摘要:为提高齿轮故障诊断的准确率,提出了核主元分析和纠错输出编码支持向量机相结合的方法。首先采用基于核主元分析方法对原始样本向量进行预处理,实现对原始样本向量的降噪及冗余消除。然后采用基于纠错输出编码矩阵构造出若干个互不相关的子支持向量机,以提高分类模型的整体容错能力。最后,把经过核主元处理后的新向量作为纠错输出编码支持向量机的训练及测试样本,实现对不同故障状态齿轮的识别。结果表明,该方法能够提取更有效的分类样本向量,故障诊断效果更好。
Abstract:, A method was proposed based on combination of KPCA with ECOC-SVM to enhance accuracy of fault diagnosis of gear. Firstly, original sample vector was preprocessed based on KPCA to eliminate noise and redundancy. Secondly, some uncorrelated SVMs were constructed based on ECOC matrix to improve whole performance of fault tolerant of classification model. Finally, the new vector got through KPCA was used as training and testing sample of ECOC-SVM to recognize different fault state of gear. The result shows this method can extract better sample vector for classification, and has better effect of fault diagnosis.