Research On Condition Monitoring And Fault Diagnosis System For Reciprocating Compressor

Reciprocating compressor is widely used in the fields of petrochemical industry,mining metallurgy,air conditioning and refrigeration.It is the key equipment in industrial production process and its operating condition determines the production efficiency of enterprises directly.However,the reciprocating compressor has a large number of parts and usually works continuously for a long time under harsh conditions,which is easy to cause various faults.Therefore,it is of great practical significance to carry out research on condition monitoring and fault diagnosis of reciprocating compressor.In this thesis,the reciprocating compressor was taken as the research object,and its vibration signal was used as the information source.The methods of fault feature extraction and fault type identification were proposed,and a set of condition monitoring and fault diagnosis system was designed and developed.The specific research contents can be divided into the following parts:(1)The research status at home and abroad of fault diagnosis of reciprocating compressor were summarized,and the common methods used for fault diagnosis were expressed.Based on the analysis of the composition and working principle of the reciprocating compressor system,the scheme of using vibration signal for fault diagnosis was determined,and four vibration excitation sources of the reciprocating compressor were analyzed in detail,then the common fault types were elaborated,which laid a foundation for the subsequent diagnosis algorithm.(2)Feature extraction is the first and crucial step in fault diagnosis.Because the vibration signal is non-linear and non-stationary,it is difficult to extract its features effectively by general time-frequency domain analysis methods.Therefore,a fault feature extraction method based on HHT was proposed in this thesis.First,the original vibration signal was filtered out by singular value decomposition to remove noise,and then the denoised vibration signal was decomposed into several IMF components by EMD.Then,the energy characteristics and marginal spectrum characteristics of the IMF were obtained by HHT respectively,and these two characteristics were combined into a complete highdimensional feature vector.Finally,the compressed low-dimensional feature vector was obtained by PCA,which was used as the input of the neural network.(3)Fault type identification is the core step of fault diagnosis.In this thesis,the global optimization capability of PSO was used to help RBF neural network to build a better performance fault identification model.In order to solve the problem that PSO is easy to fall into local optimum during iteration,a mirrored “S” shape inertia weight decreasing strategy was put forward to enhances the optimization ability of PSO.The validity and accuracy of this method were proved by comparing the two other network models(BP neural network,traditional BRF neural network).(4)A condition monitoring and fault diagnosis system for reciprocating compressor was designed and developed based on the above research results of fault diagnosis methods.The main functional modules include: system login module,thermodynamic condition monitoring module,vibration signal monitoring module,historical data query module and fault diagnosis module.(5)Finally,the work results of the thesis were summarized,and the deficiencies and improving direction of this thesis were pointed out.