Engine fault classification based on characteristic statistics of vibration signals in angle domain

Abstract

In order to improve the precision rate and success rate of engine fault classification, a new fault classification based on the characteristic statistics in angle domain was proposed. The new method used an optical encoder to acquire engine rotational vibration signals of equal angle sampling. Wavelet packet analysis and correlation coefficient method were applied to acquire the characteristic order of the engine signals in angle domain. Energy ratio, standard deviation ratio, spectrum energy ratio and spectrum mean ratio of the characteristic order were regarded as the characteristic statistics in angle domain to extract engine fault features. The method of support vector machines was also adopted to classify the engine faults. Experimental results using signals recorded from an engine with an improper fit clearance existed in the connecting rod bearing show, compared with the traditional fault classification, this new method can improve precision rate of the fault classification.