Valve fault diagnosis of a reciprocating compressor based on hybrid method

Due to strong nonlinear and non-stationary characteristics of a reciprocating compressor's valve vibration signals, the cubic spline interpolation EEMD (S-EEMD) method well utilized still has shortages of mode mixing and envelop inaccurate. Aiming at the above mentioned problems, the combined analysis method of EEMD based on the quartic Hermite interpolation (QH-EEMD) and the power spectral entropy (PSE) was proposed. The original signals were decomposed into a set of IMF components using the quartic Hermite method with advantages of shape-preserving and adjustability and the EEMD method promoting signals' continuity in different decomposing scale to improve the approximation accuracy of the interpolation curve and to decrease mode mixing. The advantages of the QH-EEMD with PSE (QH-EEMD with PSE) analysis method were verified comparing with those of the S-EEMD-PSE (S-EEMD with PSE) method and the QH-EEMD-SE (QH-EEMD with sample entropy) method. Taking common faults of a reciprocating compressor as the study objects, feature vectors of faults were extracted based on the QH-EEMD-PSE method and the faults were diagnosed accurately.