Quantification of Valve Stiction using Particle Swarm Optimisation with Linear Decrease Inertia Weight

Abstract

Valve stiction is one of the most common problems on industrial process control loops. The detection and quantification of valve stiction in control loops is therefore important to ensure the high quality of the products and maintain the reliable performance of control loops. This paper presents an algorithm for quantifying valve stiction in control loop based on linear decrease inertia weight particle swarm optimization to obtain more accurate estimates of stiction parameters. The amount of stiction present in the valve is estimated by identifying parameters of Kano model which is a two-parameter data-driven stiction modelling based on the parallelogram of MVPV phase plot. Simulation results have demonstrated the efficacy of this algorithm in valve stiction quantification and also its robustness to oscillations due to inappropriate controller tuning and external disturbances. Results are confirmed by application to real process industrial data.

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

“…The authors (Kamaruddin et al) also developed a stiction quantification method by using butterfly shapes as input to a deep convolutional neural network. Aksornsri and Wongsa employed a particle swarm optimization technique to estimate the parameters of Kano’s stiction model being used in a framework similar to the Hammerstein system. An extensive review of stiction models, stiction detection, and quantification methods was provided in Bacci di Capaci and Scali .…”