| Modern chemical processes are highly complex and operate with a large number of variables under closed loop control. Plant engineers and operators are unable to effectively detect and diagnose faults for these large scale systems using traditional process monitoring methods. However, by developing measures that more accurately characterize the state of the process, the plant engineers and operators can more effectively be incorporated into the process monitoring loop. The large amount of data available from modern processes and the computational power of today's computers enable the employment of empirical-based monitoring measures to be practical and effective. New empirical-based measures for fault detection, identification, and diagnosis are developed, evaluated, and compared with existing measures. The measures are based on methods obtained from the chemometric, pattern classification, system identification, and artificial intelligence literature. It was found by applying these measures to a simulation of a realistic chemical process that some of the new measures are better for monitoring the process compared to the existing measures. |
