【摘要】 为了准确地对轴承性能退化过程数据进行评估,将Mexican hat小波函数引入支持向量机多分类器中,提出一种小波支持向量机多分类器。并基于平移不变核函数条件,给出该小波函数为容许核函数的证明。根据"一对多"算法建立支持向量机多分类器。通过对内圈故障和滚动体故障的轴承性能恶化过程中数据的分析,表明小波支持向量机具有比BP(back propagation)神经网络、RBF(radial basis function)核函数支持向量机更高的分类正确率。更多还原

【Abstract】 In order to assess bearing performance degradation accurately,a wavelet support vector machine multi-classifier is proposed,in which the Mexican hat wavelet function is used as the kernel function.Based on the translation invariant kernel condition,the Mexican hat wavelet function is deduced to be an admissible kernel in theory.A support vector machine multi-classifier is constructed based on "one against all" algorithm.Though the assessment analysis of degraded bearing data,it is proved that the wavelet support vector machine has higher correctness rate than back propagation(BP) neural networks and RBF(radial basis function) support vector machine.更多还原