基于特征优选和GA-SVM的滚动轴承智能评估方法

针对滚动轴承等旋转机械设备零部件的退化状态识别问题,研究并提出一种基于支持向量机(support vector machine,SVM)的智能评估方法。对于在线持续输出的轴承振动信号,采用时域方法和集成经验模态分解(ensemble empirical mode decomposition,EEMD)能量熵提取轴承特征,并基于相关性、单调性和鲁棒性进行特征选择。综合考虑三个指标,计算选择准则,得到最终的退化特征。针对SVM参数选择困难问题,使用遗传算法(genetic algorithm,GA)优化SVM参数,确定参数最优值。定义轴承性能退化指标,用三种不同故障类型的轴承数据训练模型。分别输入不同故障的轴承全寿命周期数据,得到轴承故障的类型和性能退化曲线,确定早期故障点,并对方法进行对比和验证。实验表明,模型故障诊断平均准确率为97.69%,性能退化评估曲线结果准确,早期故障检测能力强。

An intelligent method for rolling bearing evaluation using feature optimization and GA-SVM

Condition monitoring and fault diagnosis of rolling bearings is one of the hotspots in the research of rotating machinery fault diagnosis.For the problem of degenerative state identification of rotating machinery such as rolling bearings,an intelligent diagnosis method based onsupport vector machine was proposed.For the bearing vibration signal of continuous output on the line,a time domain method and the ensemble empirical mode decomposition(EEMD)energy entropy were used to extract the bearing characteristics,and the feature selection was based on correlation,monotonicity,and robustness.Considering three indicators comprehensively,the selection criteria were calculated to obtain the final degradation characteristics.For the problem of parameter selection of support vector machine,the genetic algorithm(GA)was used to optimize the SVM parameters and determine the optimal values of the parameters.The bearing performance degradation state health indicators were defined,and the model was trained with bearing data from three different fault types.The bearing life cycle data of different faults were input separately,and the type of bearing fault and the performance degradation curve were obtained to determine the early fault point.Finally,the method was compared and verified.Experiments show that the average accuracy of model fault diagnosis is 97.69%,the performance degradation assessment curve is accurate,and the early fault detection capability is strong.