Closed Loop Performance Monitoring: Automatic Diagnosis of Valve Stiction by means of a Technique based on Shape Analysis Formalism

Valve stiction is an important cause of performance deterioration in control loops of industrial plants; owing to the large number of loops in complex plants and different cause of poor performance, it is important to be able to detect causes and suggest actions to perform in automatic way. The paper examines some recent techniques to detect the presence of valve stiction as root causes of oscillations, by using qualitative shape analysis formalism. Basic properties and main factors are put into evidence by application on data generated by simulation, while the reliability is checked by application on plant data sets to account for sensitivity to noise, for the effect of set point variations due to cascade or advanced control acting on the upper level. The algorithm shows to be able to detect stiction when it shows up with clear patterns (about 50% of examined cases coming from about 200 data sets). This allows a quicker detection and saving of computation time with respect to more comprehensive techniques, thus suggesting on line implementation of the technique.