Process simulation in VM-CBA alumina refinery
Thiago Franco1, Roberto Seno Jr.2, Rodrigo Moreno3, Caio van Deursen4 and Alexandre Freitas5
1. Process Engineer
2. Bauxite and Alumina Technology Development Manager
3. Process Consultant
4. Process Consultant
Votorantim Metais/CBA, Alumínio SP, Brazil
5. Process Engineer,
AJP Engenharia, Rio Verde GO, Brazil
Corresponding author: [email protected]
Abstract
Modelling of chemical processes has been developed since the seventeenth century with several
authors expressing mathematical relationships between system properties. More recently, with
the development of modern computational capabilities and by applying process models in
numerical simulations, it has been possible to make predictions and evaluate ideas prior to
making any modifications in the operation. Along with these advancements it was possible to
control and optimize processes, to evaluate process improvement projects and also to train
operators. The current work presents process modelling development and results of process
simulation in Votorantim Metais - Companhia Brasileira de Alumínio (VM-CBA) refinery.
Optimisation studies were conducted in the desilication, digestion and precipitation areas, as
well as for future project opportunities such as the conversion of dual stream to single stream
digestion.
Keywords: Process simulation; process modelling; numerical modelling; computational
programming.
1. Introduction
Companhia Brasileira de Alumínio (CBA), of the Votorantim Group, is located in Alumínio, 74
km from São Paulo city. It is the largest integrated aluminium plant in the world. CBA started
to operate in 1955 and today produces 0.475 Mt of primary aluminium.
The Bayer process can be divided into two parts, commonly referred to as the red side and white
side. To briefly summarize, on the red side alumina minerals such as boehmite, diaspore
(Al2O3.H20) or gibbsite (alumina trihydrate Al2O3.3H2O), are dissolved in a caustic solution at
temperatures between 100 and 300 °C, and bauxite residue is separated from pregnant liquor.
On the white side, alumina trihydrate is precipitated from pregnant liquor and afterwards the
removal of structural water from this hydrate is effected in the next major process step of the
Bayer Process, calcination, generating the product we call alumina (Al2O3).
To assist process control and the optimization and the predictability of process variables, tools
and procedures were developed to monitor all areas of the plant, many of them using
spreadsheets, often of great complexity. This existent modeling presents the plant operation and
shows how distant from ideality it is. On the other hand, it is rarely reliably predictive.
Aiming for better predictability and flexibility, CBA teams are now utilizing Kenwalt’s
software SysCAD, to develop improved process simulations.
1
2. Developments
2.1. Pre-desilication and digestion
Pre-desilication and Digestion are both areas in an alumina refinery where significant reactions
and complex energy exchanges take place. Firstly, the bauxite slurry from grinding is heated in
contact heaters with flashed steam from the flash tanks of digestion area. This heated slurry then
flows to a desilication tank (14C) and then to the paste heater batteries, which are heated with
reflashed live steam from digestion area. Afterwards, the slurry goes to other desilication tanks
(14A and 14B) with the some recycle flow to tank 14C. Then, the slurry flow follows to 3
digestion units.
Figure 1. CBA’s grinding, paste heaters and pre-deslication systems on SysCAD interface.
Digestion units include autoclaves that receive the bauxite slurry and the heated caustic liquor.
This equipment setup promotes the ore’s alumina trihydrate (Al2O3.3H2O) dissolution under
elevated pressure and temperature. The digested slurry from the autoclaves then flows to flash
tanks in which the temperature is reduced to avoid boiling in the atmospheric decanters, the next
step of process. The flashed steam is sent to recuperative heaters, where the strong liquor from
the test tanks is heated. After recuperative heating, the strong liquor flows to live steam heaters
and soon after to the autoclaves. The live steam flow to the live steam heaters is controlled
according to the autoclaves’ set point temperature (close to 144 °C). The live steam condensate
is reflashed in pot tanks and its steam is sent to the desilication contact heaters. The excess
steam from second flash tank in each digestion unit is sent to the paste heaters, forming a
complex energy exchange system.
2
Figure 2. CBA’s Digestion system on SysCAD interface.
Simulation of this area provides important information and insights to process, operation and
maintenance teams:
1. Heaters energy transfer performance and cleaning cycle residual life
2. Reaction performances and energy loss in desilication tanks and autoclaves
3. Evaluation of digestion productivity and production
2.2. Precipitation
The cooled pregnant liquor from the Heat Interchange Departments (HIDs, the stage responsible
for cooling the liquor from the red side), flows to the first precipitation tanks, where
agglomeration occurs after adding fine seed. After this stage, the liquor and its agglomerated
solids flow to the precipitation growth stage, via the Interstage Coolers (ISC). These coolers
consist of flash tanks, barometric columns and cooling towers. The growth tanks receive coarse
seed from the classification area. The resulting suspension flows through the precipitation chain
to the last growth tanks and later to the hydrocyclone classification system.
The first hydrocyclone batteries separate the final product from seed which, in turn, are
separated between fine and coarse seed by new hydrocyclone batteries. Both product and seed
3
are filtrated to recover liquor; the product is then sent to calcination, the fine seed to the
agglomeration tanks and the coarse seed to the growth tanks.
2.2.1. Agglomeration
a. Principles
The product size (% -45 μm), product strength and product occluded soda are largely controlled
during agglomeration. In this stage, it is very important that the fine particles are gathered in
bigger clusters, and that these clusters are cemented into strong agglomerates. Moreover, the
generation of fine particles by nucleation is not desirable.
b. Control philosophy
In the hydrate precipitation circuit, due to the dynamics of process conditions, the product
particle size varies and consequently, so do the fine and coarse seed introduced in the process.
As a consequence, the precipitation process cycles with fining or coarsening of the hydrate
particles.
When the precipitation circuit has a tendency towards fining, the following decision is taken,
either separately or altogether:
1. Reducing the fine seed charge.
2. Increasing the temperature in agglomeration.
3. Increasing residence time.
4. Increasing A/C ratio in the beginning of agglomeration.
Figure 3. Block diagram for CBA’s precipitation and classification system.
The first option is more immediate and the others may require longer planning. The coarsening
control is done by the exact opposite measures.
4
CBA’s classification system has three by-passes which allow the transfer of precipitation
product straight to seed classification (1); to the seed cyclones overflow tank (2); or to the seed
cyclones underflow tank (3), as seen in the figure below:
Figure 4. CBA’s hydrate classification schematic diagram.
2.2.2. Growth
a. Principles
During Growth, the main objective is to maximize yield. The linear growth of crystal by means
of hydrate deposition occurs, favored by relatively lower temperatures in comparison with
agglomeration.
b. Control philosophy
At CBA’s refinery, the temperature decrease is assisted by ISCs, which generate new fine
crystals by nucleation.
5
Figure 5. CBA’s precipitation system on SysCAD interface.
2.2.3. Input data system
CBA’s precipitation chain is divided into an agglomeration tank series of 1000 m3, a growth
series with tanks of 3300 and 4400 m3, and growth series tanks at the end of chain of 470 and
1000 m3, totaling 75 tanks. It also includes the cooling system for the growth tanks, the ISCs
(Interstage Cooling).
The model was created as simple as possible, considering the dynamic plant routine of inserting
and removing tanks and changing ISCs position between growth tanks. A system was
programmed that inserts or removes tanks by selecting checkboxes and choosing ISC positions.
Figure 6. Insert/remove tank system on SysCAD interface.
6
The data input to calibrate the modeling is done by transferring historical plant data from a
spreadsheet. Selected pregnant liquor, fine and coarse seed and spent liquor process variables
are sent to software instantaneously.
2.2.4. Kinetic reaction calibration system
The alumina tri-hydrate precipitation reaction, as well as other reactions, is influenced by
countless factors. It is presented as follows:
2NaAlO2 (aq) + 4H2O (l) Al2O3.3H2O (s) + 2NaOH (aq)
The hydrate precipitation driving force (supersaturation), chemically represented by the
difference between the alumina concentration in the liquor (a) and alumina concentration at
equilibrium (a*), is strongly influenced by the liquor caustic concentration. Precipitation
kinetics can be expressed by linear growth rate (G) of hydrate crystals:
G = dL/dt = Kg . f(S) (1)
Where, L is the characteristic crystal length, Kg is the growth constant and S is the
supersaturation. The growth constant is influenced by temperature (T) according with Arrhenius
equation:
Kg = k.exp(-E/RT) (2)
Where, E is the activation reaction energy of precipitation. Numerous modeling correlations
were already published:
Table 1. Kinetic models of hydrate precipitation.
SysCAD has precipitation kinetic models available for precipitation unit simulation.
2.2.4.1. Calibration principle
A value of activation energy (E) has to be adopted, then controllers find values for the growth
constant and for the ambient losses, respecting in the final tank conditions, A/C ratio and
temperatures presented in plant data. To calibrate the model developed, real plant data taken in
stable periods of operation is utilized.
7
2.2.4. Results
The mathematical simulation achieves great results in relation to calibration with plant data in
stable periods. It is possible to compare precipitation production and yield in these two
situations:
Table 2. Comparison between calibration and simulation.
Production (t/day) Yield (g/L)
Period Calibration Simulation Difference Calibration Simulation Difference
1 2713 2690 -1 % 64,7 64,2 -1 %
2 2035 2135 5% 60,1 62,8 4%
3 2494 2596 4% 69,5 72,1 4%
4 2297 2412 5% 66,3 69,4 5%
5 2529 2702 7% 68,0 72,4 6%
6 2398 2384 -1 % 65,8 65,4 -1 %
7 2129 2255 6% 59,3 62,6 5%
8 2657 2822 6% 67,6 71,7 6%
The observed difference between simulation and calibration, both for production and yield, is
fairly low, with values of 5 % on average.
3. Conclusions and Perspectives
Process simulation is a tool of great importance for refinery operation, being useful to perform
mass and energy balances, predict process behavior, calculate equipment sizing and train control
room operators. It is also essential in projects for increasing plant capacity or to implement new
technologies, supporting the estimation of the benefits to the company.
Some examples of project simulation in VM-CBA:
1. Single Stream Digestion: estimate the steam savings and predict digestion heater
performance due to the addition of bauxite slurry prior to the recuperative heaters.
2. Filter Press: estimate caustic liquor recovery, energy and mass balances after adding
press filters for the de-liquoring of bauxite residue.
3. Decanter Centrifuge: estimate the alumina recovery by adding a decanter centrifuge
receiving the thickener underflow and sending it to the first washer.
The development of modelling for process simulation at CBA has been extremely satisfactory,
with differences between simulation and model calibration data of 5 % on average. It is
expected that process simulation will become an increasingly effective tool in the refinery.
8
References
1. Misra, C. and White, E. T. “Kinetics of Crystallization of Aluminium Trihydroxide from
Seeded Caustic Aluminate Solutions”, Chemical Engineering Progress Series 67 1970 Vol
110 pp 53-65.
2. King, W. R., “Some Studies in Alumina Trihydrate Precipitation Kinetics”, Light Metals
1973 Vol 2 pp 551.
3. White, E.T. and Bateman, S.H. "Effect of Concentration on the Growth Rate of Al(OH)3
Particles", Light Metals 1988 pp 157–162.
4. Veesler, S. and Boistelle, R., “About Supersaturation and Growth Rates of Hydragillite
Al(OH)3 in Alumina Caustic Solutions”, Journal of Crystal Growth 1993 Vol 130 pp 411-
415.
