Intelligent Fault Diagnosis Methods And Experimental Verification Of Pneumatic Membrane Control Valve

Pneumatic control valves have been widely used in many industrial fields and become one of the actuators of industrial process control systems.Most of the pneumatic control valves work in high temperature,high pressure and humid environments,which usually results in faults.Faults can lead to a decline in factory efficiency or even accidental downtime and casualties.The traditional manual maintenance method usually consists of disassembling and fixing the valve,which may waste a lot of manpower,material resources and time.Therefore,the fault diagnosis method that can be carried out in the non-disassembly case has attracted much attention.Currently,the existing diagnosis methods can be classified into two classes: the expert system-based methods and the machine learning-based methods.The disadvantage of the former is that the selection of parameters is subjective and the adjustment of parameters is difficult,which usually lead to diagnosis missing or misreport.The latter is usually faced with the challenge of less samples available on site and uneven distribution.Therefore,this thesis studies the intelligent fault diagnosis and experimental verification methods of pneumatic membrane control valves.The main results are as follows:1.The common faults of pneumatic diaphragm control valve are analyzed,and the experimental platform is designed.Firstly,the common faults of pneumatic diaphragm control valves are analyzed and the design methods of manual manufacturing faults and control fault strength are given.Then,the software and hardware system requirements of the experimental system are analyzed and the overall design of the experimental platform and the selection of each module are proposed.Finally,the experimental platform is verified by using high-precision instruments.2.Aiming at the problem that the parameter selection of traditional expert system is subjective,the parameter adjustment is difficult,and the false alarm is easy to report,a fault diagnosis method based on improved expert system for pneumatic diaphragm control valve is proposed.Firstly,the expert experience of fault diagnosis for pneumatic membrane control valves is analyzed,and the rationality of expert experience is demonstrated.Then,particle swarm optimization(PSO)algorithm is used to improve the expert system,and a fault diagnosis method of pneumatic membrane control valve based on the improved expert system is proposed.Finally,the experimental platform is used to collect fault samples,and the diagnostic accuracy of the proposed method and the traditional expert system method is compared to verify the effectiveness of the proposed method.3.For the machine learning method,there are few samples available on site,the distribution is uneven,and there are often many samples of individual models,but less of other models of the same type,a pneumatic membrane control valve fault diagnosis method based on semi-supervised migration ELM and auto-encoder(AE)is proposed.Firstly,the basic principles and characteristics of machine learning algorithms such as ELM,migration learning and auto-encoder are reviewed.Secondly,according to the difference between air pressure and displacement under fault conditions and trouble-free conditions,the appropriate feature functions are selected.The abnormality of the characteristic function values in the fault condition reflects the fault,and the characteristic function values are used to form the experimental samples.Then,the fault diagnosis method of the pneumatic membrane control valve based on semi-supervised migration ELM and auto-encoder is proposed.Finally,the fault samples are collected to verify the effectiveness of the proposed method.