(359d) Stiction Compensation in Process Control Loops

Research in developing automated controller performance monitoring systems has been increasing in the past decade. One of the main objectives of such automated systems is to make use of the closed-loop controller data that is available in the plant historian database to make assessments about the health of the control loops. The performance assessment system is just one part of a process needed to optimize the controller performance. Assessment has to be typically followed by diagnostics for the badly performing loops and this is, in general, a difficult task. Deterioration of control performance may be due to several reasons such as badly tuned controllers, oscillating external disturbances or non-linearity in actuators.

Of the various reasons for poor performance, stiction in control valves is implicated in more than 30% of the cases as reported by recent industrial surveys. Hence routine identification of stiction from operational data has assumed significance over the last few years. While stiction identification by itself could have sound financial implications, the sticky loops continue to operate poorly till the subsequent production stop. Production stops are in general scheduled a priori and are quite apart in frequency. Performance of these loops can be improved prior to maintenance if it were possible to compensate and mitigate stiction within a closed loop framework. While the financial gains by stiction compensation can be substantial, the general industrial perspective seems to be that such compensation will result in drastic valve movements, which might negate the original financial incentive.

In this talk, we will present three different stiction compensation approaches. In the first approach, a well known knocker technique is optimized through the quantification of stiction. In the second approach, an explicit optimization problem is formulated and solved for calculating the stiction compensation moves. This approach is useful to find out the trade-offs between stiction compensation and drastic valve movements. Local minima and computational issues associated with this approach will be highlighted. The third approach uses an innovative methodology and is guarantied to use very gentle valve movements to compensate stiction. This is achieved by a sound understanding of the stiction mechanism. The validity of all these approaches will be demonstrated on simulation studies and also on an experimental liquid level system.