Effect of Baffle Design Parameters on Fluid Dynamic Response of a Coal Classifier

Journal of Mining World Express Volume 3 Issue 1, January 2014 www.mwe‐journal.org

doi: 10.14355/mwe.2014.0301.03

15

Effect of Baffle Design Parameters on Fluid

Dynamic Response of a Coal Classifier Hamid Khoshdast*1, Hami Khoshdast2, Vahideh Shojaei1

1 Mining Engineering Department, College of Mining and Industry, Bahonar University, Zarand, 77611‐56391 Iran 2 Mechanical Engineering Division, INVENTIVE® Mineral Processing Research Center, Zarand, 77611‐56391 Iran

*1 [email protected]; [email protected]

Abstract

The effect of baffle design parameters including length,

slope, shape, and installation position on fluid dynamic

behavior of a coal classifier was simulated using

computational fluid dynamic approach. General response of

classifier at different conditions was interpreted by referring

to turbulent patterns inside the sorting column and fluid

velocity magnitude in products outlets. Results showed that

the effect of design parameters is directly influenced by

other operating factors such as feed pulp flowrate. The

velocity of discharge streams can be used for the prediction

of particle size variations of products due to any change in

fluid dynamic pattern in classier.

Keywords

Hydraulic Classifier; Coal; Baffle; Computational Fluid Dynamic;

Modeling

Introduction

Classification is a method of separating mixtures of

minerals into two or more products on the basis of the

velocity with which the grains fall through a fluid

medium. In mineral processing, this is usually water,

and wet classification is generally applied to mineral

particles which are considered too fine to be sorted

efficiently by screening (Wills and Napier‐Munn, 2006).

Hydraulic classifiers are used in many coal and

mineral applications such as fine coal classification,

the removal of clay fines from siliceous sands, particle

size control in closed circuits with mills, fine control in

taconite pellet washing, dewatering of coal tailing

prior to centrifugation, silica removal from iron ores,

cement purification, etc. (Taggart, 1945; Sarrafi, 1987;

Kelly and Spottiswood, 1989; Priester et al., 1993;

Vijayendra, 1995; Nematollahi, 2002).

Classifiers consist essentially of a sorting column in

which a fluid is rising at a uniform rate. Particles

introduced into the sorting column either sink or rise

according to whether their terminal velocities are

greater or lesser than the upward velocity of the fluid.

The sorting column therefore separates the feed into

two products. An overflow consisting of particles with

terminal velocities lesser than the velocity of the fluid

and an underflow or spigot product of particles with

terminal velocities greater than the rising velocity

(Wills and Napier‐Munn, 2006).

Numerous researches are available in which the

influence of different operating parameters on the

performance of wet hydraulic classifiers has been

investigated (Bhardwaj et al., 1987; Klumpar, 1992;

Heiskanen, 1996; Honaker and Mondal, 2000; King,

2001; Galvin et al., 2005; Zhou et al., 2006; Feng et al.,

2008; Galvin and Zhou, 2012; Johansson and Evertsson,

2012; Tao et al., 2012). The effects of these parameters

on the cut point (d50) as a representative criteria of

classifier performance can be summarized as follows:

The cut point is directly influenced by feed

particle size;

As solid content increases, the cut point

increases due to the effect of hindered‐settling

condition;

Density difference between the particles has a

pronounced effect on classification, especially

in coarser size ranges. In general, the cut point

decreases as density of the particles increases;

The cut point increases by increasing hydraulic

(feed) flowrate;

At constant feed flowrate, two phenomena lead

to decreasing the cut point by increasing

volume of the classifier. First, large classifiers

provide a sorting column with less turbulent

regime inside the classifier. Second, particles’

retention time decreases with increase of

volume. Both of these effects improve efficiency

of classification process; therefore, the cut point

is inversely related to volume of the classifier.

These parameters influence the fluid velocity and the

turbulent condition in sorting column. One practical

solution is to install a barrier, namely baffle, inside the

column to control and moderate turbulent stream of

www.mwe‐journal.org Journal of Mining World Express Volume 3 Issue 1, January 2014

16

input pulp. The design features of baffle are key

factors which directly affect the performance of

classifier.

The aim of this paper is to investigate the effect of

design manipulation of baffle’s structure and position

on hydraulic behavior of a classifier at Zarand coal

washing plant (Kerman, Iran) using computational

fluid dynamic approach.

Modeling of Hydraulic Classifier

Theory of the Model

It is well known that the best choice for modeling flow

patterns in water‐base separators is to apply

incompressible Navier‐Stokes equation in combination

with a turbulent flow model. To predict the fluid flow

pattern in a classifier the governing equation consists

of the continuity and momentum balance equations

for the liquid phase as follow:

. 0v (1)

.( ) .( )vv p g

(2)

where ρ is the fluid density, g is the gravity, v is the

velocity of fluid and p is the static pressure. The stress

tensor τ can be calculated as below:

22( ) .

3effective v v

(3)

effective t (4)

where μ and μeffective are dynamic and effective

viscosities, respectively (Singh et al., 2006; Bhaskar et

al., 2007). The momentum equation can be solved

using a turbulent flow (TF) model. The standard k‐ε

dispersed turbulence model is a TF model commonly

used for engineering purposes. Variables k and ε

characterize turbulent kinetic energy and turbulent

dissipation rate, respectively. The k‐ε model is solved

based on equations:

!

( ) ( ) ( )! !

ii k

i j k j

μn kρk ρku μ G ρε

t x r n r x σ x

(5)

2

1 2( ) ( ) ( ) ( )ii ε k ε

i j ε j

μ k ε ερε ρεu μ C G C ρ

t x x σ x k k

(6)

where Gk is the kinetic energy due to velocity gradient

and μt is the viscosity of turbulent flow. These

parameters can be calculated as follows:

' ' jk i j

i

uG u u

x

(7)

2

t

kC

(8)

where uʹ is the velocity vector and C1ε, C2ε, σk, σε and

Cn are constant values (Nakhaei and Sam, 2009).

Modeling Procedure

The effect of structural adjustment of baffle on

hydraulic response of a coal classifier was investigated

using computational fluid dynamic (CFD) approach.

Design parameters included baffle length, l (0, 1.95

and 3.9 m), slope of baffle sheet, θ (30 and 60 degree

for length 1.95 m), baffle curvature (flat and curved),

and installation position. For each condition, two

different pulp flowrate was considered (0.05 and 1.55

m3/s). Fluid property was fixed for a pulp with 15

wt.% solid content (practical value was 16±2 wt.%).

The studied classifier was fed by a fine feed with

characteristic size (d80) of about 400±20 μm. Fig. 1

shows a schematic illustration of classifier studied in

this paper.

FIG. 1 SCHEMATICS OF HYDRAULIC CLASSIFIER AT ZARAND

COAL WASHING PLANT

The simulation used the steady state, pressure based,

implicit formulation of Fluent 6.3 software which

employs finite volume method and the physical

meshing of the classifier was constructed in pre‐

processor Gambit 2.3. The initial and boundary

conditions were set on the basis of practical data


17

measured in the plant: pulp velocity of 0.009758 and

0.3026 m/s for pulp flowrates of 0.05 and 1.55 m3/s,

respectively, pulp density of 1.099 kg/m3 and viscosity

of 1.022×10‐3 kg/ms, and atmospheric pressure (1 atm).

In order to approximate more accurate results the

residual convergence and iteration values were fixed

at 1×10‐5 and 10000, respectively (Dehghani‐Sanij, 2008;

Golshahifar, 2009).

Validation of the Model

The validation process of the model was done using the experimental data reported by Shariat and

coworkers (Shariat et al., 2010) from a laboratory

hydraulic classifier. In this regard, the water velocity

in overflow discharge gate was plotted against

predicted values. As shown in Fig. 2, the predicted

values are in good agreement with experimental

measurements. This confirms the accuracy of the

model chosen.

FIG. 2 COMPARISON BETWEEN EXPERIMENTAL DATA AND

SIMULATION RESULTS

(a) l = 0 m, Q = 0.05 m3/s (b) l = 1.95 m, Q = 0.05 m3/s (c) l = 3.9 m, Q = 0.05 m3/s

(d) l = 0 m, Q = 1.55 m3/s (e) l = 1.95 m, Q = 1.55 m3/s (f) l = 3.9 m, Q = 1.55 m3/s

FIG. 3 VELOCITY PATTERN INSIDE THE CLASSIFIER FOR DIFFERENT BAFFLE LENGTHS AND PULP FLOWRATES


18

FIG. 4 PULP VELOCITY MAGNITUDE AT OVERFLOW AND UNDERFLOW DISCHARGES FOR DIFFERENT BAFFLE LENGTHS

(a) θ = 60°, Q = 0.05 m3/s (b) θ = 30°, Q = 0.05 m3/s

(c) θ = 60°, Q = 1.55 m3/s (d) θ = 30°, Q = 1.55 m3/s

FIG. 5 VELOCITY PATTERN INSIDE THE CLASSIFIER FOR DIFFERENT BAFFLE SLOPES AND PULP FLOWRATES

Simulation Results and Discussion

Effect of Baffle Sheet Length

Fluid velocity patterns inside the classifier simulated

for different baffle lengths are shown in Fig. 3. As seen,

turbulency inside the column decreases by increasing

the baffle length, especially at high pulp flowrate.

These models clearly show the action of baffle to

control the turbulent environment. The variation in

pulp velocity of overflow and underflow is shown in

Fig. 4. Baffle decreases the turbulency by breaking and


19

dispersing the pulp stream lines. Long baffle

moderates the turbulency in sorting column and this,

in turn, decreases the underflow velocity. In contrast,

long baffle prevents pulp to freely flow inside the

column. Therefore, stream lines concentrate above

baffle sheet and are directed to overflow outlet. This

will increase the velocity of overflow. Since the

particle size distribution in products directly follows

the fluid velocity in outlets, it would be expected that

the fraction of coarse particles in overflow product

increases by increasing of baffle length. This means

that classifier cut‐size should be controlled by

manipulation of feed flowrate and/or solid content (i.e.

pulp density and viscosity).

FIG. 6 PULP VELOCITY MAGNITUDE AT OVERFLOW AND UNDERFLOW DISCHARGES FOR DIFFERENT BAFFLE SLOPES

(a) Flat sheet, Q = 0.05 m3/s (b) Curved sheet, Q = 0.05 m3/s

(c) Flat sheet, Q = 1.55 m3/s (d) Curved sheet, Q = 1.55 m3/s

FIG. 7 VELOCITY PATTERN INSIDE THE CLASSIFIER FOR FLAT AND CURVED BAFFLES


20

Effect of Baffle Slope

Effect of baffle slope on fluid dynamic response of

classifier is shown in Fig. 5. At low pulp flowrate and

baffle slope, stream lines are directed toward overflow

outlet and thus, the sorting column experiences lower

turbulency. Under this condition, fluid velocity

decreases and increases in overflow and underflow,

respectively (Fig. 6). At high pulp flowrate, classifier

gives different velocity patterns. At higher slope, there

is enough space above baffle sheet for formation of

pulp vortex, whereas at lower slope, high speed

stream of pulp will directly encounter the classifier

wall and flows toward the sorting column. Therefore,

both turbulency inside the column and underflow

velocity increase (Fig. 6). As seen from Fig. 6, changing

of baffle slope at high pulp flowrates has less affected

the fluid velocity in outlets and consequently,

products particle size. This result offers that at high

flowrates, changing of baffle length is more preferred

for controlling the particles distribution.

Effect of Baffle Curvature

Fig. 7 shows the effect of baffle sheet curvature on

classifier velocity pattern. At lower pulp flowrate, the

effect of curved baffle on column turbulency and

therefore, velocity of product streams (Fig. 8) can be

neglected. However, stream density above the curved

baffle has fairly increased. In contrast, baffle

deformation has significantly increased turbulent

pattern inside the classifier fed by high speed pulp

Curved form of baffle makes fluid streams to flow into.

sorting column faster than flat baffle and to form a

vortex pattern (capillary effect). Increased turbulency

inside the column will then increase the underflow

velocity to some extent. In addition, limited space

above curved baffle would break fluid stream lines

above the baffle and consequently, decrease the

velocity in overflow outlet (Fig. 8).

FIG. 8 PULP VELOCITY MAGNITUDE AT OVERFLOW AND UNDERFLOW DISCHARGES FOR DIFFERENT BAFFLE SHAPES

Position A Position B Position C

FIG. 9 DIFFERENT POSITIONS CONSIDERED FOR CFD SIMULATION OF CURVED BAFFLE


21

(a) P(A), Q = 0.05 m3/s (b) P(B), Q = 0.05 m3/s (c) P(C), Q = 0.05 m3/s

(d) P(A), Q = 1.55 m3/s (e) P(B), Q = 1.55 m3/s (f) P(C), Q = 1.55 m3/s

FIG. 10 EFFECT OF BAFFLE POSITION ON VELOCITY PATTERN INSIDE THE CLASSIFIER FOR DIFFERENT PULP FLOWRATES

FIG. 11 PULP VELOCITY MAGNITUDE AT OVERFLOW AND UNDERFLOW DISCAHRGES FOR DIFFERENT BAFFLE POSITIONS

Effect of Baffle Position

Effect of baffle position on fluid dynamic of classifier

was also investigated. Three different installation

points were considered for a curved baffle with

equivalent radius. These positions are illustrated in Fig.

9. As seen from simulation results shown in Fig. 10,

turbulency has been enhanced for positions B and C

since feed pulp directly flows to the sorting column.

Under these conditions, it would be expected that


22

underflow velocity increases and the velocity in

overflow outlet decreases (Fig. 11). It is also seen from

Fig. 11 that at high feed flowrate, the velocity in

underflow outlet is nearly equal for positions B and C.

This could likely be interpreted as follows; for both

classifiers, the velocity of feed pulp is equal and very

fast which directly flows into column after encounter

the curved interior face of baffles and a turbulent

saturated environment is formed inside the sorting

column. This saturated stream pattern is not observed

for position A.

Conclusions

The following points can be highlighted from the

present study:

1. CFD‐based simulation gives useful information

about fluid dynamic response of hydraulic

classifier following any operating and/or

design manipulation.

2. The action of baffle to reduce the turbulence

conditions inside the classifier is dependent on

various parameters. Baffle for a specific

classifier should be designed on the basis of

operating parameters and requested perform‐

ance of the classification practice. CFD

approach could be a useful tool for optimum

selection of baffle design aspects.

3. Effects of design parameters are directly

influenced by operating factors.

4. Results obtained from changing baffle

installation position may lead to the conclusion

that the velocity and thus, volumetric rate of

products streams is independent from

hydraulic and turbulent conditions inside the

classifier when feed stream is directed to

sorting column without any intermediate.

There are many other operating parameters such as

pulp solid content, feed particle size distribution, etc.

that affect the hydraulic performance of classifiers.

Further studies are required to investigate these effects.

ACKNOWLEDGMENT

Technical supports from Zarand Coal Washing Plant

and INVENTIVE® Mineral Processing Research Center

are acknowledged.

REFERENCES

Bhardwaj, O.P., Bandyopadhyay, A. and T.C. Rao. “Plant

Performance Studies on a Hydraulic Cone Classifier as a

Secondary Classifier.” International Journal of Mineral

Processing 21 (1987): 217–223.

Dehghani‐Sanij M.A. Numerical Simulation with Fluent 6.3

Software. Tehran: Naghoos‐e Andisheh Publisher, 2008.

Feng, Y., Liu, J. and S. Liu. “Effects of Operating Parameters

on Flow Field in a Turbo Air Classifier.” Minerals

Engineering 21 (2008): 598–604.

Galvin, K.P. and J. Zhou. “Application of the Reflux

Classifier for Measuring Gravity Recoverable Product.”

In Separation Technology, edited by C.A. Young and

G.H. Luttrell, 153–162. New York: SME, 2012.

Galvin, K.P., Callen, A., Zhou, J. and E. Doroodchi.

“Performance of the Reflux Classifier for Gravity

Separation at Full Scale.” Minerals Engineering 18 (2005):

19–24.

Golshahifar M. Practical Fluent. Tehran: Sanei Shahmirzadi

Publication, 2009.

Heiskanen, K.G.H. “Developments in wet classifiers.”

International Journal of Mineral Processing 44–45 (1996):

29–42.

Honaker, R.Q. and K. Mondal. “Dynamic Modeling of Fine

Coal Separations in a Hindered‐Bed Classifier.” Coal

Preparation 21 (2000): 211–232.

Johansson, R. and M. Evertsson. “An Empirical Study of a

Gravitational Air Classifier.” Minerals Engineering 31

(2012): 10–16.

Kelly E.G., and Spottiswood. Introduction to Mineral

Processing. Queens Land: John Wiley & Sons, 1989.

King, R.P. Modeling & Simulation of Mineral Processing

Systems. England: Butterworth‐Heinemann, 2001.

Klumpar, I.V. “Measuring and Optimizing Air Classifier

Performance.” Separation Technology 2 (1992): 124–135.

Nakhaei F., and A. Sam. “Modeling Stream Patterns in

Hydrocyclones Using Fluent Software.” Paper presented

at the third Mining Engineering Conference, Yazd, Iran,

2009.

Nematollahi H. Mineral Processing. Tehran: Tehran

University Press, 2002.

Priester M., Hentschel T., and B. Benthin. Tools for Mining:

Techniques and Processes for Small Scale Mining.

Munich: Friedr. Vieweg & Sohn Verlagsgesellschaft mbH,

1993.

Sarrafi A.R. Mineral Separation Device Supported by

Hydrocyclone Mechanism with Closed Circuit. Kerman:


23

Industrial and scientific research organization, 1987.

Shariat E., Khoshdast H., Sam A., and S.A. Jarkani.

“Combined Experimental and Simulation Analysis on

Performance of a Bench Scale Hydraulic Classifier.”

Paper presented at the second National Conference of

Modern Researches in Chemical Engineering, Mahshahr,

Iran, 2010.

Singh V., Srivastava S., Chaval R., Vitankar V., Basu B., and

M.C. Agrawal. “Simulation of Gas‐Solid Flow and

Design Modifications of Cement Plant Cyclones.” Paper

presented at the fifth International Conference of CFD

Process Industries, Queens Land, Australia, 2006.

Taggart A.F. Handbook of Mineral Dressing. New York:

John Willey & Sons, 1945.

Tao, Y., Wang, L., Li, Z. and M. Sun. “Study on Coarse Slime

Separation by Teeter Bed Separator in Chenglao Coal

Preparation Plant.” In Separation Technology, edited by

C.A. Young and G.H. Luttrell, 177–180. New York: SME,

2012.

Vijayendra H.G. Handbook of Mineral Processing. New

Delhi: New Delhi Publication, 1995.

Wills B.A., and T. Napier‐Munn. Mineral Processing

Technology. Amsterdam: Elsevier Science and

Technology, 2006.

Zhou, J., Walton, K., Laskovski, D., Duncan, P. and K.P.

Galvin. “Enhanced Separation of Mineral Sands Using

the Reflux Classifier.” Minerals Engineering 19 (2006):

1573–1579.

Effect of Baffle Design Parameters on Fluid Dynamic Response of a Coal Classifier

Documents

Transcript of Effect of Baffle Design Parameters on Fluid Dynamic Response of a Coal Classifier