Control Valve Stiction Detection Based on the Improved Linear Regression Method

Accurate detection of control valve stiction is crucial for ensuring system stability and production safety in industrial process control. The stiction detection method based on linear regression fall short in identifying signal peaks and valleys due to the fixed window size, making it difficult to accurately determine stiction. To address this, an improved detection method is proposed, in which the algorithm automatically identifies the peaks and valleys in the signal, and dynamically selects the window size in conjunction with the signal characteristics, thereby enhancing the accuracy of stiction detection. Moreover, a curve shape index is introduced to reinforce the reliability of the detection results. The effectiveness of the improved linear regression method in control valve stiction detection and its advantages over existing methods have been validated with actual industrial data.