Multifractal Signal Feature Extraction And Its Application In Fault Diagnosis Of Reciprocating Compressor

Reciprocating compressors are widely used in the transportation of petroleum, chemical and other industrial production, and play an important role in industry equipment. Currently,the reciprocating compressor condition monitoring and fault diagnosis technology has become a hot topic in the field of troubleshooting problems at home and abroad, and feature extraction is one of the difficulties in fault diagnosis technology. For nonlinear and non-stationary characteristics of reciprocating compressor vibration signal, the paper proposes the use of multiple fractal method to quantify the description of its local singularity characteristics, and uses a new fault identification method for identifying their fault classification. The results show that this method can effectively achieve fault diagnosis of reciprocating compressors and provides a new way for fault diagnosis of reciprocating compressors.First of all, through consulting massive literature material, gives a brief overview of the reciprocating compressor fault diagnosis technology research status, thus put forward the research idea of this paper. Based on the basic structure, the reciprocating compressor working principle are briefly summarized, and focus on common failure mode and failure mechanism of the main components in-depth analysis.Secondly, for the nonlinear characteristics of reciprocating compressor vibration signal,proposes the use of multifractal methods of the fault feature extraction. It chiefly introduces the fractal, multifractal theory and basic algorithm, and based on it, proposes multifractal detrended fluctuation analysis(MF-DFA) theory and analyze the vibration signal of the reciprocating compressor valve. It can verify the validity of the MF-DFA method by analyzing the vibration signal feature extraction of the reciprocating compressor.Then, for reciprocating compressors, using multifractal detrended fluctuation analysis method to extract the fault feature vector parameters, when chosen the common fault recognition method, the fault recognition rate can be low, it is difficult to distinguish the mode confusion phenomenon, therefore, cites a new fault recognition method-- incremental learning KNNModel algorithm(IKNNModel). Compared with other fault identification methods, it is proved that the method has better performance in reciprocating compressor fault classification.Finally, a reciprocating compressor fault diagnosis method based on MF-DFA and IKNNModel algorithm is proposed. It elaborated several aspects such as the vibration signal data preprocessing, fault feature extraction methods,fault feature parameter optimization andfault feature identification of reciprocating compressor extracted from the reciprocating compressor failure characteristics and recognition, and eventually made a complete reciprocating compressor fault diagnosis program. To make an analysis for 2D12 reciprocating compressor troubleshooting experimental test signal data. The results show that the method can accurately extract fault feature and identify fault type, and provides new avenues of research for fault diagnosis of reciprocating compressors.