Valve Stiction Detection and Quantification Using a K-Means Clustering Based Moving Window Approach

In this paper, a novel and effective stiction detection method is proposed by combining K-means clustering and the moving window approach. As a byproduct, the proposed stiction detection method offers an estimation for the stiction band in sticky control valves. The proposed stiction detection method is tested in industrial case studies consisting of benchmark industrial control loops and control loops from an oil sands industry. In the benchmark industrial control loops, the results of the proposed method are compared with some of the existing stiction detection methods. This comparison shows superior performance of the proposed method. It is noticed through a simulation case study and an industrial case study that the proposed method not only provides stiction band estimation but also can detect severe valve stiction or unexpected valve closures.