基于分类器组的轴承故障识别方法研究

Fault recognition method of rolling bearings based on classifier ensembles

提出一种基于设备运行数据构造分类器组用于滚动轴承故障识别的方法。在决策表上使用属性约简的遗传算法找出构成候选基分类器的较好约简,再使用多样性筛选的遗传算法找出最终的约简,以此为基础结合加权投票策略构建分类器组用于模式分类。通过轴承正常情况、内圈、外圈和滚动体故障的识别实验验证了方法的有效性,得到了较好的实验结果。

A method to construct a data-based classifier ensemble used in fault diagnosis of rolling bearings was presented.The candidate reducts which could be used to build the base classifiers were found by applying a genetic algorithm for feature reduction on a decision table,and then the other genetic algorithm for diversity evaluation was used to search the ensemble of base classifiers.Based on the result above and the weighted voting strategy,the final solution to pattern classification could be set up.It was proved by means of the diagnosis experiments including normal condition,inner race faults,outer race faults and rolling elements faults that the method proposed here is valid and the result obtained is better.