果蝇算法的MPSR-MKSVR轴承剩余寿命预测综合优化  

Comprehensive Optimization of Bearing Residual Life Prediction Based on Fruit Fly Algorithm MPSR-MKSVR

作　　者：陶沙沙 江星星[2] TAO Sha-sha;JIANG Xing-xing(Chendu Vocational and Technical College of Industry,Sichuan Chengdu 610000,China;College of Urban Rail Transit,Suzhou University,Jiangsu Suzhou 215137,China)

机构地区：[1]成都工业职业技术学院,四川成都610000 [2]苏州大学城市轨道交通学院,江苏苏州215137

出　　处：《机械设计与制造》2022年第7期52-58,共7页Machinery Design & Manufacture

基　　金：国家自然科学基金资助项目(51705349);四川省教育厅人文社会科学重点研究基地统筹城乡教育发展研究中心科研项目“基于教师评价激励机制视角的四川高职院校‘双师型’教师队伍建设策略研究”(TCCXJY-2017-B10)。

摘　　要：针对小样本条件下的轴承寿命预测问题,提出一种基于多子种群混沌自适应果蝇算法的多变量相空间重构-多核支持向量回归轴承剩余寿命预测综合优化方法。对多个特征向量进行相空间重构从而更加完整地还原轴承退化特性;利用不同的SVR核函数进行排列组合得到一个组合多核函数,从而将多核函数的构造问题转换为权值系数的优化问题;将延迟时间、嵌入维数、SVR参数以及多核函数权值作为整体参数向量,利用多子种群混沌自适应果蝇算法对训练数据预测精度的适应度函数进行优化,从而得到最优的预测模型参数。最后将得到的预测模型对轴承剩余寿命进行预测,结果证明了提出方法能够有效提升小样本条件下的轴承剩余寿命预测精度。To solve the problem of bearing life prediction under small sample size,a multiple variable phase space reconstructionmultiple kernels support vector regression(MKSVR)method based on multiple sub-population chaotic adaptive fruit fly algorithm was proposed.The phase space reconstruction of multiple eigenvectors was used to restore the bearing degradation characteristics more completely;a combined multiple kernels function was obtained by arranging and combining different SVR kernels,thus the construction of multiple kernels function was transformed into the optimization of weight coefficients;the delay time,embedding dimension,SVR parameters and the weight of multiple kernels function were taken as the whole parameter vectors,and the adaptive fruit fly algorithm based on population chaos optimizes the fitness function of the prediction accuracy of training data to obtain the optimal parameters of the prediction model was used.Finally,the prediction model was used to predict the bearing residual life.The results show that the proposed method can effectively improve the bearing residual life prediction accuracy under the condition of small samples.

关 键 词：轴承 相空间重构 支持向量回归 果蝇算法 剩余寿命预测 

分 类 号：TH16[机械工程—机械制造及自动化] V263.6[航空宇航科学与技术—航空宇航制造工程]