Control loop performance monitor
Abstract
Scope and Method of Study: Control Loop Monitoring Using Markov Chains and Binomial Statistics Findings and Conclusions: Good control practice is essential for industrial safety and profitability. Real time monitoring of industrial control processes continues to receive both industrial and academic attention due to the impact of control related faults on corporate bottom line. Controllers are tuned for optimum performance but once tuned, process conditions change with time and what was once a good controller becomes a bad one unable to control the process effectively. It will be nice to automatically monitor controller performance so that operators can take immediate action without the tedium control loop monitoring. A number of control loop monitoring products have been developed for controller assessment but most of these programs operate offline, are cumbersome to understand or are themselves fraught with inherent shortcomings making their usage inefficient. This work proposes a simple, efficient, and practicable method to automatically flag poor controller performance in real time. It uses only the run length of the actuating errors. Run length is defined as a state in a Markov chain, and transitions between states (which are binomially distributed) are then modeled using binomial statistics. Transition probabilities from operation are then compared with the control limits (estimated using binomial statistics) established from a user-defined period of good control. Initial test results indicate that the program is effective and adaptable to numerous control configurations.
Collections
- OSU Dissertations [11222]