Fault Diagnosis Method For Control Valves Based On The Cloud Model Theory

Due to the advantages of stable performance and agile response,the control valve has been widely used in various fields of industry.However,frequent mechanical actions and harsh working conditions make it easy to appear various faults,which seriously affects the safe and stable operation of the system.The traditional fault diagnosis method is time-consuming and laborious,and has been difficult to meet the needs of modern industrial systems.With the requirements of ’Made in China 2025’,’The 14 th Five-Year Plan’ and other national development strategies for promoting industrial intelligence,various intelligent diagnostic methods have been continuously proposed and improved.However,the current intelligent diagnosis methods for control valve faults still face certain challenges in terms of fault feature extraction strategy,fault data collection,fault concurrency and so on.In view of the above problems,this thesis studies the data-driven intelligent fault diagnosis method of control valve based on cloud model theory and stochastic configuration network algorithm,and the main work is summarized as follows:1.A fault feature fusion strategy based on CM-DiPCA is proposed.The different opening degree of the control valve is taken as the entry point,on the one hand,the cloud model theory is used to calculate the cloud digital features of faults in order to mine the uncertainty information.On the other hand,combined with the open-closed dynamic working characteristics of the control valve during operation,the dynamic inner principal component analysis is used to extract the dynamic features which will be fused with cloud digtal features.The CM-DiPCA method proposed by this thesis solves the problem that the fault sampling signal contains a large amount of uncertain information while the conventional feature extraction method has difficulty to excavate effectively.The effectiveness and feasibility are proved by comparative experiments.2.An ensemble stochastic configuration network algorithm on the basis of cloud based sampling is proposed.Firstly,the small scale fault data set is expanded by cloud based sampling.Secondly,several diagnostic basis models are trained in parallel by using the constructed samples under different randomness constraints.Finally,the voting decision method is used to diagnose the test samples,which improves the problem that the data-driven intelligent diagnosis method has poor diagnostic accuracy under the background of limited training samples.And the superiority of the proposed method is verified on the public data set.3.A simultaneous fault diagnostic framework based on Dual-SCN-OD is proposed.To begin with,on the basis of stochastic configuration network algorithm,two classification models named ’ counting network ’ and ’ label network ’ are trained in parallel to determine the number of faults and fault labels respectively.In the second place,the overlap degree between the sample to be diagnosed and each type of single fault selected randomly is calculated.Then the results and the output matrix of the label network are weighted and integrated into the final output matrix.Furthermore,combined with the diagnosis results of the counting network,the diagnosis of simultaneous faults is realized.Finally,comparative experiments show that the proposed method improves the diagnosis accuracy of concurrent faults.