Performance evaluation of decomposition methods to diagnose leakage in a reciprocating compressor under limited speed variation

Abstract

Reciprocating compressors (RCs) are used in pressure based applications to achieve high-pressure ratio. An efficient detection of a leakage in a RC by vibration based signal processing techniques can avoid high energy losses. Conventional signal processing methods used to examine the faults in non-stationary vibration signals have been proved to be inefficient. This paper aims to detect valve leakage in a RC by evaluating the performance of signal decomposition techniques i.e. empirical mode decomposition (EMD) and recently developed variational mode decomposition (VMD). These non-stationary signal processing techniques have been widely applied for condition monitoring of mechanical systems. In the present study, the vibration signals of RC at different levels of leaks under limited speed variation were acquired, decomposed and compared. To detect the fault features exhibiting leakage, the vibration signals were decomposed by EMD and VMD respectively. Post decomposition, FFT analysis along with RMS and kurtosis were evaluated from decomposed signals. The characteristics frequencies were clearly exhibited by VMD and the responses of kurtosis were also better with VMD. The fault detection performance of VMD was found better than EMD. © 2018 Elsevier Ltd