Improving process optimisation with predictive modelling

February 27th, 2017

Dr Mark CookseyBy Dr Mark Cooksey

In high volume metal production, process optimisation can be an expensive and time-consuming exercise. While improving efficiency by a small percentage would result in significantly higher returns, you may find that the risks and costs associated with process optimisation are often too large to justify.

Plant processing efficiencies are often highlighted when conducting site comparisons and setting benchmarking targets within your operation. If another operation is achieving higher returns performing the same function, you need to understand why this is occurring, and how you can achieve similar performance.

To achieve these high returns, you need to undertake well-controlled process optimisation. However, well-controlled optimisation using a trial and error approach is difficult and time-consuming to execute in a large operaton where there are significant variation in inputs and conditions.

In place of these optimisation methods, which rely on process conditions to be repeatedly varied and the resulting process performance to be measured, you could implement predictive modelling, such as computer fluid dynamics (CFD) modelling. A shift towards this method of optimisation allows you to build a representation of your process, and more easily alter variables to rapidly gain a reliable view of how any process changes will affect performance.

Managing your risk

While CFD modelling does minimise the risks associated with process optimisation, it is not entirely risk free. You need to:

  • Identify the important parameters in your process for inclusion in the model.
  • Establish reliable baseline data.
  • Validate model predictions of baseline and improved performance, with laboratory and plant measurements.

The benefits of CFD modelling

Computer fluid dynamic model of alumina concentration

CFD model prediction of alumina concentration in an aluminium reduction cell.

CFD modelling allows you to rapidly investigate the effect of multiple variables on process performance, without having to devote time to multiple plant trials. The benefits include:

  • Investigations are not confounded by process variation, a common issue in plant trials, so you can obtain higher confidence in the results.
  • Process changes can be investigated without making expensive and time-consuming plant modifications
  • Process optimisation is accelerated. You no longer have to wait for the results of multiple plant trials. Far more process conditions can be investigated.
  • Tests that would otherwise be forgone due to high risk can be conducted.
  • You are able to understand your entire process and, as a result, you can take a more active and innovative approach to process optimisation and even redesign.

CSIRO’s CFD Modelling Solutions

CSIRO has extensive CFD modelling capabilities which have been used to accelerate process optimisation and minimise risk

For example, CSIRO has used CFD modelling to optimise alumina feeding in aluminium reduction cells, where there are complex flows of gas and molten bath. There are typically 4-6 alumina feeders in a cell, so it possible is to adjust the amount of alumina fed from each feeder to better balance the alumina concentration throughout the cell.

Some smelters have attempted to optimise the alumina feeders through plant trials, but this is time-consuming (months) and it can be difficult to be confident in the results, because of other variation occurring in the process

CSIRO has built a CFD model of an aluminium reduction cell, and used this to evaluate the effect of different feed rates on alumina distribution in the cell. It was found that using different feed rates for different feeders could produce a more even alumina distribution. This understanding was reached much quicker than it would have been through plant trials.

Contact the Responsible Metal Production and Recycling team on +61 3 9545 8865 or email, and let us assist you in planning and executing your process optimisation using predictive modelling.

Subscribe and receive updates on our sustainable metal and processing research.