A Novel Scheme For Fault Detection Of Reciprocating Compressor Valves Based On Basis Pursuit, Wave Matching And Support Vector Machine

Reciprocating compressors are one of the most widely used machines in petroleum and petrochemical industry. They generally play a key role on the production line and their working status is directly related to the efficiency of the factory. The structure of compressors is complex and many of their parts work in a high-temperature and high-pressure condition and under reciprocating forces. As a result, failures often occur and even lead to severe consequences. The valve is the most frangible part of reciprocating compressors and it fails more frequently than the other parts. Therefore, the effective and accurate fault diagnosis for compressor valves is highly necessary and crucial in terms of accident prevention, maintenance decision-making and cost minimization.The vibration signal of the compressors contains plenty of periodic and transient components and it shows clear non-stationary features. The parts of compressors have the same motion period, so their characteristic frequencies mix with each other and are not able to be identified on the frequency spectrum. In this paper, the mathematical model of the vibration signal of valves is built based on the analysis of the working principle of the valve. After that, a novel scheme for fault detection of compressor valves based on basis pursuit, wave matching and support vector machine is proposed from the perspective of time-domain. Basis pursuit is applied to extract the main vibration component in the signal and to suppress background noise. Wave matching is a new feature extraction method proposed in this paper. Instead of extracting features through commonly used indicators such as statistic measures or information entropy, wave matching extracts features by matching the vibration signal with parameterized waveform optimized by differential evolution algorithm. It only produces a small number of features and the features have clear physical meaning. Support vector machine is employed in the fault classification because of its superiority in dealing with small sample problems. And then, the above scheme is tested by vibration signals of the compressor valve from both experiment and factories. The results of experiment signals confirm that the proposed scheme can differentiate three different compressor valve conditions (normal condition, spring deterioration and valve plate deformation) with high accuracy and reliability. The results of on-spot signals from factories show that the proposed scheme can effectively identify the health status of the valve. Finally, an analysis on the advantages and the limits of the scheme is given and the further research prospects based on the scheme is discussed.