Aluminium Smelter Benefits from MP Consultancy

This article is based on a talk given by Professor Miles Nicholls of Swinburne University of Technology to the Operational Research Society conference.

The Portland Aluminium Smelter

The Portland Aluminum Smelter 200 miles west of Melbourne is the largest aluminium smelter in the southern hemisphere. Built on a greenfield site in the 1980s at a cost of A$1.5 billion, it produces 330 million tonnes of aluminium a year.

The plant is adjacent to a deepwater port and the entire production is exported. The alumina (aluminium oxide derived from the bauxite ore) is imported so the primary local input is electricity. This comes from a power station burning brown coal 400 miles away. The plant is operated by Portland Aluminium Company on behalf of a consortium of owners, including Alcoa of Australia (the parent of Portland Aluminium) with 45%, and the Government of Victoria with 25%.

Professor Miles Nicholls of Swinburne University of Technology, Melbourne, was called in to assist in modelling the plant and improving its performance. Although the plant operates in the private sector, it does not use conventional measures of profit. Portland Aluminium Company does not own the raw materials which are being processed or the aluminium produced. Because of this the plant operates on a value added basis, aiming to produce the maximum amount of aluminium for the smallest quantity of inputs, i.e. to maximise efficiency.

Areas of the Plant

Within the plant there are four areas of the production process which operate as autonomous business units:

  • the pot rooms, where the alumina is reduced to molten aluminium in iron baths by electrolysis (280,000 amps at approximately 4 volts DC);
  • the ingot mill, where aluminium syphoned off from the pots is cast into ingots and stored;
  • the rodding area, where used anodes from the pots are broken up for recycling;
  • the anode area, where fresh anodes are manufactured from old ones, together with coke and pitch.

These are shown in Figure 1.

Figure 1: Schematic of an Aluminium Smelter

Although these areas operate autonomously, they are related by the way in which aluminium is produced. The anodes used in the pots consist mainly of carbon suspended from a metal bar. The process of electrolysis causes the carbon to be consumed. The anodes must be withdrawn from the pots before there is so little carbon left on them that the metal bar could contaminate the electrolyte. Carbon from "used" anodes is also an essential ingredient in making new anodes, so the used anodes must be withdrawn sufficiently early to provide enough "recycled" carbon for the anode area.

Decision Variables

There are two main variables which can be controlled in the system:

  • the electric current kA in the pot rooms;
  • the setting cycle, which is the time for which the anodes remain in the pots.

The quantity of aluminium produced is directly proportional to the kA used. The primary aim of the pot rooms is therefore to maximise kA. The pot rooms also like a high setting cycle, because it means less work for them.

The ingot mill aims to maximize its throughput, within the limits of its capacity. It can therefore be viewed as an extension to the pot rooms, constraining the maximum rate of production.

The rodding area aims to minimize its activity and costs. These are directly related to the number of used anodes which it processes, so it likes to see the highest possible setting cycle.

Similarly, the anode area likes to see a high setting cycle. But its position is more complicated. In a new anode the proportion of pitch is fixed. There are upper and lower limits on the proportion of recycled carbon which must come from used anodes. The amount of recycled carbon available is determined by what has been happening in the pot rooms, so the only free agent is the amount of coke to be used.

Although the plant was new, Portland had followed the established practice among aluminium smelters of allowing the pot rooms to determine both the kA and the setting cycle. There was no general recognition that this could cause problems for the rodding and anode areas and that, in the worst case, a substantial increase in costs could occur.

The Model

The task which Professor Nicholls faced was first and foremost to model the entire plant and determine which were the decision variables and how they related to one another. To build an integrated model he had to overcome conceptual difficulties which were entrenched in the ethos of the operators of aluminium smelters.

The model itself was to serve a variety of purposes: to determine the optimal solution when the plant was running in a steady state; to enable management to perform "what if?" studies; and to assist in planning. It had to be integrated with the plant's existing computer systems. This led to the use of Powerhouse for data handling, Cobol for maths on the Powerhouse data, SAS/OR for the user interface and, eventually, Fortran for the expanded mathematical model and its optimization.

The Results

Figure 2 shows a typical chart of the relationship between the decision variables kA and SC, the setting cycle.

Figure 2: Relationship between kA and SC

There are upper and lower bounds kAMAX and kAMIN on kA, and the setting cycle can lie between SCMIN and SCMAX. The curved line BC represents the relationship between kA and SC that there must be some minimum proportion of the anode left when it is removed from the pot at the end of the setting cycle. The shaded area therefore represents the feasible region for the decision variables, provided the other constraints are satisfied. If this is so, the optimal solution is the point B which represents the best combination of kA and SC.

Now suppose that at point B the solution requires more coke than is available for making new anodes. If we were to reduce the setting cycle while keeping kA constant (i.e. move towards A) this would increase the requirement for coke. We need to decrease it. We cannot increase the setting cycle for this value of kA because we are already hard up against a constraint. We must therefore reduce the kA, moving away from B down the curve BC to the point R. Similarly we need to check that there is the correct proportion of recycled carbon available for making new anodes and this may cause us to move away from point R, potentially into the interior of the shaded region.

Impact of the Model

The development of the model has led to much greater understanding of the way that the plant works and the interactions between the four areas. Pot rooms have engaged in discussions with other areas to determine the setting cycle. Greater cooperation has followed and with it reductions in cost, better working practices and greater output.

Although the model is notionally concerned with finding the "best solution", its greatest benefit has been to promote greater understanding of the way the plant works. This has facilitated many improvements, for instance to the way that maintenance is scheduled. The model has also been used to evaluate possible developments at the plant such as expanding the pot rooms. The model's worth may be judged from the continuing programme of development and research which the plant is sponsoring.

Related articles include Optimizing the Supply of Bulk Gases and Prize-Winning Planning at Harris Semiconductors. To find other articles, refer to the MP in Action page.