A modern complex industrial process, such as petroleum refinery, petrochemical plant, pulp & paper process and power plant, usually include hundreds or thousands of control loops. It is a well-known fact that many controllers in a process plant are not tuned properly (Desborough and Miller, 2001). Badly tuned controllers would lead to loss in production as well as quality (Bialkowski, 1993; Ender, 1993). Therefore, it is necessary to detect the controllers that are poorly tuned and diagnose their behavior as sluggish or aggressive (oscillatory) and make the operators aware of it so that appropriate retuning actions can be initiated. There are methods in the literature for diagnosing aggressive (Thornhill and Horch, 2007) or sluggish (Hugglund, 1995; Kuehl and Horch, 2005) controller behavior. However, each of these methods has its own limitations and none of them addresses the improper controller tuning issues (sluggish and aggressive behavior) in a unified framework. Here, we have proposed a non-invasive method to automatically detect the badly tuned controller and identify its tuning issue as sluggish or aggressive directly from the routine plant operation data. The effectiveness of the proposed method is demonstrated on both simulated as well as industrial control loop examples.
展开▼
