many mid-range particles are produced. This insight is
interesting because, while the intermediate layer might be expected to behave similarly to
aleurone and pericarp as part of the bran layers, the data suggest that the shearing effect
applied to this soft wheat causes the intermediate layer to crumble quite easily into small
particles, while the pericarp and aleurone on either side remain relatively intact.
Chapter 6
179
Figure 6.12 shows the concentration functions resulting from dividing the fitted DNKB
functions using Equation 6.23, for all four components, compared with the original
experimental data for each component’s concentration. Similar to Mallacca data, the
experimental Consort data was used to generate the compositional breakage functions, so
the reverse analysis (which is what the ratio of the composition and particle size DNKBFs
is) would be expected more or less to recreate the experimental data. Similar to Mallacca
wheat results, Figure 6.12 reassures that the analysis does indeed reveal genuine features,
while allowing continuous functions to be formulated that could not readily be formulated
from the raw compositional data.
(a) (b)
(c) (d)
Figure 66 Concentration functions for pericarp, intermediate layer, aleurone and starchy
endosperm, compared with experimental data, for Consort wheat milled under a Sharp-to-Sharp
disposition.
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
y pe
(z)
z
Pericarp
PericarpConcentration function
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
yIn
lay (z
)
z
Intermediate layer
Intermediate layerConcentration function
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
ya
l (z
)
z
Aleurone
Aleurone
Concentration function0
10
20
30
40
50
60
70
80
90
100
0 0.2 0.4 0.6 0.8 1
ye
n (z
)
z
Starchy endosperm
Starchy endospermConcentration function
6.12 Figure 6.12 Concentration functions for pericarp, intermediate layer, aleurone and starchy
endosperm, compared with experimental data, for Consort wheat milled under a Sharp-to-Sharp
disposition.
Chapter 6
180
Again, some observations need to be considered. As noted above, although dividing one
wiggly function by another wiggly function gives an even more wiggly function, the
curves obtained seem to agree with the trends in the experimental data. The curves and
data beyond 2000 µm (z = 0.5) should be largely ignored, as there was only one data point
covering this entire range. But below 2000 µm (z = 0.5), the concentration of pericarp as
shown by the curve is very low initially (low production of dust) and increases in the larger
particles. This increment is observed in the peak for the mid-range particles and begins to
decrease again, features that are again reflected in the experimental data. The curves and
experimental data for aleurone show the same general pattern, although with more scatter.
The curves and data for the starchy endosperm show an inverse trend with lower
concentrations in the finest and the mid-range particles. The trend is less pronounced
because the endosperm dominates the composition of all the particles. Meanwhile, the
overall trend is downwards, consistent once again with the expectation that larger particles
are less concentrated in endosperm than smaller particles. The intermediate layer seems to
show a decreasing trend of concentration with particle size and then increases a little bit in
the larger particles.
Figures 6.13-6.15 show the equivalent results for the samples milled under a D-D
disposition. The fitted DNKBF parameters are again reported in Table 6.2.
Chapter 6
181
(a)
(b) (c)
(d) (e)
Figure 67 Cumulative particle size and component distributions, for Consort wheat milled under a
Dull-to-Dull distribution.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
P2(z
)
z
PSD
PSDType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Yp
e(z
)
z
Pericarp
PericarpType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
YIn
lay(
z)
Particle size x (µm)
Intermediate layer
Intermediate layerType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Yal
(z)
Particle size x (µm)
Aleurone
AleuroneType 1Type 2DNKBF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 0.2 0.4 0.6 0.8 1
Ye
nd(z
)
z
Starchy endosperm
Starchy endospermType 1Type 2DNKBF
6.13 Figure 6.13 Cumulative particle size and component distributions, for Consort wheat milled under
a Dull-to-Dull disposition.
Chapter 6
182
(a)
(b) (c)
(d) (e)
Figure 68 Non-cumulative particle size and component distributions, for Consort wheat milled
under a Dull-to-Dull distribution.
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρ2(z
)
z
PSD
PSDType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρp
e(z
)
z
Pericarp
PericarpType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρIn
lay(
z)
z
Intermediate layer
Intermediate layerType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρal
(z)
z
Aleurone
AleuroneType 1Type 2DNKBF
0
1
2
3
4
5
6
0 0.2 0.4 0.6 0.8 1
ρe
n(z
)
z
Starchy endospermStarchy endospermType 1Type 2DNKBF
6.14 Figure 6.14 Non-cumulative particle size and component distributions, for Consort wheat milled
under a Dull-to-Dull distribution.
Chapter 6
183
(a) (b)
(c) (d)
Figure 69 Concentration functions for pericarp, aleurone, endosperm and intermediate layer,
compared with experimental data, for Consort wheat milled under a Dull-to-Dull disposition.
The results obtained are independent of those discussed above, although this is the same
wheat; the size fractions were generated and analysed independently of those produced
from milling Consort under S-S. It is again encouraging that many of the features seen in
the S-S data also appear here: the higher concentrations of pericarp and aleurone in mid-
range Type 1 particles, and higher concentration of endosperm in smaller particles. A
notable difference is the absence of evident pericarp in the very fine dust, although there is
still evidence of aleurone material in this fine dust. The intermediate layer shows a high
concentration of dust in the very small particles. In the slightly larger small particles there
is higher concentration of the intermediate layer which then decreases in the mid-range and
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
y pe
(z)
z
Pericarp
Pericarp
Concentration function
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
yIn
lay (z
)
z
Intermediate layer
Intermediate layer
Cocentration function
0
2
4
6
8
10
12
14
16
18
0 0.2 0.4 0.6 0.8 1
yal
(z)
z
Aleurone
Aleurone
Concentration function0
10
20
30
40
50
60
70
80
90
100
0 0.2 0.4 0.6 0.8 1
y en
d (z
)
z
Starchy endosperm
Starchy endospermConcentration function
6.15 Figure 6.15 Concentration functions for pericarp, aleurone, endosperm and Intermediate layer,
compared with experimental data, for Consort wheat milled under a Dull-to-Dull disposition.
Chapter 6
184
larger particles. This probably reflects limitations in this small set of experimental data, but
could conceivably reflect differences in the nature of breakage under Dull-to-Dull
compared with Sharp-to-Sharp milling. Based on the description of Galindez-Najera and
Campbell (2014) (Chapter 4), it is observed that aleurone and intermediate layer are
generating more dust than pericarp, which seems to show very little or no dust production
under D-D milling. Under S-S milling, the production of aleurone dust is less compared
with D-D milling, although pericarp dust is higher and intermediate layer seems to be even
more. All these features are contrasting to the harder Mallacca wheat, in which overall, the
bran dust production is considerable higher under both dispositions compared with the soft
Consort wheat, and particularly higher under D-D disposition as explained before. The
breakage mechanism observed here seems to suggest a more effective scraping of
endosperm from the inside of the large bran particles, this scraping generating aleurone and
intermediate layer material in the finest particles, but not getting as far as pericarp.
Figure 6.16 collects the pericarp, intermediate layer and aleurone distributions together on
the same graph, for both wheats under both dispositions.
Chapter 6
185
(a) (b)
(c) (d)
Figure 70 Pericarp, Intermediate layer and aleurone distributions for Mallacca (a,b) and Consort
(c,d) wheats milled under (a,c) Sharp-to-Sharp; and (b,d) Dull-to-Dull dispositions.
0.0000
0.0005
0.0010
0.0015
0.0020
0.0025
0 500 1000 1500 2000
i (x
)
x (µm)
Mallacca S-S
Pericarp
Intermediate layer
Aleurone
DNKBF Pericarp
DNKBF Intermediate layer
DNKBF Aleurone
0.0000
0.0005
0.0010
0.0015
0.0020
0.0025
0 500 1000 1500 2000
i (
x)
x (µm)
Mallacca D-D
Pericarp
Intermediate layer
Aleurone
DNKBF Pericarp
DNKBF Intermediate layer
DNKBF Aleurone
0.0000
0.0005
0.0010
0.0015
0.0020
0.0025
0 500 1000 1500 2000
i (x
)
x (µm)
Consort S-S
Pericarp
Intermediate layer
Aleurone
DNKBF Pericarp
DNKBF Intermediate layer
DNKBF Aleurone
0.0000
0.0005
0.0010
0.0015
0.0020
0.0025
0 500 1000 1500 2000
i (
x)
x (µm)
Consort D-D
Pericarp
Intermediate layer
Aleurone
DNKBF Pericarp
DNKBF Intermediate layer
DNKBF Aleurone
6.16 Figure 6.16 Pericarp, Intermediate layer and aleurone distributions for Mallacca (a,b) and Consort
(c,d) wheats milled under a Sharp-to-Sharp (a,c) and Dull-to-Dull (b,d) dispositions.
Chapter 6
186
Gathering together the data from all four conditions highlights certain consistent patterns
and some distinctive differences that together give a degree of confidence that the apparent
effects are genuine. Most striking is the contrast between the hard Mallacca wheat and the
soft Consort wheat, which is more striking than the difference between the S-S and D-D
dispositions. There are some intriguing and tantalising patterns within the compositional
data for Mallacca, most notably the aleurone peak being shifted to the right compared with
the pericarp peak (which is also evident for Consort under S-S), and the apparent
production of pericarp/intermediate layer/aleurone “dust” under S-S, but only intermediate
layer/aleurone dust, without pericarp, under D-D, which may point to subtleties in the
mechanisms of breakage. But more striking than these small differences is the relative
uniformity of the Mallacca compositions in relation to pericarp, intermediate layer and
aleurone, which vary in broadly consistent ways with particle size. This is in marked
contrast to Consort, in which the relative proportions of these three components appear to
vary substantially in particles of different size, pointing to very different breakage origins.
It appears that in the hard wheat, essentially the bran layers break “together”, with
subsequent minor variations in composition as bits are knocked off. This is consistent with
the general understanding that in hard wheats, the bran “breaks together with the
endosperm” (Fang and Campbell, 2002a,b, 2003a), and the breakage patterns being
dominated by the endosperm physical properties. By contrast, in the soft wheat, which
naturally produces much larger bran particles (Campbell and Muhamad, 2007) these large
flat particles are then scraped by the rollers in ways that alter their composition profoundly,
and more so under D-D than under S-S. The behaviour of these large bran particles is
therefore dictated much more by the properties and structure of the bran layers by
themselves than by the hardness of the endosperm.
Perhaps most interesting is the evidence that when a large flat bran particle produced from
a soft wheat is scraped by the differential action of the rollers, the intermediate layer
appears to crumble into smallish particles, while the pericarp, and to a lesser extent the
aleurone, manage to stay predominantly in large particles. This is evident under S-S, while
under D-D, the contrast between the pericarp and intermediate layer is even more evident,
with aleurone tending more towards smaller particles in this case. This idea that the
intermediate layer, which is physically located between the pericarp and aleurone layers,
appears to crumble into small particles whilst the layers either side remain more intact, has
profound consequences for understanding the nature of wheat breakage and differences
Chapter 6
187
between the milling performances of different wheats. It may be that this crumbly
intermediate layer is specific to this particular Consort sample, and not a general feature of
soft wheats, in which case the implications are even more profound, particularly for
Second break milling which is devoted to scraping of large flat bran particles (Mateos-
Salvador., 2013). Variations in the breakage patterns of the intermediate layer could be
exploited for developing wheats, or conditioning regimes, or first break/second break roll
gap combinations that lead to noticeably enhanced separation during second break milling.
Throughout this discussion it has been careful to highlight limitations in the scope and
accuracy of the study, and clearly these tentative suggestions would be more conclusive if
based on a wider range of wheats and roll gaps (if the scraping of large flat bran particles
has such profound effects on bran particle composition, it would have been interesting to
complement these results with those from a smaller roll gap, for which scraping would be
expected to be more severe). Nevertheless, the observed patterns are sufficiently similar in
certain respects and sufficient different in others, in ways that are consistent with the
known effects of wheat hardness and disposition on breakage (Fang and Campbell,
2002a,b, 2003a; Campbell and Muhamad, 2007), that there can be confidence that the new
insights are at least plausible. A greater understanding of the subtle effects of the physical
properties of bran and endosperm and their interaction with roll gap and disposition has the
potential to lead to more effective wheat breeding and flour processing. Meanwhile, this
work has demonstrated the new insights and quantitative understanding that can be
accessed through the compositional breakage equation approach.
Figure 6.17 shows the distributions of all four tissues (pericarp, intermediate layer,
aleurone and starchy endosperm) plotted together on the same graph, for both wheats under
both dispositions. In this graph the distributions have been multiplied by the proportions of
each component, such that Figure 6.17 is the equivalent of Figure 6.2. The distributions
therefore add up to give the overall particle size distribution, 2(x), i.e the figure is the
graphical representation of Equation 6.19, the compositional breakage equation in its non-
cumulative form.
Chapter 6
188
(a) (b)
(c) (d)
Figure 717 Pericarp and aleurone distributions for Mallacca (a,b) and Consort (c,d) wheats milled
under (a,c) Sharp-to-Sharp; and (b,d) Dull-to-Dull dispositions.
In these four graphs the starchy endosperm content dominates and dilutes the other three
botanical components, which makes sense considering that starchy endosperm is contained
in higher proportions in the whole wheat kernel than pericarp, intermediate layer and
aleurone, as observed previously in Figure 6.3 (Pomeranz, 1988; Barron, 2011; Barron et
al., 2011).
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0 500 1000 1500 2000
Xi
i (x
)
x (µm)
Mallacca S-S
PericarpIntermediate layerAleuronestarchy EndospermDNKBF PericarpDNKBF Intermediate layerDNKBF AleuroneDNKBF starchy EndospermPSD
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0 500 1000 1500 2000
Xi
i (x)
x (µm)
Mallacca D-D
PericarpIntermediate layerAleuronestarchy EndospermDNKBF PericarpDNKBF Intermediate layerDNKBF AleuroneDNKBF starchy EndospermPSD
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0 500 1000 1500 2000
Xi
i (x
)
x (µm)
Consort S-S
PericarpIntermediate layerAleuronestarchy EndospermDNKBF PericarpDNKBF Intermediate layerDNKBF AleuroneDNKBF starchy EndospermPSD
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0 500 1000 1500 2000
Xi
i (x
)
x (µm)
Consort D-D
PericarpIntermediate layerAleuronestarchy EndospermDNKBF PericarpDNKBF Intermediate layerDNKBF AleuroneDNKBF starchy EndospermPSD
6.17 Figure 6.17 Pericarp, Intermediate layer, aleurone and starchy endosperm distributions for
Mallacca (a,b) and Consort (c,d) wheats milled under (a,c) Sharp-to-Sharp (a,c), and Dull-to-Dull
(b,d) dispositions.
Chapter 6
189
Figure 6.17(a) shows a dashed line in Mallacca and Consort wheats milled under S-S
disposition, as examples of particles of different composition. To illustrate how
compositions can be calculated, for the Mallacca wheat milled under S-S disposition, the
values of the pericarp, intermediate Layer, aleurone and starchy endosperm for particles of
size 500 µm (shown by the dashed line in Figure 6.16(a)) are:
Tot
enal
Inlaype
gdx
gdxenenXgdxalal
X
gdxInlayInlay
XgdxpepeX
078.0071.0003.0001.0003.0)500(2
071.0)500(,003.0)500(
,001.0)500(,003.0)500(
From these values, the composition of particles of 500 m is determined as follows:
Toten
Total
TotInlay
Totpe
ggeny
ggal
y
ggInlay
y
ggpey
/
/
/
/
910.0078.0/071.0)500(
0385.0078.0/003.0)500(
013.0078.0/001.0)500(
0385.0078.0/003.0)500(
which means that 3.85% is pericarp, 3.85% is aleurone, 1.3% is intermediate layer, and
91.0% is starchy endosperm.
Similarly, using a contrasting example, for the Consort wheat milled under S-S disposition,
the values of the pericarp, intermediate Layer, aleurone and starchy endosperm for
particles of size 1500 µm (shown by the dashed line in Figure 6.16(c)) are:
Tot
enal
Inlaype
gdx
gdxenenXgdxalal
X
gdxInlayInlay
XgdxpepeX
0910.00721.00099.00012.00078.0)1500(2
0721.0)1500(,0099.0)1500(
,0012.0)1500(,0078.0)1500(
Then,
Toten
Total
TotInlay
Totpe
ggeny
ggal
y
ggInlay
y
ggpey
/
/
/
/
7923.00910.0/0721.0)1500(
1088.00910.0/0099.0)1500(
0132.00910.0/0012.0)1500(
0857.00910.0/0078.0)1500(
Chapter 6
190
leading to a composition for these particles of 8.6% pericarp, 1.3% intermediate Layer,
1.1% aleurone and 79.2% starchy endosperm.
The approach presented here, allowing the particle size distribution and the component
distributions to be described by Double Kumaraswamy Functions, the ratios of which give
the concentration functions, is a practical way to describe, quantify and interpret the effects
of breakage on component distributions. This approach also represents the continuous
equivalent of the discrete compositional breakage matrices introduced by Fistes and
Tanovic (2006), yielding greater predictive power and greater mechanistic insights in
wheat breakage.
More work is needed to evaluate the accuracy of the spectroscopic predictions for this sort
of application, and to apply the approach to a wider range of milled samples in order to
lead to more confident conceptions of the physical breakage mechanism operating during
roller milling of wheat.
Chapter 6
191
6.6 Summary
The distributions of wheat kernel components within eight milled fractions of Mallacca
and Consort wheats milled under S-S and D-D dispositions have been investigated by PLS
models developed by Barron (2011) and then, the concentration functions found by using
Double Kumaraswamy Functions fitted to the particle size distribution and to the
compositional distributions. The DNKBF was found to describe the data well for the four
botanical components studied: pericarp, aleurone, intermediate layer and starchy
endosperm, for both wheat types. For the Mallacca wheat, the pericarp and aleurone layer
compositions mostly varied with particle size in similar ways, consistent with these layers
fusing together as “bran” and breaking together, although with possibly a subtle difference
around the production of very fine particles under D-D milling. Although the data
predicted for the intermediate layer by the spectroscopic model was less accurate compared
with the other botanical tissues, the results show a broadly similar pattern to those for
pericarp and aleurone in the Mallacca wheat, adding confidence that the features observed
are genuine. However, for Consort wheat, the intermediate layer behaved differently from
pericarp and aleurone, suggesting a different breakage mechanism which could be
associated with how the wheat hardness affects breakage of the bran and the production of
large flat bran particles. This finding gives new insights into the nature of wheat breakage,
and the contribution of the intermediate Layer tissues to breakage, that could have
implications for wheat breeding and for flour mill operation.
The data from both wheats under the two milling dispositions highlight certain consistent
patterns and some distinctive differences that together give a degree of confidence that the
apparent effects are genuine. The contrast between the hard Mallacca wheat and the soft
Consort wheat is more evident than the difference between the S-S and D-D dispositions.
Some interesting patterns within the compositional data for Mallacca are observed like the
aleurone peak being shifted to the right compared with the pericarp peak, which is also
evident for Consort under S-S, and the apparent production of pericarp/intermediate
layer/aleurone dust under S-S, but only intermediate layer/aleurone dust, without pericarp,
under D-D, which may point to subtleties in the mechanisms of breakage. The relative
uniformity of the Mallacca compositions in relation to pericarp, intermediate layer and
aleurone, which vary in roughly consistent ways with particle size, is notable. This is in
contrast to Consort, in which the relative proportions of these three components appear
Chapter 6
192
to vary substantially in particles of different size, pointing to very different breakage
origins.
It is suggested tentatively that in the hard wheat the bran layers break “together”, with
subsequent minor variations in composition as bits are knocked off. By contrast, in the
soft wheat, which naturally produces much larger bran particles, these large flat particles
are then scraped in such a way that their composition changes profoundly, and more so
under D-D than under S-S. The behaviour of these large bran particles is therefore dictated
much more by the properties and structure of the bran layers than by the hardness of the
endosperm.
Chapter 7
193
CHAPTER 7
CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE
WORK
7.1 Introduction
Wheat milling aims to separate endosperm from bran and germ to produce high and
consistent yield and quality of white flour using a dry milling process that is efficient and
cheap, always trying to fulfil the specifications of the buyers. First break milling is a
critical point in the overall milling process, because the distribution of broken particles
dictates the flows through the rest of the mill, affecting directly the subsequent system
arrangement and machine settings and thus determining the yields and quality of flour.
First break milling opens up and scrapes the wheat grain, enabling the separation of bran
from endosperm. The breakage equation describing first break roller milling of wheat was
developed to predict the particle size distribution produced from first break roller milling.
This equation considers input grain characteristics such as diameter and hardness, and
processing parameters like roll gap. However, the broken output particles produced by
wheat breakage vary in size and in composition. In first break milling, the large particles
produced are associated with bran, while the small particles are related to endosperm
(Campbell, 2007). Therefore, a model of first break milling that considers particle
composition as well as size is needed. The milling industry aims to produce flour relatively
free of bran. With the aim to achieve a more efficient milling technology, the quantitative
and qualitative analysis of botanical elements in milled fractions would be particularly
useful. Therefore the main objective of the current work was to extend the DNKBF during
first break roller milling to include particle composition, by using biochemical markers
from pericarp, aleurone, intermediate layer and endosperm, which, in principle, would help
to indicate and predict the distribution of these components within different size fractions
obtained during the roller milling operation.
Chapter 7
194
7.2 Progress made in the current Thesis.
Five key advances were achieved in the present work: 1) the Double Normalised
Kumaraswamy breakage function was used to model first break milling of debranned
wheat; 2) a breakage mechanism was proposed to give insightful understanding of the
nature of the breakage mechanisms during first break milling; 3) the isolated wheat
botanical tissues along with the milled fractions were characterised with FTIR and in
relation to their sugar profiles using HPLC; 4) a compositional breakage equation for
wheat milling was developed and its compositional breakage functions fitted to
experimental data; and 5) the application of the compositional breakage equation to hard
and soft wheats suggested new insights into the mechanisms of wheat breakage and the
effects of wheat hardness and bran properties on breakage.
7.2.1 Modelling first break milling of debranned wheat using the
Double Normalised Kumaraswamy Breakage function.
The DNKBF was applied to model the first break milling of debranned wheat. Two
representative UK wheat types, Mallacca (hard) and Consort (soft) were conditioned,
debranned for different times and milled under two dispositions at three roll gaps. Type 1
breakage was found to increase at longer debranning times for both wheats under both
dispositions. S-S disposition produced more Type 1 breakage than D-D disposition, and the
Mallacca wheat broke to give more Type 1 mid-range particles than the softer Consort. It
was proposed that removal of bran decreases the production of large particles from which
very small particles can be scraped, thus increasing the proportion of Type 1, mid-range
particles. Therefore the effect of debranning was more dramatic under conditions that
favour production of Type 2 particles (i.e soft wheats under D-D milling). The proposed
breakage mechanism explained why Type 2 breakage accounts for both the very large
particles and the very small particles; the production of the latter by scraping depends on
the initial creation of the former. These results gave a better understanding of the physical
meaning of the DNKBF parameters, and also a more insightful understanding of the nature
of the breakage mechanisms during first break milling.
Chapter 7
195
7.2.2 Characterisation of wheat botanical tissues and milled
fractions with FTIR and sugar profiles using HPLC
The four major botanical components of the wheat grain, pericarp, aleurone, endosperm
and germ, were hand-dissected from Mallacca (hard) and Consort (soft) wheats. Despite
the good quality tissues isolated, checked with microscopic and FTIR analysis, the
aleurone layer appeared to be slightly contaminated with endosperm due to limitations of
the manual dissection technique. The main lipid, carbohydrate and protein regions of the
spectra were clearly distinguished, and the peak intensities reflected the known make up of
the different tissues in wheat. After applying PCA to the spectra of the botanical
components, some regions in the spectra were more intense in each wheat component,
showing, specific wavenumbers that could aid to their possible identification and
quantification in the milled samples. PCA was computed for milled samples, from which
the largest variance separated the fine particles (<212, 212 and 500 µm) from the coarse
particles (850, 1180, 1400, 1700, 2000 µm) along PC1. The Loading associated with PC1
showed the strongest absorbance in the carbohydrate region, mainly related to the cell wall
polysaccharides which are associated with the pericarp tissue. Milled samples of size 850,
1400 and 1700 µm were closely clustered, suggesting that these samples were similar in
composition. These results demonstrated the potential of analytical and mathematical
techniques to quantify the botanical distribution within milled fractions.
Three attempts at quantification of the botanical tissues in milled fractions were carried
out:
i) By selecting eight specific peaks observed in the spectra of all the samples
(botanical tissues and milled fractions), but in different proportions. The heights
of each peak in each sample were measured, then the relative contribution of
each wheat component in milled fraction samples was calculated. However,
only a few samples were tested and implausible results were obtained. Another
approach was performed by selecting the whole spectrum instead of specific
peaks but the same unsatisfactory results were found.
ii) By PLS, however, the required quantity of samples for building up a proper
calibration curve was not enough due to technical limitations, thus, this
statistical technique was abandoned.
Chapter 7
196
iii) By sugar profiles quantified by HPLC, in which, despite the promising results
obtained, the technique was complex and limited by the detection limit of
HPLC, in particular for arabinose and xylose content in samples rich in
endosperm particles.
Despite the interesting and promising findings, the results of these approaches were
insufficiently accurate and convenient for the quantification of the botanical components in
milled fractions and hence, not appropriate for the formulation of the compositional
breakage equation. However, although the FTIR approach was unsuccessful for the reasons
given, the principle remains sound, and has been implemented more comprehensively and
successfully in the PLS models developed by Barron (2011). Therefore a collaboration
with Barron was established to exploit these existing predictive models.
7.2.3 Development of a compositional breakage equation for wheat
milling
The botanical distributions within eight milled fractions of Mallacca and Consort wheats
milled under S-S and D-D dispositions were investigated by PLS models developed by
Barron (2011) and their concentration functions found using the DNKBF fitted to the
particle size distribution and to the compositional distributions. Overall, the DNKBF was
found to describe the data well for the four botanical components studied: pericarp,
aleurone, intermediate layer and starchy endosperm. Although the data predicted for the
intermediate layer by the spectroscopic model was less accurate compared with the other
botanical tissues, the results showed a broadly similar pattern to those for pericarp and
aleurone in the Mallacca wheat, adding confidence that the features observed in the graphs
for these two components are genuine. However, for Consort wheat, the intermediate layer
behaved differently from pericarp and aleurone, suggesting a different breakage
mechanism which could be associated with how the wheat hardness affects breakage of the
bran and the production of large flat bran particles.
The data from all four conditions showed consistent patterns and some distinctive
differences, giving a degree of confidence that the apparent effects are genuine. The
contrast between the hard (Mallacca) wheat and the soft (Consort) wheat was more evident
than the difference between the S-S and D-D dispositions. For Mallacca wheat, the
Chapter 7
197
aleurone peak was shifted to the right compared with the pericarp peak, which was also
evident for Consort under S-S, and the apparent production of pericarp/intermediate
layer/aleurone dust under S-S, but only intermediate layer/aleurone dust, without pericarp,
under D-D, which may point to subtle differences in the mechanisms of breakage. The
relative consistency in composition in relation to pericarp, intermediate layer and aleurone,
which vary in roughly regular ways with particle size, in Mallacca wheat fractions was
notable. This was in contrast to Consort, in which the relative proportions of these three
components appeared to vary substantially in particles of different size, pointing to very
different breakage origins.
A breakage mechanism was proposed tentatively in which the bran layers break together in
the hard wheat, with subsequent slight variations in composition as small pieces are
knocked off. By contrast, in the soft wheat, which naturally produces larger bran particles,
these large particles are then scraped in a way that their composition changes deeply, and
more so under D-D than under S-S. The behaviour of these large flat bran particles is
therefore more a consequence of the properties and structure of the bran layers than the
hardness of the endosperm. This mechanism was consistent with the earlier observed
effects of debranning on breakage (Galindez-Najera and Campbell, 2014). Thus the
application of the breakage equation to debranned wheat and of the compositional
breakage equation to hard and soft wheats has suggested new insights into the mechanisms
of wheat breakage and the effects of wheat hardness and bran properties on breakage.
7.3 Recommendations for future work
The approach presented in the current work is a practical way to describe, quantify and
interpret the effects of breakage on component distributions. It allows the particle size
distribution and the component distributions to be described by Double Kumaraswamy
Functions, with their ratios giving the concentration functions. More work is now needed
to evaluate the accuracy of the spectroscopic predictions for this type of application, and to
apply the approach to a wider range of milled samples in order to lead to more confident
conceptions of the physical breakage mechanisms operating during roller milling of wheat.
Spectroscopic techniques along with multivariate methods are useful tools to understand
the chemical bases of organic components, in this case, the wheat grain components. The
Chapter 7
198
potential of these techniques to follow the fate of the botanical components in the milling
process has been probed and its improvement is continual. In the current work, despite the
promising results obtained with PCA, some issues couldn’t be addressed successfully to
establish a PLS model for the prediction of the botanical tissues in milled fractions due to
problems including:
i) Difficulties in the isolation of aleurone, hence, low yield of this particular tissue.
ii) A large number of wheat kernels need to be dissected to obtain sufficient amounts
of each botanical component to be analysed, not only for creating a calibration
curve for the PLS analysis (different combinations of different proportions of each
component are required), but also for analytical techniques that are needed to
support the results obtained by spectroscopy.
iii) For an accurate and meaningful analysis of the spectra collected by FTIR with
Principal Component Analysis (PCA) and Partial Least Square (PLS), it is critical
to have the spectra treated, removing noise and undesired data.
The first and second of these issues are a somewhat inevitable consequence of the small
size and close structure of the wheat kernel and the limitations of manual dissection, to
which a large part of the solution is experience and careful hard work, particularly to
separate more effectively the aleurone layer. Perhaps instead of water, a different organic
solvent could be used, such as ethanol or acetone, or maybe these mixed with water in
different proportions. The problem of not getting enough material to build up an
appropriate calibration curve is a consequence of the difficulties of the hand-isolation
which again can be addressed by extensive and time-consuming work.
For the third issue identified, the pre-treatment of the data for computing PCA was
sufficient to find useful features to identify each botanical tissue; however, these
smoothing pre-treatments (base-line correction, normalization and 2nd
derivative using the
Savitsky-Golay method) may need some improvements before applying PLS to the data.
For example, changes in the Savitsky-Golay method used should be explored such as
different polynomial order and different smoothing points. Although the FTIR approach
was unsuccessful for the reasons given, the principle remains sound that profiling of the
botanical components of wheat should allow profiling of the composition of milled
fractions.
Chapter 7
199
Choomjaihan (2008) attempted to apply the principle using mineral profiling,
unsuccessfully because the soaking procedure allowed minerals to move around or from
tissues too readily, such that dissected and milled fractions were no longer directly
comparable. An alternative approach employing the same principle was investigated, based
this time on sugar profiles measured by HPLC analysis, in the hope that sugars would be
less mobile during soaking than minerals. As was explained in Chapter 5, only two wheat
components were selected for analysis: endosperm and bran (in this work, pericarp is
treated as the bran layer as it is easier to isolate and analyse, although it is recognised that
bran is composed of several layers of which pericarp is only). These components were
considered to contain the most representative and easiest sugars to be quantified by HPLC:
glucose for endosperm, and arabinose and xylose for bran. However, despite the potential
of this approach and its sound basis, some issues were identified in terms of limitations in
the accuracy with which sugars could be quantified by HPLC, and the time-consuming
nature of the approach.
For the second issue, the FTIR technique represents a much faster and easier technique
compared with HPLC but, as discussed earlier, FTIR needs direct chemical analysis to
serve as reference. For this reason, the collaboration made with Dr. Barron from INRA,
France was crucial for the analysis of the samples in the current work, exploiting the
predictive models already developed (Barron et al., 2007; Hemery et al., 2009; Barron et
al., 2011; Barron, 2011), in order to demonstrate the principles of the compositional
breakage equation and the insights it can draw from such compositional data.
Imaging analysis like the methods described in Chapter 3, such as fluorescence,
environmental scanning electron microscopy, X-ray and atomic force, are powerful
techniques that can support the results obtained in the current work in particular to perform
a qualitative analysis to identify certain botanical tissues in milled fractions. Combining
these methods with multivariate analysis may enable the development of prediction models
like those of Barron (2011).
The compositional breakage equation developed in the present work was focused on
monitoring the wheat botanical tissues present in milled fractions during first break roller
milling, and could be extended to model breakage during second, third and fourth breaks).
Meanwhile, there are numerous other industries that employ comminution processes in
which the broken particles may vary in composition as well as size, such as in ore
Chapter 7
200
extractions, cosmetic and pharmaceutical powders; the mathematical framework
demonstrated here could be applied to understand and improve these processes as well.
The progress achieved in the present thesis gives new insights into the nature of roller
milling of wheat breakage, compositional analysis and modelling. A significant step
forward to understand the wheat milling process, in particular first break roller milling, has
been reached, with potential application to other particle comminution processes. The
compositional breakage equation developed in the present work has the potential to help
millers process wheat more efficiently to fulfil the increasing demands of wheat as a raw
material for food and non-food products.
References
201
REFERENCES
AACC (1995a) Approved Methods of the American Association of Cereal Chemists, 9 ed.
(1), Minnesota, American Association of Cereal Chemists, Inc.
AACC (1995b) Approved Methods of the American Association of Cereal Chemists, 9 ed.
(1), Minnesota, American Association of Cereal Chemists, Inc.
Adams, R. M. 1966. The evolution of urban society: Early Mesopotamia and prehispanic
Mexico. Second printing 2007. New Jersey, USA: Aldine Transaction.
Al-Mogahwi, H. W. H. and Baker, C. G. J. 2005. Performance evaluation of mills and
separators in a commercial flour mill. Food Bioprod Process. 83:25-35.
Antoine, C., Peyron, S., Mabille, F., Lapierre, C., Bouchet, B., Abecassis, J. and Rouau, X.
2003. Individual contribution of grain outer layers and their cell wall structure to
the mechanical properties of wheat bran. J Agric Food Chem. 51:2026-2033.
Antoine, C., Peyron, S., Lullien-Pellerin, V., Abecassis, J. and Rouau, X. 2004. Wheat
bran tissue fractionation using biochemical markers. J Cereal Sci. 39:387-393.
Atchison, J., Head, L. and Gates, A. 2010. Wheat as food, wheat as industrial substance;
comparative geographies of transformation and mobility. Geoforum. 41:236-246.
Austin, L. G. 1971. A review: Introduction to the mathematical description of grinding as a
rate process. Powder Technol. 42: 91-109.
Austin, L. G. and Rogers, R. S. C. 1985. Powder technology in industrial size reduction.
Powder Technol. 42: 91-109.
Austin, L. G., Shoji, K. and Bell, D. 1982. Rate equations for non-linear breakage in mills
due to material effects. Powder Technol. 31: 127-133.
Barron, C. 2011. Prediction of relative tissue proportions in wheat mill streams by Fourier
Transform Mid-infrared spectroscopy. J Agric Food Chem. 59: 10442–10447.
Barron, C., Samson, M.-F., Lullien-Pellerin, V. and Rouau, X. 2011. Wheat grain tissue
proportions in milling fractions using biochemical marker measurements:
Application to different wheat cultivars. J Cereal Sci. 53:306-311.
Barron, C. and Rouau, X. 2008. FTIR and Raman signatures of wheat grain peripheral
tissues. Cereal Chem. 85:619-625.
Barron, C., Surget, A. and Rouau, X. 2007. Relative amounts of tissues in mature wheat
(Triticum aestivum L.) grain and their carbohydrate and phenolic acid composition. J
Cereal Sci, 45:88-96.
Barron, C., Parker, M. L., Mills, E. N. C., Rouau, X. and Wilson, R. H. 2005. FTIR
imaging of wheat endosperm cell walls in situ reveals compositional and
architectural heterogeneity related to grain hardness. Planta. 220: 667–677.
References
202
Batchelor, S., Booth, E. J., Walker, K. G. and Cook, P. 1993. The potential for bioethanol
production from wheat in the UK. Vol.H-GCA. Research review No. 29. London,
Home-Grown Cereals Authority.
Beaugrand, J., Cronier, D., Debeire, P. and Chabbert, B. 2004. Arabinoxylan and
hydroxycinnamate content of wheat bran in relation to endoxylanase susceptibility. J
Cereal Sci. 40:223-230.
Bechtel, D. B., Abecassis, J., Shewry, P. R. and Evers, A. D. 2009. Development, structure
and mechanical properties of the wheat grain. In: Wheat chemistry and technology.
Pp. 51-95. Edited by Khan, K. and Shewry, P. R. 4th
Edition. Minesota, USA: AACC
International, Inc.
Bekes, F., Gianibelli, M.C. and Wrigley, C. 2004. Wheat proteins and flour quality. In:
Encyclopedia of Grain Science. Edited by Wrigley, C., Walker, C., and Corke, H.
Oxford, UK: Elsevier.
Belderok, B., Mesdag, J. and Donner, D.A. 2000. The wheat grain. In: Bread-making
quality of wheat: a century of breeding in Europe. Pp. 15-18. Edited by Donner, D.A.
Dordrecht, The Netherlands: Kluwer Academic Publishers.
Beta, T., Nam, S., Dexter, J.E. and Sapirstein, H.D. 2005. Phenolic content and antioxidant
activity of pearled wheat and roller mill fractions. Cereal Chem. 82:390-393.
Blanco-Romía, M. and Alcalá-Bernández, M. 2009. Multivariate calibration for
quantitative analysis. In: Infrared spectroscopy for food quality analysis control. Pp.
51-82.Edited by Da-Wen Sun. Oxford, UK: Elsevier Inc.
Bradshaw, J. 2004. Debranning. Grain Feed Mill Tech. July-August: pp. 10-13.
Bradshaw, J. 2005. Developments in semolina milling. Grain Feed Mill Tech. July-August:
pp. 14-17.
Bonnin, S., Besson, F., Gelhausen, M., Chierici, S. and Roux, B. 1999. A FTIR
spectroscopy evidence of the interactions between wheat germ agglutinin and N-
acetylglucosamine residues. FEBS Letters. 456: 361-364.
Boserup, E. 1965. The conditions of agricultural growth: The economics of agrarian
change under population pressure. Reprinted in 2005. New Jersey, USA: Aldine
Transaction.
Bottega, G., Caramanico, R., Lucisano, M., Mariotti, M. and Ambrogina-Pagani, M. 2009.
The debranning of common wheat (Triticum aestivum L.) with innovative abrasive
rolls. J Food Eng. 94:75–82.
Brouns, F., Hemery, Y., Price, R. and Mateo Anson, N. 2012. Wheat Aleurone: Separation,
Composition, Health Aspects, and Potential Food Use. Critical Reviews in Food
Science and Nutrition. 52:553-568.
Campbell, G. M. 2007. Roller milling of wheat. In: Handbook of Powder Technology. Pp.
391-428. Edited by Salmon, A. D., Ghadiri M., and Hounslow, M. J. Amsterdam,
The Netherlands: Elsevier.
References
203
Campbell, G. M., Bunn, P. J., Webb, C. and Hooks, S. C. W. 2001a. On predicting roller
milling performance II: The breakage function. Powder Technol. 115:243:255.
Campbell, G. M., Fang, C., Bunn, P. J., Gibson, A. 2001b. Wheat flour milling: A case
study in processing of particle foods. In: Powders and Solids: Developments in
Handing and Processing Technologies. Pp. 95-111. Edited by Hoyle, W. Cambridge,
UK: The Royal Society of Chemistry.
Campbell, G. M., Fang, C. and Muhamad, I. I. 2007. On predicting roller milling
performance VI: Effect of kernel hardness and shape on the particle size distribution
from first break milling of wheat. Food Bioprod Process. 85:7-23.
Campbell, G. M. 2008. A history of aerated foods. In: Bubbles in food 2: Novelty, Health
and Luxury. Pp. 1-21. Minnesota, USA: Eagan Press.
Campbell, G. M. and Martin, P. J. 2011. Bread dough aeration and rheology. In:
Breadmaking: Improving Quality. Pp. 1-36. Edited by Cauvain S.P. Woodhead
Publishing Ltd., Cambridge, UK.
Campbell, G. M., Sharp, C., Wall, K., Mateos-Salvador, F., Gubatz, S., Huttly, A. and
Shewry, P. 2012. Modelling wheat breakage during roller milling using the Double
Normalised Kumaraswamy Breakage function: Effects on kernel shape and
hardness. J Cereal Sci. 55:415-425.
Campbell, G. M. and Webb, C. 2001. On predicting roller milling performance I: The
breakage equation. Powder Technol. 115:234-242.
Carson, G. C. and Edwards, N. M. 2009. Criteria of wheat and flour quality. In: Wheat
chemistry and technology . Pp. 97-118. Edited by Khan, K. and Shewry, P. R. 4th
Edition. Minesota, USA: AACC International, Inc.
Chalmers, J. M., Edwards H. G. M. and Hargreaves, M., 2012. Vibrational Spectroscopy
Techniques: Basics and Instrumentation. In: Infrared raman spectroscopy in forensic
science. Pp 9-40. Published by John Wiley & Sons, Inc. Hoboken, New Jersey, USA.
Chan, L. E., Chalmers, J. and Griffiths, P. 2010. Part One: Introduction and Basic
Concepts. In: Applications of vibrational spectroscopy in food science, Vol.1. Pp 3-
89. Published by Jhon Wiley & Sons, Inc. Hoboken, New Jersey, USA.
Chantapet, P., Kunanopparat, T., Menut, P. and Siriwattanayotin. 2013. Extrusion
processing of wheat gluten bioplastic: Effect of the addition of kraft lignin. J Polym
Environ. 21:864-873.
Childe, V. G. 1934. New light on the most ancient East: The oriental prelude to European
prehistory. London, UK: Keagan Paul.
Choomjaihan, P. 2008. Extending the breakage equation on first break milling of wheat to
include particle composition. The University of Manchester, UK.
Courtin, C.M. and Delcour, J.A., 2002. Arabinoxylans and endoxylanase in wheat flour
bread-making. J Cereal Sci, 35:225–243.
References
204
Cui, S. W. and Wang, Q. 2009. Cell wall polysaccharides in cereals: chemical structures
and functional properties. Struct Chem. 20:291-297.
Decker, E., Beecher, G., Slavin, J., Miller, H. E. and Marquart, I. 2002. Whole grains as a
source of antioxidants. Cereal Foods World. 47:370-373.
DEFRA, 2014. [Online]. Available by the UK government department responsible for
policy and regulations on the environment, food and rural affairs. Access date on
March 20th
, from www.defra.gov.uk/.
Delcour, J. A. and Hoseney, R. C. 2010. Principles of cereal science and technology.
Minnesota, USA: AACC International.
Delcour, J. A., Rouau, X., Courtin, C. M., Poutanen, K. and Ranieri, R. 2012.
Technologies for enhanced exploitation of the health promoting potential of cereals.
Trends Food Sci Technol. 25:78-86.
Dexter, J. E. and Sarkar, A. K. 2004. Wheat: dry milling. In: Encyclopedia of grain
science. Pp. 363-374. Edited by Wrigley, C., Corke, H., and Walker, C. Oxford, UK:
Elsevier.
Dexter, J. E. and Wood, P. J. 1996. Recent applications of debranning of wheat before
milling. Trends Food Sci Technol. 7:35-41.
Dixon, J. 2007. The economics of wheat: Value chains from research to field to fork. Pp.
9-22. In: Proc. 7th
Int. Wheat Conf. Buck, H. T., Nisi, J. E. and Salomón N., Eds.
Springer, The Netherlands.
Dobraszcyk, B. J. and Morgenstern, M. P. 2003. Rheology and the breadmaking process. J
Cereal Sci. 38:229-245.
Dobraszczyk, B.J., Campbell, G.M. and Gan, Z. 2000. Bread—a unique food. In:
Dobraszczyk, B.J., Dendy, D.A.V. (Eds.), Cereals and Cereal Products: Technology
and chemistry, Aspen Publishers, USA.
Dobraszczyk, B.J., Smewing, J., Albertini, M., Maesmans, G. and Schofield, J.D. 2003.
Extensional rheology and stability of gas cell walls in bread doughs at elevated
temperatures in relation to breadmaking performance. Cereal Chemistry 80:218–224.
Du, C., Campbell, G. M., Misailidis, N., Mateos-Salvador, F., Sadhukhan, J., Mustafa, M.
and Weightman, R. M. 2009. Evaluating the feasibility of commercial arabinoxylan
production in the context of a wheat biorefinery principally producing ethanol. Part
1. Experimental studies of arabinoxylan extraction from wheat bran. Chem Eng Res
Des. 87:1232-1238.
Dubcovsky, J. and Dvorak, J. 2007. Genome plasticity: a key factor in the success of
polyploid wheat under domestication. Science. 316:1862-1866.
Dukor, R. K. 2002. Vibrational spectroscopy in the detection of cancer. Biomedical
Applications, 5: 3335–3359.
References
205
Esbensen, K. H., Guyot, D., Westad, F. and HoumØller, L. P. 2002. Multivariate data
nnalysis – in practice: An introduction to multivariate analysis and experimental
design. Fifth edition. Published by CAMO, ASA. Oslo, Norway.
Eugster, W. and Gerschwiler, O. 2006. Method and installation for cleaning cereal. US
patent application publication US 2006/0147591 A1. Applicant, Buhler AG.
Evans, L. T. 1993. Crop evolution, adaptation and yield. Cambridge, UK: Cambridge
University Press.
Evers, A. D. 2004. Grain and feed milling technology. Pp. 6–9.
Evers, A. and Millar, S. 2002, Cereal grain structure and development: some implications
for quality. J Cereal Sci. 36:261-284.
Fabian, H., Jackson, M., Murphy, L., Watson, P.H., Fichtner, I. and Mantsch, H.H. 1995.
A comparative infrared spectroscopic study of human breast tumors and breast tumor
cell xenografts. Biospectroscopy. 1: 37–45.
Fabriani, G. and Lintas, C., 1988. Carbohydrates of durum wheat. In: Durum wheat:
chemistry and technology. Pp. 121-138. Edited by Fabriani, G., and Lintas, C.
Minnesota, USA: AACC International, Inc.
Fanali, C., Dugo, P., Mondello, L., D’Orazio, G. and Fanali, S. 2013. Recent developments
in High-Performance Liquid Chromatography. In: Food analysis by HPLC. Pp.1-31.
Edited by Nollet, L. M. L. and Toldrá F. 3rd edition. CRC Press. Boca Raton, U.S.
Fang, C. and Campbell, G. M. 2002a. Effect of roll fluting disposition and roll gap on
breakage of wheat kernels during first break roller milling. Cereal Chem. 79:518-
522.
Fang, C. and Campbell, G. M. 2002b. Stress-strain analysis and visual observation of
wheat kernel breakage during roller milling using fluted rolls. Cereal Chem. 79:511-
517.
Fang, C. and Campbell, G. M. 2003a. On predicting roller milling performance IV: Effect
of roll disposition on the particle size distribution from first break milling of wheat. J
Cereal Sci. 37:21-29.
Fang, C. and Campbell, G. M. 2003b. On predicting roller milling performance V: Effect
of moisture content on the particle size distribution from first break milling of wheat.
J Cereal Sci. 37: 31-41.
Field, L. D, Sternhell S. and Kalman, J.R. 2007. Organic structures from spectra. Fourth
edition. Published by Jhon Wiley & Sons, Inc. Hoboken, New Jersey, USA.
Fistes, A. and Tanovic, G. 2006. Predicting the size and compositional distributionsof
wheat flour stocksfollowing first break roller milling using the breakage matrix
approach. J Food Eng. 75:527-534.
Fuh, K. F., Coate, J. M. and Campbell, G. M. 2014. Effects of roll gap, kernel shape and
moisture on wheat breakage modelled using the Double Normalised Kumaraswamy
Breakage Function. Cereal Chem. 91:8-17.
References
206
Fulcher, R.G., O’Brien, T. P. and Lee, J. W.1972. Studies on the aleurone layer. I.
Conventional and fluorescence microscopy of the cell wall with emphasis on phenol-
carbohydrate complexes in wheat. Australian Journal of Biological Science. 25: 23-
24.
Galindez-Najera, S. P. and Campbell, G. M. 2014. Modelling first break milling of
debranned wheat using the Double Normalised Kumaraswamy Breakage function.
Cereal Chem. (In press) doi.org/10.1094/CCHEM-02-14-0028-R.
Goormaghtigh, E., Cabiaux, V. and Ruysschaert, J.M. 1994 in: Physical Methods. In: The
Study of Biomembranes (Hilderson, H.J. and Ralston, G.B., Eds.), Vol. 23, pp. 329-
362, Plenum Press, New York.
Greffeuille, V., Abecassis, J., Bar l’Helgouac’h, C. and Lullien-Pellerin, V. 2005.
Differences in the aleurone layer fate between hard and soft common wheats at grain
milling. Cereal Chem. 82:138-143.
Griffiths, P. R. and De Haseth, J. A. 2007. Fourier Transform Infrared spectrometry.
Second edition. Published by Jhon Wiley & Sons, Inc. Hoboken, New Jersey, USA.
Gooding, M. J. 2009. The wheat crop. In: Wheat Chemistry and technology. Pp. 19-49.
Edited by Khan, K. and Shewry, P. R. 4th
Edition. Minnesota, USA: AACC
International, Inc.
Guttieri, M. J., Souza E. J. and Sneller, C. 2008. Non-starch polysaccharides in wheat flour
wire-cut cookie making. J Agric Food Chem, 56:10927-10932.
Gys, W., Gebruers, K., Sфrensen, J. F., Courtin, C. M. and Delcour, J. A. 2004a.
Debranning of wheat prior to milling reduces xylanase but not xylanase inhibitor
activities in wholemeal and flour. J Cereal Sci. 39, 363–369.
Gys, W., Courtin, C. M. and Delcour, J. A. 2004b. Reduction of xylanase activity in flour
by debranning retards syruping in refrigerated doughs. J Cereal Sci. 39:371-377.
Hair, J. F., Anderson, R. E. and Tatham, R. L. 2006. Multivariate data analysis. Sixth
edition. Publisher Upper Saddle River, N.J. London:Prentice Hall.
Hamer, R. J. and van Vliet, T. 2000. Understanding the structure and properties of gluten:
an overview. In: Wheat gluten. Proceedings of the 7th
International workshop gluten
2000. Editors, Shewry, P. R. and Tatham, A. S. Cambridge, UK. Royal Society of
Chemistry.
Hamer, R. J., MacRitchie, F. and Weegels, P. L. 2009. Structure and functional properties
of gluten. In: Wheat chemistry and technology. Pp. 153-178. Edited by Khan, K. and
Shewry, P. R. 4th
Edition. Minnesota, USA: AACC International, Inc.
Hareland, G.A. 2003. Effects of pearling on falling number and alpha-amylase activity of
preharvest sprouted spring wheat. Cereal Chem. 80:232–237.
Heard, P. J., Feeney, K. A., Allen, C. F. and Shewry, P. R. 2002. Determination of the
elemental composition of mature wheat grain using a modified secondary ion mass
spectrometer (SIMS). The Plant Journal. 30:237-245.
References
207
Hemery, Y., Rouau, X., Dragan, C., Bilici, M., Beleca, R. and Dascalescu, L. 2009.
Electrostatic properties of wheat bran and its constitutive layers: Influence of particle
size, composition, and moisture content. J Food Eng. 93:114-124.
Hemerey, Y., Rouau, X., Lullien-Pellerin, L., Barron, C. and Abecassis, J. 2007. Dry
processes to develop wheat fractions and products with enhanced nutritional quality.
J Cereal Sci. 46:327-347.
Himmelsbach, D. S., Khahili, S. and Akin, D. E. 1998. FT-IR microspectroscopic imaging
of flax (Linum usitatissum L.) stems. Cell Mol Biol. 44:99-108.
Hsieh, F. H., Martin, D. G., Black, H. C. and Tipples, K. H. 1980. Some factors affecting
the first break grinding of Canadian wheat. Cereal Chem. 57:217-223.
Huleihel, M., Salman, A., Erukhimovich, V., Ramesh, J., Hammody, Z. and Mordechai, S.
2002. Novel optical method for study of viral carcinogenesis in vitro. Journal of
Biochemical and Biophysical Methods. 50:111–121.
Izydorczyk, M. S., McMillan, T. L., Kletke, J. B. and Dexter, J. E. 2011. Effects of
pearling, grinding conditions, and roller mill flow on the yield and composition of
milled products from hull-less barley. Cereal Chem. 88:375-384.
Jackel, S. S. 1995. Foreword. In: Wheat end uses around the world. Edited by Faridi, H.,
and Faubion, J. M. Minnesota, USA: AACC International, Inc.
Jaillais, B., Bertrand, D. and Abecassis, J. 2012. Identification of the histological origin of
durum wheat milling products by multispectral imaging and chemometrics. J Cereal
Sci. 55:210-217.
Jamme, F., Robert, P., Bouchet, B., Saulnier, L., Dumas, P. and Guillon, F. 2008. Aleurone
cell walls of wheat grain: high spatial resolution investigation using synchrotron
infrared microspectroscopy. Appl. Spectrosc. 62: 895–900.
Jensen, S. A. and Martens, H. 1982. The botanical constituents of wheat and wheat milling
fractions. II. Quantification by amino acids. Cereal Chem. 60:172-177.
Jensen, S. A., Munck, L. and Martens, H. 1982. The botanical constituents of wheat and
wheat milling fractions. I. Quantification by autofluorescence. Cereal Chem.
59:477-484.
Jollife, I. T. 1986. Principal Component Analysis; Springer-Verlag:New York.
Kacurakova, M. and Wilson, R. H. 2001. Developments in mid-infrared FTIR
spectroscopy of selected carbohydrates. Carbohydr Polym. 44: 291–303.
Karunakaran, C., Gaillard, C., Bouchet, B., Gnaegi, H., Buleon, A., Wang, J. and
Hitchcock, A. P. 2009. Characterization of wheat grain tissues by Soft X-Ray
Spectromicroscopy. Activity Report. Life Sciences.
Kumaraswamy, P. 1980. A generalized probability density function for double-bounded
random processes. J Hydrol. 46:79-88.
References
208
Laca, A., Mousia, Z., Diaz, M., Webb, C. and Pandiella, S. 2006. Distribution of microbial
contamination within cereal grains. J of Food Eng. 72: 332-338.
Liu, Z., Wang, H., Wang, X. E., Xu, H., Gao, D., Zhang, G., Chen, P. and Liu, D. 2008.
Effect of wheat pearling on flour phytate activity, phytic acid, iron, and zinc content.
Food Sci Technol-LEB. 41:521-527.
MacMasters, M. M., Hinton, J. J. C. and Bradbury, D. 1971. Microscopic structure and
composition of the wheat kernel. pp 193-201 in: Wheat: Chemistry and Technology,
2nd. Y. Pomeranz, ed. AACC International: St. Paul, MN.
Maes, C. and Delcour, J. A. 2002. Structural characterisation of Water-extractable and
Water-unextractable arabinoxylans in wheat bran. J Cereal Sci. 35:315–326.
Mark, H. and Workman, J. 2007. Derivatives in Spectroscopy. In: Chemometrics in
Spectroscopy Pp. 339-378. Oxford, UK: Academic Press. Elsevier.
Martinez-Hernandez, E., Ibrahim, M. H., Leach, M., Sinclair, P., Campbell, G. M. and
Sadhukhan, J. 2013. Environmental sustainability analysis of UK whole-wheat
bioethanol and CHP systems. Biomass and Bioenergy. 50:52-64.
Mateos-Salvador, F., Sadhukan, J. and Campbell, G.M. 2011. The Normalised
Kumaraswamy Breakage function: A simple model for wheat roller milling. Powder
Technol. 208:144-157.
Mateos-Salvador, F., Sadhukhan, J. and Campbell, G. M., 2013. Extending the Normalised
Kumaraswamy Breakage function of roller milling of wheat flour stocks to Second
break. Powder Technol. 237:107-116.
McGee, B. 2006. Color sorting. World Grain. 24:74-78.
McGee, B. C. 1995. The PeriTec process and its application to durum wheat milling. Oper.
Millers Bull., Mar. Pp 6521-6528.
McGee, B. C., 1996. A new rollermill and debranner for use in a compact mill.,
Association of Operative Millers Bulletin, 6674–6675.
Misailidis, N., Campbell, G. M., Du, C., Sadhukhan, J., Mustafa, M., Mateos-Salvador, F.
and Weightman, R. M. 2009. Evaluating the feasibility of commercial arabinoxylan
production in the context of a wheat biorefinery principally producing ethanol Part 2.
Process simulation and economic analysis. Chem Eng Res Des. 87:1239–1250.
Mousia, Z., Edherly, S., Pandiella, S. S. and Webb, C. 2004. Effect of wheat pearling on
flour quality. Food Research International. 37:449–459.
NABIM. 2014. The UK flour milling [Online]. Available by National Association of
British and Irish Millers. Access date 20th
May, from http://faostat.fao.org/2014.
Nagata, Y. and Burger, M. M. 1974. Wheat germ agglutinin. Molecular characteristics and
specificity for sugar binding. J Biol Chem. 249: 3116-3122.
References
209
Neethirajan, S., Thomson, D. J., Jayas, D. S. and White, N. D. G. 2008. Characterization of
the surface morphology of durum wheat starch granules using atomic force
microscopy. Microscopy Research and Technique. 71:125-132.
Ordaz-Ortiz, J. J. and Saulnier L. 2005. Structural variability of arabinoxylans from wheat
flour. Comparison of water-extractable and xylanase-extractable arabinoxylans. J
Cereal Sci. 42:119-125.
Osborne, B. G. and Anderssen, R. S. 2003. Review: Single kernel characterization
principles and applications. Cereal Chem. 80:613-622.
Paulsen, G. M. and Shroyer, J. P. 2004. Wheat: Agronomy. In: Encyclopedia of grain
science. Pp. 337-3447. Edited by Wrigley, C., Walker, C., and Corke, H. Oxford,
UK: Elsevier. Vol. 3.
Peyron, S., Surget, A., Mabille, F., Autran, J. C., Rouau, X. and Abecassis, J. 2002.
Evaluation of tissue dissociation of durum wheat grain (Triticum durum Desf.)
generated by the milling process. J Cereal Sci. 36:199-208.
Piergiovanni, A. R. and Volpe, N. 2002. Capillary electrophoresis of gliadins as a tool in
the discrimination and characterization of hulled wheats (Triticum dicoccon Schrank
and T. spelta L.). Cereal Chem. 80:269-273.
Pomeranz, Y. and Williams, P.C. 1990. Wheat hardness: its genetic, structural, and
biochemical background, measurement, and significance. In: Advances in cereal
science and technology. Edited by Pomeranz Y. Pp. 471-529. Minnesota, USA:
AACC International, Inc.
Posner, E. S. 2009. Wheat flour milling. In: Wheat chemistry and technology. Pp. 119-152.
Edited by Khan, K. and Shewry, P. R. 4th
Edition. Minesota, USA: AACC
International, Inc.
Posner, E. and Benjamin, C. 2003. Milling: principles of milling. In: Encyclopedia of food
sciences and nutrition. Pp. 3980-3986. Edited by Caballero, B., Trugo, L. and
Finglas, P. Oxford, UK: Academic Press.
Posner, E. S. and Hibbs, H. N. 2005. Wheat Flour Milling. Pp. 127-129. Second edition.
Minnesota, USA: AACC International, Inc.
Rinnan, A., Norgaard, L., van der Berg, F., Thygesen, J., Bro, R., and Engelsen, S.B. 2009.
Data Pre-processing. In: Infrared spectroscopy for food quality analysis control. Pp.
51-82.Edited by Da-Wen Sun. Oxford, UK: Elsevier Inc.
Robert, P., Marquis, M., Barron, C., Guillon, F. and Saulnier, L. 2005. FTIR investigation
of cell wall polysaccharides from cereal grains. Arabinoxylan infrared assignment. J
Agric Food Chem. 53: 7014–7018.
Salmeron-Ochoa, I. 2010. Chemical and sensorial properties of cereals fermented with
human derived lactic acid bacteria. The University of Manchester, UK.
References
210
Sapirstein, H. D., Wang, M. and Beta, T. 2013. Effects of debranning on the distribution
of pentosans and relationships to phenolic content and antioxidant activityof wheat
pearling fractions. Food Sci. Technol. 50:336-342.
Saulnier, L., Guillon, F., Sado, P. E. and Rouau, X. 2007. Plant cell wall polysaccharides
in storage organs: Xylans (food applications). In: Comprehensive glycoscience. Pp.
653–689. Edited by Kamerling, J., Boons, G. J., Lee, Y., Suzuki, T. A., Taniguchi,
N., and Voragen, A. G. J. Oxford, UK: Elsevier. Vol. 2.
Scudiero, L. and Morris, C. F. Field emission scanning electron and atomic force
microscopy, and raman and X-ray photoelectron spectroscopy characterization of
near-isogenic soft and hard wheat kernels and corresponding flours. J Cereal Sci.
52:136-142.
Sharp, C. 2010. Modelling Wheat breakage during first break roller milling based on
Single Kernel Characteristics. MEng Dissertation. The University of Manchester,
UK.
Shewry, P. R. 2009. Wheat. J Exp Bot. 60:1537-1553.
Singh, S. and Singh, N. 2010. Effect of debranning on the physico-chemical, cooking,
pasting and textural properties of common and durum wheat varieties. Food Res Int.
43:2277–2283.
Smith, B.C. 2011. Fundamentals of FourierTransform Infrared spectroscopy. 2nd
Edition.
Published by CRC Press, Taylor and Francis Group, LLC. USA.
Smith, L. I., 2002. A tutorial on principal components analysis. Cornell University, USA.
Smits, A. L. M., Ruhnau, F. C., Vliegenthart, J. F. G. and van Soest, J. J. G. 1998. Ageing
of starch based systems as observed with FT-IR and solid state NMR spectroscopy.
Starch-Staerke. 50:478-483.
Sovrani, V., Blandino, M., Scarpino, V., Reyneri, A., Coïsson, J. D., Travaglia, F.,
Locatelli, M., Bordiga, M., Montella, R. and Arlorio, M. 2012. Bioactive compound
content, antioxidant activity, deoxynivalenol and heavy metal contamination of
pearled wheat fractions. Food Chem. 135:39-46.
Stallknecht, G.F., Gilbertson, K.M. and Ranney, J.E. 1996. Alternative wheat cereals as
food grains: Einkorn, emmer, spelt, kamut, and triticale. In: Progress in new crops.
Pp. 156-170. Edited by Janick, J. ASHS Press, Alexandria.
Stone, B. and Morell, M. K. 2009. Carbohydrates. In: Wheat Chemistry and Technology.
Pp. 299-362. Edited by Khan, K. and Shewry, P. R. 4th
Edition. Minesota, USA:
AACC International, Inc.
Storck, J. and Teague, W. D. 1952. Flour for Man’s Bread. Minneapolis, USA.
Sun, L. J., Zhou, G. Y., Zhi, G. A. and Li, Z. G. 2007. Effects of different milling methods
on flour quality and performance in steamed breadmaking. J Cereal Sci. 45:18–23.
References
211
Sudgen, T.D. 2001. Wheat Flour Milling. In: Cereals and cereals products: Chemistry and
technology. Pp. 140-172. Edited by Dendy, D. A. V., and Dobraszczyk, B. J.,
Maryland, USA: Aspen Publishers, Inc.
Symons, S. J. and Dexter, J. E. 1992. Estimation of milling efficiency: Prediction of flour
refinement by the measurement of pericarp fluorescence. Cereal Chem. 69:137-141.
Symons, S. J. and Dexter, J. E. 1991 Computer analysis of fluorescence for the
measurement of flour refinement as determined by flour ash content, flour grade
color, and tristimulus color measurements. Cereal Chem. 68: 454–460.
Symons, S. J. and Dexter, J. E. 1993. Relationship of flour aleurone fluorescence to flour
refinement for some Canadian hard common wheat classes. Cereal Chem. 70: 90–95.
Symons, S. J. and Dexter, J. E. 1994. Aleurone and pericarp fluorescence as estimators of
mill stream refinement for various Canadian wheat classes. J Cereal Sci. 23: 73–83.
Tkac, J. J. 1992. US Patent 5082680.
Toone, E.J. 1994. Structure and energetics of protein-carbohydrate complexes. Structural
Biology. 4:719-728.
Shewry, P. R. 2009. Wheat. J Exp Bot. 60:1537-1553.
Singh, S. and Singh, N. 2010. Effect of debranning on the physico-chemical, cooking,
pasting and textural properties of common and durum wheat varieties. Food Res Int.
43:2277–2283.
Smith, B.C. 2011. Fundamentals of FourierTransform Infrared Spectroscopy. 2nd
Edition.
Published by CRC Press, Taylor and Francis Group, LLC. USA.
Smith, L. I., 2002. A tutorial on principal components analysis. Cornell University, USA.
Smits, A. L. M., Ruhnau, F. C., Vliegenthart, J. F. G. and van Soest, J. J. G. 1998. Ageing
of starch based systems as observed with FT-IR and solid state NMR spectroscopy.
Starch-Staerke. 50:478-483.
Sovrani, V., Blandino, M., Scarpino, V., Reyneri, A., Coïsson, J. D., Travaglia, F.,
Locatelli, M., Bordiga, M., Montella, R. and Arlorio, M. 2012. Bioactive compound
content, antioxidant activity, deoxynivalenol and heavy metal contamination of
pearled wheat fractions. Food Chem. 135:39-46.
Stallknecht, G.F., Gilbertson, K.M. and Ranney, J.E. 1996. Alternative wheat cereals as
food grains: Einkorn, emmer, spelt, kamut, and triticale. In: Progress in new crops.
Pp. 156-170. Edited by Janick, J. ASHS Press, Alexandria.
Stone, B. and Morell, M. K. 2009. Carbohydrates. In: Wheat Chemistry and Technology.
Pp. 299-362. Edited by Khan, K. and Shewry, P. R. 4th
Edition. Minesota, USA:
AACC International, Inc.
Storck, J. and Teague, W. D. 1952. Flour for Man’s Bread. Minneapolis, USA.
References
212
Sun, L. J., Zhou, G. Y., Zhi, G. A. and Li, Z. G. 2007. Effects of different milling methods
on flour quality and performance in steamed breadmaking. J Cereal Sci. 45:18–23.
Sudgen, T.D. 2001. Wheat Flour Milling. In: Cereals and Cereals Products: Chemistry and
Technology. Pp. 140-172. Edited by Dendy, D. A. V., and Dobraszczyk, B. J.,
Maryland, USA: Aspen Publishers, Inc.
Symons, S. J. and Dexter, J. E. 1992. Estimation of milling efficiency: Prediction of flour
refinement by the measurement of pericarp fluorescence. Cereal Chem. 69:137-141.
Symons, S. J. and Dexter, J. E. 1991 Computer analysis of fluorescence for the
measurement of flour refinement as determined by flour ash content, flour grade
color, and tristimulus color measurements. Cereal Chem. 68:454–460.
Symons, S. J. and Dexter, J. E. 1993. Relationship of flour aleurone fluorescence to flour
refinement for some Canadian hard common wheat classes. Cereal Chem. 70:90–95.
Symons, S. J. and Dexter, J. E. 1994. Aleurone and pericarp fluorescence as estimators of
mill stream refinement for various Canadian wheat classes. J Cereal Sci. 23:73–83.
Tainter, J. A. 1988. Introduction to collapse. In: The collapse of complex societies. Pp. 1-
18. Cambridge, UK: Cambridge University Press.
Tkac, J. J. 1992. US Patent 5082680.
Toone, E.J. 1994. Structure and energetics of protein-carbohydrate complexes. Structural
Biology. 4:719-728.
Van der Kamp, J. W. 2012. Paving the way for innovation in enhancing the intake of
whole grain. Trends in Food Science and Technology. 25:101-107.
Van Vliet, T., Janssen, A.M., Bloksma, A.H. and Walstra, P., 1992. Strain hardening of
dough as a requirement for gas retention. J Texture Studies. 23:439–460.
Waduge, R. N., Xu, S., Bertoft, E. and Seetharaman, K. 2013. Exploring the surface
morphology of developing wheat starch granules by using Atomic Force
Microscopy. Starch. 65:398-409.
Warren, F. J., Royall, P. G., Gainsford, S., Butterworth, P. J. and Ellis, P. R. 2011.
Carbohydrate Polymers. 86:1038– 1047.
Wellman, W., 1992. Wheat milling process. US Patent 5 089 282. Applicant, ConAgra Inc.
Wrigley, C. W. 2009. Wheat: A unique grain for the world. In: Wheat Chemistry and
Technology. Pp. 1-17. Edited by Khan, K. and Shewry, P. R. 4th
Edition. Minnesota,
USA: AACC International, Inc.
Wrigley, C. W. 2004. Cereals, Overview. In: Food science Australia and wheat CRC,
North Ryde, NSW, Australia: Elsevier.
Yuan, J., Flores, R.A., Eustace, D. and Milliken, G.A. 2003. A systematic analysis of the
break subsystems of a wheat flour milling. Transactions of The International of
Chemical Engineers. 81: 170-179.
References
213
Zanyar, M., Shazza, R. and Ihtesham ur R. 2008. Fourier Transform Infrared (FTIR)
spectroscopy of biological tissues. Applied Spectroscopy Reviews. 43: 134–179.
Zohary, D., Hopf, M., and Weiss, E. 2012. Domestication of plants in the Old World. 4th
Edition. New York, USA: Oxford University Press Inc.
Website pages
http://www.britishmuseum.org/explore/highlights/highlight_objects/pe_prb/q/quern_stone.
aspx (Access date May 2014)
http://www.cerealsdb.uk.net/cerealgenomics/WheatBP/Documents/DOC_Milling.php
(Access date May, 2014)
http://www.chem.agilent.com/cag/cabu/terms&def.htm (Access date March, 2014)
http://www.fao.org/docrep/x2184e/x2184e04.htm (Access date February, 2014)
http://www.rsc.org/CFOF (Access date February, 2014)
http://www.satake.com.au/ (Access date February, 2014)
http://www.kswheat.com (Access date June, 2014)
http://www.defra.gov.uk/ (Access date February, 2014)
Apendices
214
APPENDICES
APPENDIX 1
Table A1.1 Fitted DNKBF parameters for debranned wheat.
MALLACCA
Sharp-to-Sharp
Time (s) a α m1 n1 m2 n2
0 0.462 0.329 4.271 177.54 0.983 4.486
5 0.504 0.404 3.751 130.55 0.992 5.028
10 0.456 0.428 4.110 193.96 1.035 5.779
20 0.466 0.499 3.837 175.29 1.004 5.897
30 0.518 0.522 3.862 243.14 1.015 6.398
35 0.493 0.508 3.886 250.16 1.032 6.533
40 0.524 0.550 3.731 225.94 0.981 6.195
50 0.512 0.516 3.982 349.32 1.023 6.522
60 0.485 0.538 3.871 299.71 0.996 6.244
MALLACCA
Dull-to-Dull
0 0.383 0.196 4.34 79.5 0.861 2.686
5 0.427 0.332 3.785 70.077 0.888 3.536
10 0.374 0.335 4.19 123.805 0.915 3.816
20 0.430 0.534 3.033 52.484 0.836 4.169
30 0.442 0.612 2.865 56.383 0.758 3.876
35 0.467 0.566 3.052 72.317 0.822 4.402
40 0.513 0.566 3.136 102.808 0.829 4.512
50 0.472 0.566 3.052 72.317 0.822 4.402
60 0.494 0.617 2.986 95.406 0.755 3.987
CONSORT
Sharp-to-Sharp
0 0.350 0.064 5.004 77.26 0.890 3.123
5 0.431 0.115 4.659 115.9 0.952 3.676
10 0.523 0.202 3.177 43.22 0.965 4.546
20 0.519 0.219 3.516 80.45 1.033 5.137
30 0.455 0.318 2.688 34.02 0.990 5.268
35 0.51 0.376 2.625 38.59 0.966 5.264
40 0.519 0.374 2.444 30.64 0.978 5.517
50 0.432 0.564 2.039 20.36 0.856 4.577
CONSORT
Dull-to-Dull
0 0.400 0.062 6.274 60.00 0.653 1.637
5 0.424 0.067 6.080 93.25 0.759 2.230
10 0.384 0.097 4.936 51.19 0.822 2.600
20 0.452 0.112 4.652 92.15 0.928 3.636
30 0.460 0.358 2.420 11.19 1.005 6.472
35 0.437 0.607 1.582 7.926 0.707 3.535
40 0.510 0.620 1.659 9.843 0.701 3.602
50 0.399 0.662 1.603 8.466 0.679 3.429
60 0.464 0.667 1.619 10.26 0.699 3.772
Apendices
215
APPENDIX 2
PSD as described by the DNKBF from Mallacca wheat under both milling dispositions at
0, 5, 10, 20, 30, 35, 40 50 and 60 s of debranning.
Mallacca S-S
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakagetype 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0 s 5 s 10 s
20 s 30 s 35 s
40 s 50 s 60 s
Apendices
216
APPENDIX 2
PSD as described by the DNKBF from Mallacca wheat under both milling dispositions at
0, 5, 10, 20, 30, 35, 40 50 and 60 s of debranning.
Mallacca D-D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage
Type 2 breakage
Combined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage
Type 2 breakage
Combined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage
Type 2 breakage
Combined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage
Type 2 breakage
Combined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage
Type 2 breakage
Combined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage
Type 2 breakage
Combined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0 s 5 s 10 s
20 s 30 s 35 s
40 s 50 s 60 s
Apendices
217
APPENDIX 2
PSD as described by the DNKBF from Consort wheat under both milling dispositions at 0,
5, 10, 20, 30, 35, 40 50 and 60 s of debranning.
Consort S-S
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0 s 5 s 10 s
20 s 30 s 35 s
40 s 50 s 60 s
Apendices
218
APPENDIX 2
PSD as described by the DNKBF from Consort wheat under both milling dispositions at 0,
5, 10, 20, 30, 35, 40 50 and 60 s of debranning.
Consort D-D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakageType 2 breakageCombined
0 s 5 s 10 s
20 s 30 s 35 s
40 s 50 s 60 s
Apendices
219
APPENDIX 3
Type 1 and Type 2 breakages profiles for Mallacca and Consort wheats under S-S and D-D
dispositions.
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage, Mallacca S-S0 s5 s10 s20 s30 s35 s40 s50 s60 s
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage, Mallacca D-D0 s5 s10 s20 s30 s35 s40 s50 s60 s
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 2 breakage, Mallacca S-S0s5s10s20s30s35s40s50s60s
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 2 breakage, Mallacca D-D0s5s10s20s30s35s40s50s60s
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage, Consort S-S0 s5 s10 s20 s30 s35 s40 s50 s60 s
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 1 breakage, Consort D-D0 s5 s10 s20 s30 s35 s40 s50 s60 s
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 2 breakage, Conort S-S0s5s10s20s30s35s40s50s60s
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
0.0 0.2 0.4 0.6 0.8 1.0
ρ(z
)
z
Type 2 breakage, Consort D-D0s5s10s20s30s35s40s50s60s
Apendices
220
APPENDIX 4
Experimental data and cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35, 40,
50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca S-S
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0 s 0 s 0 s
5 s 5 s 5 s
10 s 10 s 10 s
Apendices
221
APPENDIX 4
Experimental data and cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35, 40,
50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca S-S
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
20 s
35 s 35 s 35 s
20 s 20 s
30 s 30 s 30 s
Apendices
222
APPENDIX 4
Experimental data and cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35, 40,
50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca S-S
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
40 s
60 s 60 s 60 s s
40 s 40 s
50 s 50 s 50 s
Apendices
223
APPENDIX 4
Experimental data and non-cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca S-S
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
DNKBF
0.7 mm
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
) ρ
(z)
ρ(z
)
ρ(z
)
0 s 0 s 0 s
5 s 5 s 5 s
10 s 10 s 10 s
Apendices
224
APPENDIX 4
Experimental data and non-cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca S-S
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
20 s
35 s 35 s 35 s
20 s 20 s
30 s 30 s 30 s
Apendices
225
APPENDIX 4
Experimental data and non-cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca S-S
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
40 s
60 s 60 s 60 s
40 s 40 s
50 s 50 s 50 s
Apendices
226
APPENDIX 4
Experimental data and cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35, 40,
50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca D-D
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0 s 0 s 0 s
5 s 5 s 5 s
10 s 10 s 10 s
Apendices
227
APPENDIX 4
Experimental data and cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35, 40,
50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca D-D
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
20 s
35 s 35 s 35 s
20 s 20 s
30 s 30 s 30 s
Apendices
228
APPENDIX 4
Experimental data and cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35, 40,
50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca D-D
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
40 s
60 s 60 s 60 s
40 s 40 s
50 s 50 s 50 s
Apendices
229
APPENDIX 4
Experimental data and non-cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca D-D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
% S
mall
er
than
χ/χ
max
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
0 s 0 s 0 s
5 s 5 s 5 s
10 s 10 s 10 s
Apendices
230
APPENDIX 4
Experimental data and non-cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca D-D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
20 s
35 s 35 s 35 s
20 s 20 s
30 s 30 s 30 s
Apendices
231
APPENDIX 4
Experimental data and non-cumulative DNKBF for Mallacca wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Mallacca D-D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
40 s
60 s 60 s 60 s
40 s 40 s
50 s 50 s 50 s
Apendices
232
APPENDIX 4
Experimental data and cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35, 40, 50
and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort S-S
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0 s 0 s 0 s
5 s 5 s 5 s
10 s 10 s 10 s
Apendices
233
APPENDIX 4
Experimental data and cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35, 40, 50
and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort S-S
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
20 s
35 s 35 s 35 s
20 s 20 s
30 s 30 s 30 s
Apendices
234
APPENDIX 4
Experimental data and cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35, 40, 50
and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort S-S
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
40 s
60 s 60 s 60 s
40 s 40 s
50 s 50 s 50 s
Apendices
235
APPENDIX 4
Experimental data and non-cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort S-S
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
0 s 0 s 0 s
5 s 5 s 5 s
10 s 10 s 10 s
Apendices
236
APPENDIX 4
Experimental data and non-cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort S-S
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
20 s
35 s 35 s 35 s
20 s 20 s
30 s 30 s 30 s
Apendices
237
APPENDIX 4
Experimental data and non-cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at S-S. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort S-S
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
40 s
60 s 60 s 60 s
40 s 40 s
50 s 50 s 50 s
Apendices
238
APPENDIX 4
Experimental data and cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35, 40, 50
and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort D-D
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0 s 0 s 0 s
5 s 5 s 5 s
10 s 10 s 10 s
Apendices
239
APPENDIX 4
Experimental data and cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35, 40, 50
and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort D-D
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
20 s
35 s 35 s 35 s
20 s 20 s
30 s 30 s 30 s
Apendices
240
APPENDIX 4
Experimental data and cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35, 40, 50
and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort D-D
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5
DNKBF
0
10
20
30
40
50
60
70
80
90
100
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7
DNKBF
40 s
60 s 60 s 60 s
40 s 40 s
50 s 50 s 50 s
Apendices
241
APPENDIX 4
Experimental data and non-cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort D-D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
) ρ
(z)
ρ(z
)
ρ(z
)
0 s 0 s 0 s
5 s 5 s 5 s
10 s 10 s 10 s
Apendices
242
APPENDIX 4
Experimental data and non-cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort D-D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
) ρ
(z)
ρ(z
)
20 s
35 s 35 s 35 s
20 s 20 s
30 s 30 s 30 s
Apendices
243
APPENDIX 4
Experimental data and non-cumulative DNKBF for Consort wheat at 0, 5, 10, 20, 30, 35,
40, 50 and 60 s of debranning at D-D. From left to right, roll gap 0.3, 0.5 and 0.7 mm.
Consort D-D
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.5 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.3 mm
DNKBF
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0.0 0.2 0.4 0.6 0.8 1.0
P(z
)
z
0.7 mm
DNKBF
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
ρ(z
)
40 s
60 s 60 s 60 s
40 s 40 s
50 s 50 s 50 s
Apendices
244
APPENDIX 5
Partial Least Square (PLS)
Partial Least Square is one of several methods used in multivariate analysis for
construction of predictive models. PLS first models the variants in the X matrix (measured
values) that best explains the variants in Y matrix. Similar to PCA, each matrix contains in
each raw a sample and in each column a variable. In PLS, the X and Y matrices are
modelled simultaneously, hence, a model is constructed that can give the best fit of Y
response to X (Smith, 2002; Blanco-Romía and Alcalá-Bernárdez, 2009). Although the
PLS models are “PCA-like”, which means that the structure is the same to PCA but the
scores and loadings are calculated from a different criteria. In PLS the criteria is that the
scores on X and Y need to have maximum covariance, in other words, to have a maximum
linear correlation between X and Y (Esbensen et al., 2002; Hair et al., 2006). One of the
biggest advantages of PLS is that it can handle missing data, noisy data, and it builds a
model for both the X and the Y space (Hair et al., 2006). Figure A3.1 shows the two data
sets used for the PLS, an input data matrix (X) and an output data matrix (Y).
Figure 72 Typical data set for a PLS. Adapted from Esbensen et al., 2002.
PLS is frequently applied to FTIR data for quantitative analysis of food and
pharmaceutical products, histological tissues, fuels, etc. In general, the use of multivariate
statistics for extraction, exploration and modelling of data of interest in order to solve
problems in chemistry, biochemistry, biology, medicine and chemical engineering is most
widely known as Chemometrics (Esbensen et al., 2002; Smith, 2002; Chan et al., 2010).
X
Y
Rows: Cases, observations, batches
Columns: Variables, tags, classes
Model
Figure A5.1 Typical data set for a PLS. Adapted from Esbensen et al., 2002.
244
