摘要
为提高轴承剩余使用寿命预测精度,提出一种基于改进PSO-SVR的轴承剩余使用寿命预测方法。选取轴承水平和垂直方向振动信号均方根、峰值因子、峭度因子等参数构造多维退化特征,建立基于SVR的轴承剩余使用寿命预测模型;针对SVR参数优化问题,设计一种动态自适应异步粒子群优化算法,引入G worst修正了速度位置更新公式,改进了一种基于倒S形函数的自适应惯性权重系数和一种基于惯性权重系数的异步自适应学习因子,能够有效克服局部最优,加快收敛效率,提高回归精度。仿真实验结果表明:提出的方法与GS-SVR、GA-SVR、PSO-SVR、MPSO-SVR相比,具有较高的预测效率和预测精度,预测精度均优于GBDT、RF、DT、GP等经典回归预测方法。
In order to improve the prediction accuracy of the bearing remaining useful life,a prediction method of bearing remaining useful life based on improved PSO-SVR model was proposed.The degradation characteristics were constructed by selecting the root mean square,crest factor,kurtosis factor of bearing horizontal and vertical vibration signals,and the prediction model of bearing remaining useful life based on SVR was established.In order to optimize the parameters of SVR,a dynamic adaptive asynchronous particle swarm optimization algorithm was designed,G worst was introduced to correct the velocity and position formula,an adaptive inertial weight coefficient based on the inverted S-type function was proposed,and an adaptive asynchronous learning factor based on the inertial weight coefficient was presented,which could effectively overcome the local optimization capability and accelerate the convergence efficiency,enhance the regression accuracy.The simulation results show that the proposed method has more prediction efficiency and accuracy than GS-SVR,GA-SVR,PSO-SVR,MPSO-SVR,it is superior to the classical regression methods,such as GBDT,RF,DT and GP.
作者
张成龙
刘杰
李想
ZHANG Chenglong;LIU Jie;LI Xiang(Department of Brewing Engineering Automation,Moutai Institute,Renhuai Guizhou 564500,China;School of Mechanical Engineering,Guizhou University,Guiyang Guizhou 550025,China)
出处
《机床与液压》
北大核心
2020年第16期206-211,共6页
Machine Tool & Hydraulics
基金
国家自然科学基金面上项目(51475097)
贵州省教育厅青年科技人才成长项目(黔教合KY字[2018]458)。
关键词
轴承
剩余使用寿命
支持向量回归
粒子群优化算法
健康管理
Bearing
Remaining useful life
Support vector regression
Particle swarm optimization
Health management
