Post on 22-Mar-2018
TAMPERE UNIVERSITY OF TECHNOLOGYI n s t i t u t e o f P a p e r C o n v e r t i n g
Prediction of WVTR with General Regression Models
Kimmo LahtinenSession 9.1
Paper x.y Speakers name 2
1. IntroductionTarget
• TARGET: To establish a practical, fast and easy-to-use computer-aided prediction model for water vapourbarrier of extrusion coated paper
• Computer-aided prediction model creates a base formaterial selectioncost estimationoptimization
of a new packaging material.Already existing packages: Modelling eases the load of experimental testing.
Paper x.y Speakers name 3
Regression models
• Results in this study are based on statistical findings.Experimental testsRegression analysis
• Regression models are sort of “black-box type” models.No theoretical linkages between variables
• In technology, regression models are used when more deterministic models are not efficient due to complexity and disturbances.
Paper x.y Speakers name 4
Background of water vapour permeation
• Mathematical treatment of water vapour transmission rate (WVTR)
Fick’s first law: Steady state diffusion → D does not depend on penetrant’s concentration.The product DS is called coefficient of permeation (P)Henry’s law: c = Sp →
→ The determination of WVTR: Unit: g/m2/24h
dxdcDJ −=
( ) ( )l
ppPl
ppDSl
ccDJ s
010101 )( −=
−=
−=
LppP
dtdQ
AWVTR
)(1 12 −==
Paper x.y Speakers name 5
Three external factors influencing moisture barrier of polymer film
• temperature; effect on P• humidity; effect on (p2-p1)• thickness; effect on L
• The effect of temperature is controlled by the Arrheniusrelationship as follows:
LppP
dtdQ
AWVTR )(1 12 −==
)/exp(0 RTEPP p−=
Paper x.y Speakers name 6
2. Materials and methodsPilot line
Paper x.y Speakers name 7
Materials
• Modelled polymersLDPE, density 923 kg/m3
HDPE, density 941 kg/m3
PPCOC
• Paper One-side pigment coated paper to offer a smooth substrate for coating polymer (practically no influence on WVTR)
Paper x.y Speakers name 8
WVTR test method
• Cup method (SCAN-P22:68)
• The advantage: capable to carry multitude of samples at the same time
• Accurate enough for a statistical study
Paper x.y Speakers name 9
Test series
• Regression modelling requires extensive experimental testing for statistics.
5 set points with different coating weights for each coating.4 parallel measurements with each coating weight giving 20 results total for each polymer.16 different atmospheric conditions (T and RH):
1. conditions 2. conditions 3. conditions 4. conditionsSeries 1 23°C 50% 30°C 50% 38°C 50% 45°C 50%Series 2 23°C 63% 30°C 63% 38°C 63% 45°C 63%Series 3 23°C 77% 30°C 77% 38°C 77% 45°C 77%Series 4 23°C 90% 30°C 90% 38°C 90% 45°C 90%
Paper x.y Speakers name 10
WVTRs were measured for exact 20 g/m2 coating weight to achieve an accurate comparison between the results in different atmospheric conditions.
y = 423,67x-1,0349
R2 = 0,9888
0
10
20
30
40
50
0 10 20 30 40 50 60
coating weight (g/m2)
WVT
R (g
/m2 /2
4h)
LDPEStandard tropical conditions
38°C, RH 90%
Applied method• Power law of
regression
Paper x.y Speakers name 11
3. ResultsWVTR as a function of T and RH
3D Surf ace Plot (Spreads heet2.s ta 13v *16c)
WVTR = Distance Weighted Least Squares
35 30 25 20 15 10 5
50% RH 63% RH 77% RH 90% RH 23°C 3,03 3,45 4,30 4,89 30°C 5,11 6,86 8,16 9,74 38°C 9,41 12,13 15,37 19,08 45°C 14,74 18,92 25,73 32,59
WVTR results for 20 g/m2 LDPE coating
Paper x.y Speakers name 12
Definition of mixing ratio• Relative humidity is not
the actual water concentration of surroundings.
• Mixing ratio (ω) is defined as the ratio of the amount of water (kg) and the amount of dry air (kg).
• T and RH determine mixing ratio from the h,ωdiagram of humid air. (basic thermodynamics)
Paper x.y Speakers name 13
• Mixing ratio as a function of T and RH:
where µ = MH20 / Mair = 18,015/28,964 = 0,6220, p = normal air pressure = 1 bar and ph’(T) = saturated vapour pressure (function of temperature)
( )( )TpRH
pTp
h
h
'
'
−= µω
Paper x.y Speakers name 14
WVTR as a function of T and ω
Observations:1) Linear correlation
between WVTR and mixing ratio
2) Temperature influences slightly on the slope of the WVTR-mixing ratio curve
3) Most likely suitable for regression estimation
WVTR vs. mixing ratio20 g/m2 LDPE coating
23°Cy = 267,24x + 0,5836
R2 = 0,9887
30°Cy = 404,66x - 0,2514
R2 = 0,9953
38°Cy = 532,34x - 2,1323
R2 = 0,9961
45°Cy = 659,17x - 6,5855
R2 = 0,9916
0
5
10
15
20
25
30
35
0 0,02 0,04 0,06 0,08
Mixing ratio
WVT
R (g
/m2 /2
4h)
Paper x.y Speakers name 15
Model development
• Step by step scheme for calculations
Regression
1)
2)
The influence of:
i. Temperature
ii. Mixing ratio
WVTR of 20 g/m2 single layer
The influence of iii. coating weight
3)The influence of multilayers
WVTR of a single layer
WVTR of a multilayer structure
Paper x.y Speakers name 16
• We define temperature, mixing ratio and coating weight as independent variables (x1, x2 and x3, respectively) and WVTR as a dependent variable (y)
• Step 1:Several first- and second-order models were tested to obtain results for 20 g/m2 single layer.Equation including all possible terms:
y = b0 + b1x1 + b2x2 + b3x1x2 + b4x12 + b5x2
2
We apply a spreadsheet or statistical computer program to solve the b-values and reliabilities of different models.
Paper x.y Speakers name 17
List of the tested models and the corresponding standard errors (the best values bolded)
Model Std error LDPE
Std error HDPE
Std error PP
Std error COC
y = b0 + b1x1 + b2x2 0,897 0,590 0,590 0,649 y = b0 + b1x1 + b2x2 + b3x1x2 0,689 0,435 0,355 0,447 y = b0 + b1x1 + b2x2 + b3x1
2 0,908 0,589 0,623 0,626 y = b0 + b1x1 + b2x2 + b3x2
2 0,353 0,203 0,234 0,310 y = b0 + b1x1 + b2x2 + b3x1x2 + b4x1
2 0,440 0,298 0,295 0,346 y = b0 + b1x1 + b2x2 + b3x1x2 + b4x2
2 0,284 0,162 0,223 0,323 y = b0 + b1x1 + b2x2 + b3x1
2 + b4x22 0,279 0,170 0,222 0,320
y = b0 + b1x1 + b2x2 + b3x1x2 + b4x12 + b5x2
2 0,291 0,170 0,232 0,331 y = b0 + b1x1 + b2x1
2 + b3x22 0,519 0,463 0,434 0,313
y = b0 + b1x2 + b2x12 + b3x2
2 0,410 0,238 0,214 0,313 y = b0 + b1x1 + b2x1x2 + b3x2
2 0,660 0,520 0,320 0,310 y = b0 + b1x2 + b2x1x2 + b3x2
2 0,476 0,286 0,278 0,316 y = b0 + b1x1
2 + b2x22 0,782 0,627 0,483 0,303
Paper x.y Speakers name 18
Results of step 1
Model: WVTR=b0+b1*Temp+b2*Mix+b3*Temp*Temp+b4*Mix*Mix
z=(-6,9493)+(,427809)*x+(203,756)*y +(- ,00454)*x*x+(5101,53)*y *y
40 35 30 25 20 15 10 5
16
15
14
12
1113
10
8
9
76
4
5
23 1
WVTR as a function of T and ω for 20 g/m2 LDPE coating
Reliability indicatorsSSE 0,857234S 0,279160S2 0,077930R2 0,999216
Paper x.y Speakers name 19
Step 2
• Influence of coating weightCoating weight has an inverse proportion on WVTRThus
for LDPE
( )( )213
,20 xxfx
y =
( )224
21322110
3
20 xbxbxbxbbx
y ++++=
Paper x.y Speakers name 20
Step 3
• Influence of multilayers• Provided that
All the P-values of the layers are independent of pressure and concentrationThere are no barriers to diffusion due to interfacial phenomena between layers
Multilayer film obeys the equation
• As partial pressure difference stays as a constant in the WVTR test
...3
3
2
2
1
1
PL
PL
PL
PL
tot
tot ++=
...1111++=
321 WVTRWVTRWVTRWVTRtot
Paper x.y Speakers name 21
4. The end result• A Labview based WVTR estimation computer program
User-selected input values:– Temperature (T) – Relative humidity
(RH)– Polymers of layers
1-5 and the corresponding coating weights
Computer aided results:– WVTR of chosen
structure in selected conditions
– 3D graphs; WVTR of chosen structure in different conditions
– WVTR of chosen structure in standard conditions
Paper x.y Speakers name 22
Report of the WVTR calculation program
Paper x.y Speakers name 23
5. Acknowledgements
• Many thanks to the companies that kindly arranged their materials for the study
Stora EnsoBorealis PolymersTopas Advanced Polymers
Thank you! Questions please…