论文部分内容阅读
为了降低滚动轴承故障智能分类的训练时间并提高分类精度,提出了一种滚动轴承正常、内、外环故障及不同故障严重程度的多状态分类方法。该方法首先采用峭度值结合相关系数法确定集合经验模态分解结果中包含主要状态信息的固有模态函数;再将其组成特征矩阵,利用奇异值分解所得奇异值作为特征向量;最后在采用改进分类规则的超球多类支持向量机分类时,提出由各状态超球球心间距中的最值来确定多类分类器核参数的选取范围,缩小选取区间,最终实现滚动轴承的多状态分类。实验结果表明,提出的滚动轴承多状态分类方法可以减少分类器的训练时间,提高分类精度。
In order to reduce the training time of intelligent fault classification of rolling bearing and improve the classification accuracy, a multi-state classification method of rolling bearing normal, inner and outer ring faults and different fault severity is proposed. In the method, the natural modal function containing the main state information in the empirical mode decomposition result of the set is determined by using the kurtosis value and the correlation coefficient method. The eigenvector is composed of the characteristic matrix and the singular value obtained by the singular value decomposition as the eigenvector. Finally, In the classification of hypersphere multi-class support vector machines based on improved classification rules, it is proposed to determine the range of multi-classifier kernel parameters by selecting the best value of the hypersphere center-to-center distances in each state, to narrow the selection interval and finally to realize multi-state classification of rolling bearings . The experimental results show that the proposed rolling bearing multi-state classification method can reduce the training time of classifier and improve the classification accuracy.