Diagnos av stiktion i reglerventiler för prediktivt underhåll

Sammanfattning: This master thesis reviews advances in detection and diagnosis of static friction (stiction) in control valves. Control valves suffering from stiction results in oscillating control loops which in turn decreases both the process stability and reliability. Perstorp AB sees a necessity to automatically detect stiction in valves before the effects of the malfunctioning valves have become too severe. This thesis presents an introduction to maintenance, control loops and how stiction in valves can affect performance of control loops. A selection of methods presented in the literature are introduced and examined using normal operating data from the site in Perstorp. Common for all selected methods is their intuitive ideas and that they are based on normal operating data and minimum amount of process knowledge. For the adaptive segmentation method a further development is presented where among other things a test quantity is proposed. This thesis shows that none of the selected methods can reliably diagnose stiction in control valves from normal operating data. Noise in process value and irregular disturbances results in weakened trend characteristics. Finally, the thesis has tested a root cause analysis of plant-wide oscillating control loops found on two plants in Perstorp AB.