Diagnosis of Compound Faults in Reciprocating Compressors Based on Modulation Signal Bispectrum of Current Signals

This paper studies induction motor current signatures to detect and di-agnose faults of a two-stage reciprocating compressor (RC) which creates a varying load to the motor. It also examines the influences of stator winding faults on differ-ent common faults of the compressor. Both the conventional spectrum analysis and the state of the art modulation signal bispectrum (MSB) analysis are used to process the current signals for attaining an accurate characterisation of the modulation in-duced by the variable loads and thereby developing reliable diagnostic features. The experimental studies examine different RC faults including valve leakage, inter-cooler leakage, stator asymmetries and their compounds. The results demonstrated that the MSB has a better performance in differentiating spectrum amplitudes caused by different faults especially the compound fault. Thus the MSB based fea-tures are demonstrated to be more reliable and accurate as the analysis techniques for motor current based diagnostics.