多尺度变异粒子群优化MK-LSSVM的轴承寿命预测

Rolling element bearing life prediction based on multi-scale mutation particle swarm optimized multi-kernel least square support vector machine

提出一种基于多尺度变异粒子群优化(MSPSO)算法和多核最小二乘支持向量机(MK-LSSVM)的预测新方法用于滚动轴承寿命预测。提取小波包相对能量特征对轴承性能衰退予以描述,提出MSPSO算法对MK-LSSVM模型参数进行优化选取,构造融合多核函数的LSSVM模型实现轴承寿命估计。MK-LSSVM中多核函数的引入克服了单核LSSVM对核函数类型强依赖性的弱点,MSPSO算法中种群全局大尺度均匀变异与个体局部邻域小尺度变异搜索联合策略的提出在增强种群多样性的同时保证了粒子群局部精确搜索的能力。利用实测滚动轴承振动数据分析,验证了所提MSPSO算法在模型参数优化及优化MKLSSVM模型在滚动轴承寿命预测应用中的有效性。

A novel prediction method based on both multi-scale mutation particle swarm optimization（ MSPSO） and multi-kernel least square support vector machine（ MK-LSSVM） is proposed for the life prediction of rolling element bearing. The relative wavelet packet energy features are extracted to characterize the patterns of bearing performance degradation. A particle swarm optimization（ PSO）algorithm with multi-scale mutation operation is proposed and used to optimally select the parameters of MK-LSSVM model. The LSSVM model that fuses multiple kernel functions is constructed to realize the bearing life prediction. The introduction of multiple kernel functions in MK-LSSVM overcomes the strong dependence of single kernel LSSVM on the type of the kernel functions,the proposing of the combined strategy of population global large-scale random mutation and particle local neighborhood small-scale mutation（ searching）in MSPSO algorithm not only increases the population diversity,but also ensures the accurate particle swarm local searching capability.The analysis results of actual rolling bearing vibration data verify the effectiveness of the proposed MSPSO algorithm in model parameter optimization and the optimized MK-LSSVM model in rolling bearing life prediction.