Abstract :
Manufacturers in process industries face the daily challenge of squeezing the most out of their assets. Whether the squeeze is on cost, throughput, quality or environmental compliance, the harder the push, the greater is the risk of a process deviation. Control engineers have a variety of automation strategies available to meet the challenge; likewise, production and quality managers have a range of statistical tools at hand to monitor and provide early warning of process upsets. While both approaches are complementary and the goal is the same i.e. improved manufacturing performance, there remains a gap between how statisticians and process control engineering approach the task. This presentation aims to go some way to narrowing the gap. This presentation discusses an industrial application that reveals the usual spectra of issues facing processing industries; varying raw material properties, fluctuating operating conditions due to equipment and process degradation and changing production needs. The fluidised bed dryer provides a useful framework to demonstrate the complexities of process modelling and monitoring in the presence of significant process disturbances, interacting controllers, limited measurement accuracy, intermittent equipment failures and last but not least, a tight project budget. The application demonstrates when and how the statistical tools aid the initial analysis of the process, highlight process problems and finally integrate with the model predictive control system to improve overall robustness to deliver an on-line system with a payback period of 4 months
