Faults Detection And Stiction Compensation For Pneumatic Control Valve

In the process control loop, pneumatic valve is a most common used actuator,and whether it can work in good condition or not will definitely affects the control performance of the loop. However, with the growth of the running time of the control loop, the valve will slowly rise to various failures. As to the detection and classification of common faults in spring diaphragm pneumatic valves, this paper proposed an approach to detect the valve faults according to the accumulated residual method with properly threshold setting. After a fault was detected, the particle swarm optimized BP neural network algorithm was used to complete the fault classification. Valve fault conditions and fault types were investigated on the simulation platform DAMADICS with the use of the proposed method, and is verified by the actual valve fault data from a factory. Simulation analysis and real-plant data verification results shows that the detection algorithm can effectively detect the existence of valve fault and the optimized classification algorithm can achieve high classification accuracy rate than that of the traditional BP neural network classification algorithm, and the accuracy rate has been improved. As to the stiction fault, this paper proposes the use of TS type fuzzy controller instead of the traditional PI controller to eliminate such oscillations. When the valve is sticky, using the relations between the status information of the controlled object and the change rate of the controller output to build the fuzzy control rules. Adjust the integrator coefficient of traditional PI controller so that the valve can quickly get out of the sticky area and ultimately compensate for stiction. The actual level control results have shown that the compensation effect of the controller is obvious and the controller is robust enough to adapt to different set values.