中文摘要：

支持向量机（SVM）的分类性能受样本的特征以及SVM本身参数的选择影响较大。针对这种情况，基于Shannon能量熵、SVM和小生境遗传算法（NGA），提出了一种基于NGA优化SVM的滚动轴承故障诊断方法。该方法采用容错性强的Shan—non能量熵作为特征参数，对信号进行EMD分解提取出前3个IMF分量作为特征信号，分别计算其Shannon能量熵作为特征向量得到样本集，作为多类别SVM的输入。在用样本训练SVM时，构造一种新的核函数，并采用NGA对SVM的核函数参数进行全局优化，使SVM获得最佳的分类性能，提高其分类识别的正确率。最后采用凯斯西储大学的滚动轴承故障样本进行了分类识别，并与其他几种方法进行了对比，结果表明该方法具有更好的可靠性和分类准确率。

英文摘要：

Aiming at the fact that the classification performance of support vector machine （SVM） highly depends on the sample feature and SVM parameter selection, a rolling bearing fault diagnosis method is proposed based on niche geuatic algorithm （NGA） optimized SVM. The method adopts strong fault-tolerant Shannon entropy as the character- istic parameter,the first three IMFs of the signal are extracted with SVD-morphology filter-EMD （ empirical mode de- composition） and used as the characteristic signals;their Shannon entropies are calculated and used as the character- istic vectors;and the sample set is obtained,which is then served as the input of the multi-class SVM classifier to achieve better de-noising performance. Then,in the SVM training process, a new kernel function is constructed, and the kernel function parameters are optimized with NGA to achieve better classification capability and improve the classification correct rate of the SVM. The rolling bearing fault sample data from Case Western Reserve University were used to carry out classification experiment;the experiment results were compared with those of several other methods. The results prove that the proposed method has better reliability and classification correct rate.