Research And Application Of Approximate Entropy To Fault Diagnosis Of Reciprocating Compressor

The reciprocating compressor is the key equipment of petrochemical industry, and its state control and fault diagnosis have being paid more and more attention, but the traditional diagnostic methods are unsatisfactory because of the strong complexity and nonlinearity of signals, while the approximate entropy measures signals from the view of the complexity and nonlinearity. Base on the former achievements and the real project, this paper introduces the approximate entropy theory, and verifies that this method is effective and feasible by experiments, which introduces a new way for the reciprocating compressor fault diagnosis.Firstly, the basic structure and working principle of reciprocating compressor are introduced, and the reciprocating compressor in the actual project is analyzed. On this basis, common fault diagnosis methods are compared and summarized, and the application of approximate entropy to the fault diagnosis is introduced. The physical meaning and characterization of complex signal capacity are analyzed by signal simulation experiments.According to the approximate entropy principle and definitions, the practical program is designed with LabVIEW. Based on the program, the value of the approximate entropy of several real vibration signals is calculated and analyzed. The analysis results show that approximate entropy has good capability in showing the signal's complexity. But this method has the disadvantages of algorithm complexity and tardiness, so the improved approximate entropy method is introduced and the its practical program is designed. Through the analysis of the vibration signals, this paper verifies that the improved method not only keeps the advantages of the original algorithm but also speeds up the calculation. It is an effective method for state monitoring and fault diagnosis for reciprocating compressors.The fault diagnosis system for reciprocating compressor is designed with LabVIEW and can get and analyze vibration signal data from the database. The analysis results include the vibration signal values in time and frequency domain, the approximate entropy values, as well as the relevant information of the equipment. The working conditions of equipments can be mastered visually and comprehensively through browsing these results. This system provides a good platform for equipments state control and fault diagnosis.