气动调节阀粘滞故障检测与参数辨识方法研究OA北大核心CSTPCD

To discuss the stiction characteristics of pneumatic control valves,a method of valve stiction detection and parameter identification was proposed with CHEN's stiction model as the basic stiction model.A stiction experimental platform was built to simulate the movement of valve stem with different degrees of stiction to reveal the generation mechanism of valve stiction.The random forest algorithm was used to classify oscillation sources and detect stiction.The particle swarm optimization algorithm based on Hammerstein model was used to get the optimal solution of stiction parameters.Besides,a method for determining the range of values of double stiction parameters was proposed.The proposed method has achieved the stiction parameter identification in three conditions,i.e.under-compensation,no compensation and over-compensation.The results indicate that the degree of valve stiction is related to dynamic and static friction force on the valve stem.If the effects of external factors are not considered,the proposed method of stiction detection achieves a classification accuracy of 99.026 8%.The improved method of stiction parameter identification achieves an error of less than 7%.The research results provide theoretical methods for valve stiction detection and parameter identification,and provide practical reference value for the improvement of stiction models.