Thesis for the Degree of Master of Science


Year 2001
Susanna Nevalainen

Control performance indexes in petrochemical plants

The purpose of this thesis was to develop indices to evaluate the performance of a model predictive controller. The aim is to distinguish between different situations that cause the deviation of controlled variables. In addition to a tuning problem, the deviation can result from a device fault, a poorly tuned lower level controller or restrictions in the process.

In the literature part, the performance indices of SISO-controllers were studied and different ways of evaluating the performance of a model predictive controller were surveyed. Other methods used in performance evaluation were also studied.

In the experimental part, the control performance indices were developed and tested. Applications developed by Neste Engineering were used. The simulated process was a dearomatization unit named LARPO. The model predictive controller was implemented to control the quality of the product. Control performance calculations were implemented in the controller. In particular the applicability of the prediction error variances for controller performance evaluation were studied. A performance index that accounts for restrictions of controlled and manipulated variables was also developed. Indices were also developed for supervision of error signal and control moves.

The performance indices were tested in ordinary process conditions such as stationary state, set point change and disturbances with well and poorly tuned controller. With a well-tuned controller, cases of device fault and poorly tuned lower-level controller were simulated.

Using prediction error variance to evaluate control performance gave promising results. With prediction error variance, device faults were detected. Modeled and un-modeled disturbances were distinguished with the performance indices developed for supervising the error signal. The performance indices applied to control signals were able to detect aggressive control action. Indices developed for supervising process constraints were able to detect situations where the value of a controlled or manipulated variable was at its boundary.

Keywords: control performance, model predictive control, constraints

This info last modified 20 Jun 2018 - contact webmaster