A pilot project was launched at BHP Billiton Worsley Alumina which investigated the potential of an automated uni- and multivariate statistical process control system. The potential of such an automated system was evaluated by selecting a well-understood facility for the pilot and comparing its findings against expert process knowledge.
The causticisation facility was selected, which is part of the first phase of the Bayer alumina process. The causticisation process converts carbonate to caustic soda through the addition of lime. The project was awarded to CSense Systems and Motherwell Automation and the solution was implemented using CSense software.

In the initial stage of the project, troubleshooting was done on historical data. This consisted of data preparation and cleaning, visualization and modeling of data to analyse relationships and determine possible causes of poor performance. After this a knowledge base was established, a system was implemented which automatically and rapidly processes data in real-time. The state of the system was summarised in easily understood web-based dashboard reports. These reports included multivariate causal analyses, expert rules and univariate monitoring statistics according to six sigma principals.

The software demonstrated great potential. Initial fault diagnosis identified a problem with the calculation of the required lime to the process. The real-time reporting system has been running for a period of months and has consistently pointed out a problem with flow measurements. All of these findings were in line with expert knowledge at Worsley, which has confirmed the dependability of such a monitoring system.