Fault Feature Extraction And Identification Of Reciprocating Compressor Based On Multifractal

Reciprocating compressor is used in all kinds of industry field widely, its condition monitoring and fault diagnosis is one of focus of the fault diagnosis studies. Aiming at the nonlinear characteristic of reciprocating compressor vibration signals, multifractal was applied to extract fault feature parameter of reciprocating compressor, after that the parameter was identified by Support Vector Machine, and the new method was provided for the fault diagnosis technology of reciprocating compressor.Multifractal theory is spread of self-similar general fractals, it surmounts disadvantage that single fractal dimension is difficult to characterize signals fully, and could be used to describe and extract intrinsic character of non-stationary signals exactly. This paper compared different arithmetic of multifractal, and the improved multifractal arithmetic was used to compute fault feature parameter of reciprocating compressor. The computing results showed that general dimension Dq give a good presentation for reciprocating compressor working condition, and Support Vector Machine identifying samples was constructed based on general dimension Dq as fault feature parameter. This paper also introduced the choice of kernel function and optimization of parameter about Support Vector Machine, and a Support Vector Machine multi-classifier was constructed based on them.This paper shows the concrete steps of fault feature extraction and identification of reciprocating compressor clearly, and compressor gas valves and axletree fault were identified and analyzed based on actual measuring signals. The identification result shows that this method surmounts localization of traditional method, and could identify kinds of compressor fault exactly. In the end, the method of reciprocating compressor classification diagnosis was brought forward in this paper, the measure have been to compare the difference ofΔD qwhich have been acquired by computing fault and normal condition signals of compressor two-stage system, and the bigger one was remained with fault stage; the validity of this method was proved by analyzing kinds of compressor fault condition.