Fault Mechanism Analysis And Intelligent Diagnosis Of Pneumatic Control Valves

Pneumatic control valves are widely used in petrochemical industry,power industry,metallurgical industry and other industries due to its simple structure,ease of use,and excellent explosion-proof performance.However,various faults of pneumatic control valves are common because of mechanical wear and corrosion and the disgusting working environment,which affects the industrial production.In the early time,periodic inspection was usually used to overhaul the pneumatic control valves.But there is a serious problem of 'over-maintenance' in this method,which usually causes a large amount of equipment damaged.The data-driven fault diagnosis methods have been widely favored by scholars in recent years.However,the issues of the robustness of the fault diagnosis,the difference in data distribution and the simultaneous faults are not considered in the existing data-driven fault diagnosis methods of pneumatic control valves.In view of the above-mentioned issues,the fault mechanism analysis and the intelligent diagnosis methods of pneumatic control valves are studied in this thesis.The results of the thesis are summarized as follows:1.The mechanism of common faults of pneumatic control valves is analyzed.Firstly,the mechanism of common faults is analyzed according to the operating principle and status of pneumatic control valves.Then,a mechanism model of pneumatic control valve is established to analyze the characteristic and phenomenon of the single fault and multiple simultaneous faults.Finally,the fault mechanism is verified by the mechanism model,which lays the foundation for the fault data acquisition and the study of fault diagnosis methods.2.A robust fault diagnosis approach for the single fault of pneumatic control valves based on variational mode decomposition-multiple multiscale entropy(VMD-MMSE)and robust random vector functional link network(RRVFLN)is proposed in this thesis.Firstly,the feature extraction method based on VMD-MMSE is used to extract the fault characteristics of the pneumatic control valves.Then,RRVFLN is constructed by minimizing the mean and variance of modeling errors based on the conventional random vector functional link network(RVFLN),which is used for the classification of pneumatic control valve faults to improve the robustness.Finally,the fault data is collected by the experimental system,and the performance of the proposed method is verified by comparing with the existing methods.3.A domain adaptation fault diagnosis approach for the simultaneous faults of pneumatic control valves based on the paralleled domain adaptation random vector functional link network(Paralleled-DA-RVFLN)is proposed in this Thesis.Firstly,the maximum mean difference(MMD)is introduced to construct a novel domain adaptation random vector functional link network(DA-RVFLN).Then,a simultaneous faults diagnosis frame based on Paralleled-DA-RVFLN is proposed,which is used for the fault diagnosis of pneumatic control valves to solve the issues of simultaneous faults and domain adaptation.Finally,the performance of the proposed method is verified by the experimental system.The thesis has 33 figures,13 tables and 80 references.