Abstract: The vibration signals of the air valve of the reciprocating compressor is interfered by strong gas fluctuations. Aiming at such characteristics, this paper proposes a fault diagnosis method based on improved maximum correlation kurtosis deconvolution (MCKD) and refined composite multi-scale fuzzy entropy (RCMFE). Deconvolution of vibration signals of reciprocating compressor air valve using improved maximum correlation kurtosis deconvolution can extract the shock components of the signals effectively; after the deconvolution, the signals are analyzed by adopting the refined composite multiscale fuzzy entropy quantization to obtain the feature vectors of the fault diagnosis of reciprocating compressor air valve and the feature vectors are input in the support vector machine for fault feature identification. The analysis of experimental data of reciprocating compressor air valve faults shows that this method can extract the fault information of reciprocating compressor air valve effectively and realize the accurate diagnosis of air valve faults of reciprocating compressor.

Key words: maximum correlation kurtosis deconvolution, reciprocating compressor, refined composite multi-scale fuzzy entropy, air valve, fault diagnosis