R B Kahane

Worsley Alumina Pty Ltd

PO Box 344



Computational Fluid Dynamics (CFD) Modelling of process unit operations is a tool that is being used increasingly by mineral processing industries to reduce operating and capital costs and increase throughput.

Worsley Alumina first became involved with CFD modelling through AMIRA thickener project 266A in Q4 1994 with the CSIRO Division of Minerals as part of that project.

The benefits of obtaining a better understanding of flow patterns in thickeners using this type of modelling became obvious and a number of one on one projects commenced in Q3 1995 and are still continuing.

The CFD modelling has led to development of innovative changes in settlers and washers that have resulted in improved process stabilisation, reduced chemical costs, and very large savings in capital requirements for our major expansion.

The paper outlines details of an individual CFD modelling project and cost benefits achieved from completed projects.




R B Kahane


The Bayer alumina process consists of a number of unit operations that are continuous and interconnected. These individual unit operations can successfully be modelled using computational fluid dynamics (CFD) to improve understanding of flow patterns of the multivariable streams involved.

The results of the CFD modelling can then be tested by tracer studies and if validated can be used to develop process improvements in that unit operation.


The Worsley Bayer alumina refinery process consists of four major steps: Digestion of ground bauxite in hot caustic solution, separation of the residue from the liquor in settlers, seeding the cooled supersaturated liquor and growing the hydrate and calcining the hydrate to product. An overview of the Bayer Circuit is shown in Figure 1.




2.1 Residue Handling Capacity

Most of the development work using CFD modelling has occurred in the clarification area, where the residue is separated from the liquor and washed to recover soda, and the causticisation area. This has occurred because of the need to handle greater residue flow


without an increase in major equipment. The increased capacity achieved in the residue washing area has been described in a number of previous publications (Kahane et al 1997, Kahane 1998).

Figure 2 is a plot of the digestion residue, which indicates the extra handling capacity required over the last ten years.


In Causticisation, lime is added to heated first washer overflow liquor in a mixing reaction tank. The lime forms calcium carbonate and converts sodium carbonate (soda) to sodium hydroxide (caustic). The reaction tank discharge is fed to the causticiser settler so that the solids can be separated from the liquor before the causticised overflow dilution liquor is pumped to the settlers.

3.1 Feed Flow

Higher production rates from an increase in digestion flow requires a corresponding increase in dilution flow from the causticiser to maintain target settler liquor caustic concentration. This increase in dilution must pass through the causticiser settler and therefore the chances of process upsets are increased from the higher flow. Figure 4 shows the increase in flows that has occurred.


3.2 Peripheral Feed Design

The original causticiser settler design is shown in Figure 5. Problems were experienced at higher flow rates due to the increase in rising velocity and the difficulty of achieving good flocculation.

A number of attempts had been made to convert the causticiser settler to centre feed because of the increase in liquor flow through the tank and the need to increase lime addition to the causticiser reaction tank. This increase to the reaction tank solids feed also increases the solids feed to the causticiser settler.

3.3 Lime Flow

The lime flow to the causticiser reaction tank (Figure 6) was often constrained because the rake torque, tank clarity or overflow solids were excessive and therefore causticisation could not be optimised especially at high flow. This in turn caused a production loss by allowing the caustic to soda ratio in the liquor to fall. Lime flow is now controlled by process chemistry requirements.

The main factor preventing the tank from being switched from peripheral to centre feed was poor flocculant control. Initially starch was used when the tank was first commissioned because the reaction tank did not exist and slow settling was required so that the causticisation reaction could take place in the tank. After the reaction tank was commissioned in 1988, flocculant was used for a time to improve underflow density but because it was added to the feed tank or dilution line with foam and solids present it was not very effective. An increase in underflow and overflow pump capacity was carried out in 1991 and an attempt made to switch the tank to centre feed with the flocculant dosing into the feed line. Control was not achieved and the tank remained on peripheral feed. Starch dosing was again used to reduce rake torque. Overflow solids again became a problem – Flocculant was again used from mid 1995.

3.4 Rake Torque

A good indication of the performance of the causticiser settler is the rake torque. If the lime residue is properly flocculated and forms a compacted bed in the tank that is easily raked to the periphery of the tank for discharge then a relatively steady rake torque will be obtained. Figure 7 is a plot of historic rake torque data.

The rising rake torques of 1996 and 1997 are an indication of poor tank control. The rake torque became steadier once conversion to centre feed had taken place. During the cleaning outage in Q3 1998 the electric drive motors on the rake hydraulic power pack were upgraded and this has contributed to the lower rake torque. This upgrade had been desired since 1992 but due to unscheduled rake failures could not be planned into the tank maintenance programme.


Because of the location and design of the peripheral feed system (Figure 5) a trial to switch the tank to centre feed has to be successful within a few days because the peripheral feed line starts to block up rapidly when no flow passes through it.


An extensive CFD modelling programme had been carried out between 1995 and 1997 on the settlers and washers of the CCD circuit. The modelling of the settler was validated using a lithium ion tracer added to the feed line and sampled from the feedwell at four locations. The averaged normalised data of lithium analysis confirmed the CFD model solids flow pattern predictions. This was despite the fact that a number of assumptions have been made in development of the model. These include a simple flocculant adsorption step, a single average particle diameter and a fixed settling rate.

With the successful outcome of CFD modelling and implementation on the washers to significantly increase their capacity, an opportunity was seen to model the causticiser settler and determine if it could be switched to centre feed.

4.1 Causticiser Settler CFD Modelling

The modelling was subsequently carried out but it was difficult to determine flow patterns in the body of the tank and therefore a risk of rising bed levels was a concern. Loss of causticisation efficiency in the causticisers from the higher settling rate was also a possibility and extensive sampling was carried out to determine base case chemical reaction kinetics. Because of the confidence that had been developed in previous implementations of the model outcomes no tracer tests were carried out to confirm the model predictions. This was relatively safe in this application, but elsewhere others have been caught by relying too heavily on unvalidated models (Brown 1997). Figure 8 shows the solids flow pattern in the feedwell.


On the positive side the CFD modelling of the causiticiser feedwell clearly showed where the natural upward dilution would enter the feedwell,(Johnston et al 1996) and therefore where the flocculant sparge would need to be located. These upward dilution regions are the darker areas in Figure 9 and are located on either side of the feed inlets. The floc sparge location was chosen to be in an area of consistent flow and reduced scaling risk.


The flocculant sparge locations suggested by the CFD modelling came as a surprise to those with extensive feedwell flocculation expertise and was different than expected due to the lower S.G. of the calcium carbonate compared with bauxite residue. The flocculant sparge and pipe were installed while the tank was out of service for cleaning in June 1997.

No other modification was required and the tank was brought into service on peripheral feed. Approximately two weeks later the tank was switched to centre feed and the flocculant sparge brought into service.

4.2 Flocculant Dose Rate

High rake torques were initially experienced but once flocculant dose had been cut by 80% torques fell to normal levels.

Figure 10 shows the reduction of flocculant that occurred when the tank was switched to centre feed and the flocculant sparge location selected based on CFD modelling predictions used.

The tank ran at more stable operation than had previously been experienced for the required twelve months service life.

Causticiser settler overflow solids have reduced since conversion to centre feed although excursions still occur. These are now caused by either poor settling lime reaction product low liquor temperature or excess residue in the feed rather than poor flocculation in the causticiser settler.

4.3 Future Design Changes

This causticiser settler process change is one of the many successful CFD projects that have been implemented at Worsley. Further CFD modelling work is expected on the causticiser settler to cope with expected increased flows that can not be handled with current design.

Therefore to maintain required clarity from more lime addition at increased flows, further feedwell modifications will be required. These will be developed by suggesting to the modeller a number of practical alternative changes which when CFD modelled will indicate the best option. It is possible that the CFD modelling may show that feedwell modifications will not achieve the required results and CFD modelling to increase the tank size may be the next step. The final step may be the construction of a second causticiser settler at considerably more cost than simple feedwell modifications.

In this section one of the many successful CFD modelling projects have been described.


Table 1 shows the CFD modelling history at Worsley from commencement through the AMIRA 266A thickener project in Q4 of 1994 up to the date of writing this paper.

It took almost two years to implement the first modification on a settler based on the CFD modelling. Once confidence had been gained implementation usually took no more than three – six months after commencement of modelling and some of these projects cost significant amounts of money to implement on a trial basis. No pilot scale testing is carried out on these design changes. Development changes go straight from CFD modelling to full scale implementation.

Table 1

CFD Modelling History



Model Date



Tracer tests / CFD

Q4 1994



Improved Settler Design

Q3 1995 – Q2 1996

Q3 1996 – Q3 1998

Spare Settler Washer

Significantly Increased Capacity

Q2 1996

Q3 1996

Washer (1st)

Significantly Increased Capacity

Q3 1996

Q4 1996 – Q4 1997

Causticiser Settler

Centre Feed

Q1 1997

Q2 1997

Seed Thickener

Improve O/Flow Clarity

Q1 1998

Q2-Q4 1998

Washer (3rd & 4th)

Significantly Increased Capacity

Q1 1998

Q2 1998 – Q3 1999

Seed Thickener

Increase Capacity

Q3 1998

Q1 1999 ?

Causticiser Settler

Increase Capacity

Q4 1998 ?

Q3 1999 ?

5.1 Cost Benefits

The bottom line is the cost benefit with any project (Batterham et al 1997). CFD modelling projects have the cost of the modelling plus the costs of capital to implement the project.

Benefits from the projects listed in table 1 include substantial increases in liquor and solids feed capacity with improved control in settlers washers and thickeners. Chemical costs have also been reduced. Because of these reduced costs and increased capacity , the projects have adequate cost benefits.

An annual consultant budget is maintained by Worsley for CFD modelling because as soon as a project is successfully completed another project is developed. Some of the projects proposed are quite different from those already tested.


CFD modelling has maintained Worsley’s position in the forefront of innovative change to process equipment and allowed us to gain a greater understanding of flow patterns and fluid interactions during flocculation, mixing and solid liquid separation processes.

The CFD modelling has been so successful and cost effective that continued support for new CFD modelling projects is assured.


The author wishes to acknowledge the contribution of T.Nguyen (CSIRO) for carrying out the CFD modelling runs and to CSIRO Minerals for permission to publish figures 8 & 9.



R.J.Batterham and G.J.Hardie, "CFD and the Real Needs of the Minerals and Metals Processing Industry", Proc Computational Fluid Dynamics in Minerals and Metal Processing and Power Generation 1997 Melb.

G.J.Brown, "Numerical Simulation of Separated Flows in Three-Dimensional Industrial Geometries: A Case Study", Proc Computational Fluid Dynamics in Minerals and Metal Processing and Power Generation 1997 Melb.

R.M.Johnston, T.Nguyen, M.P.Schwarz and K.Simic, "Fluid Flow and Natural Dilution in Open-Type Thickener Feedwells", 4th International Alumina Quality Workshop 1996 Darwin.

R.B.Kahane, M.P.Schwarz, R.M.Johnston, "Residue Thickener Modelling at Worsley Alumina", Proc Computational Fluid Dynamics in Minerals and Metal Processing and Power Generation 1997 Melb.

R.B.Kahane, "Control Development Alternatives", 4th International Conference on Optimization : Techniques and Applications (ICOTA’98) 1998 Perth.