基于小波包和进化支持向量机的齿轮早期诊断研究

GEAR INCIPIENT DIAGNOSING BASED ON WAVELET PACKET AND GENETIC-SUPPORT VECTOR MACHINE

针对齿轮早期故障的特征不明显,提出了一种基于小波包和进化支持向量机的齿轮故障诊断方法,该方法既充分利用了小波包优良的时频局部化特性,又利用了支持向量机在小样本情况下出色的学习性能和良好的推广特性,以及遗传算法的全局优化能力。在齿轮试验台上的应用结果表明,经过特征提取和参数优化后,提高了支持向量机的分类能力。

Due to incipient fault attributes of gear being not obvious,a hybrid diagnosis model to gear diagnosing based on wavelet packet and genetic-support vector machine is proposed.At first,the model makes full use of the time-frequency localization ability of wavelet packet;secondly,it utilizes support vector machines(SVM)which can make the model have good learning and developing ability in situation of small sample;finally,it uses global optimization of genetic algorithm.The proposed model is applied to a gear testing system,and the results show the fault classification performance of the SVM is improved after feature extraction and parameter optimization.