【摘要】 提出了基于时间序列参数模型和核Fisher判别分析(KFDA)的滚动轴承故障诊断方法.该方法首先通过自相关算法对轴承振动信号建立自回归(AR)模型,将自回归模型的参数作为特征向量并映射到高维核空间.然后在高维核空间中进行Fisher判别分析,求出Fisher判别分析的最优投影向量以及各类状态的Fisher判别值.最后获取未知状态轴承的高维核空间特征向量,求出其在最优投影向量上的投影值,通过与判别值进行距离判别来识别轴承所处的状态.实验结果验证了所用方法的有效性.更多还原

【Abstract】 A fault diagnosis approach to rolling bearing based on parameter model of time-series and kernel Fisher discriminant analysis(KFDA) is proposed.Firstly,auto-regressive(AR) model of the vibration signal of rolling bearing is established,and the auto-regressive parameters are regarded as the feature vectors and mapped into a high dimension kernel space.Then Fisher discriminant analysis is performed in the high dimensional space,the optimal projection vector and Fisher discriminant value of every state are obtained.Finally,the unknown state bearing′s feature vectors in the high dimension kernel space and projection value on the optimal projection vector are obtained,then the projection are compared with the discriminant value to identify the bearing′s state by distance discrimination.The experiment result shows the effectiveness of the proposed approach.更多还原