The Friction Compensation Analysis And Diagnosis Of Pneumatic Control Valve

During the process of modern industry, automation level rises quickly, and has achieved remarkable results. Increased automation is reflected in the increase in the number of controllers and control loops. A plant may have several or even hundreds of control loops, therefore, the diagnosis and monitoring of control loop performance is an important guarantee for the production of safe and efficient.The pros and cons of the control loop performance are essential to guarantee high yield and low consumption of the product. The normal lose of components or changes in operating conditions are likely to result in the controller performance. The pneumatic control valve is a typical actuator in the most process control loops. It is the only moving part in the control loop, which control the implementation of the decision-making. Even if the controller design is reasonable, the control valve failure is bound to result in performance degradation of the control loop.The shock of the controlled variable caused by the control valve is the main reason for the performance of the control loop. Control loop cover the authenticity of the control loop performance, the existence of the fault valve led to the decline in product quality, defective rate, the increase in energy consumption, the reduce of production capacity, less profitable. Fault detection on pneumatic control valve is necessary.In this paper, poor control loop performance caused by control valve nonlinearity which is produced by stiction are studied. And a method based signal process about the fault diagnosis of spring-diaphragm type valve is suggested. It includes the detection, quantification and compensation of valve’s stiction.Firstly, we diagnose the control loop’s performance. Using Non-Gaussianity Index (NGI) and Non-Linearity Index (NLI), we can detect signal’s Non-Gaussianity and Non-Linearity, and diagnose the causes of poor control loop performance.Secondly, introduce two parameters data-driven model of valve stiction. The model is simple, yet powerful enough to properly simulate the complex valve stiction phenomena. Both open and closed loop results have been presented and validated to show the capability of the model. Finally, we use the ellipse-fitting method and PV-OP plot to judge the existence of stiction. Then we use the separable least-squares and global search algorithms to estimate the valve’s stiction.at the same time, a method of compensate stiction is introduced. The objective is to compensate the valve’s stiction, and improve the performance of control loop.The whole process on friction diagnosis, including the detection, quantification and on-line compensation of friction were studied, the single parameter identification method of Hammerstein model is deduced to the dual parameter identification method which is suitable for the friction model. Some cases’ researches have demonstrated the wide applicability and practicality of this method as a useful diagnostic method in troubleshooting poor control performance.