基于累积KL散度和改进粒子滤波的滚动轴承剩余使用寿命预测  

Remaining Useful Life Prediction of Rolling Bearings Based on Cumulative KL Divergence and Improved Particle Filter

作　　者：徐中 王继承 刘东林 曾春 麻浩军 XU Zhong;WANG Jicheng;LIU Donglin;ZENG Chun;MA Haojun(China Nuclear Power Operation Management Co.,Ltd.,Haiyan 314396,China;School of Mechanical Engineering,Xi'an Jiaotong University,Xi'an 710049,China)

机构地区：[1]中核核电运行管理有限公司,浙江海盐314396 [2]西安交通大学机械工程学院,陕西西安710049

出　　处：《机械制造与自动化》2024年第5期183-190,共8页Machine Building & Automation

摘　　要：针对大量传统退化指标未考虑轴承在服役过程中的累积退化特性以及常规粒子滤波算法存在粒子退化和粒子多样性不足导致剩余使用寿命预测困难的问题,提出一种基于累积KL散度退化指标结合改进粒子滤波的轴承剩余使用寿命预测方法。利用累积缩放变换将从轴承振动信号中提取的原始KL散度转换为映射特征以优化其单调性与趋势性,构建累积KL散度退化指标;根据退化指标建立双指数退化模型,并利用灰狼算法优化粒子滤波的采样过程,引入残差重采样方法解决粒子退化问题,实现改进粒子滤波递推预测轴承剩余寿命。分别在6312/C3轴承全寿命实验数据与XJTU-SY公开轴承数据集上进行验证,利用对比实验证明了所提出的累积KL散度退化指标结合改进粒子滤波预测方法相比常规粒子滤波预测方法具有更高的预测精度。In view of the difficulty in predicting remaining useful life of bearing due to the neglection of cumulative degradation attribute of in-operation bearings in traditional degradation indexes and the particle degradation of conventional particle filter algorithm and insufficient particle diversity,a remaining useful life prediction method of rolling bearings was proposed based on feature cumulative KL divergence combined with improved particle filter.The cumulative KL divergence degradation index was constructed by converting the original KL divergence extracted from the bearing vibration signal into a mapping feature to optimize its monotonicity and tendency.A double exponential degradation model was established according to the degradation index,the sampling process of particle filter was optimized by using the gray wolf algorithm,and the residual resampling method is introduced to solve the particle degradation problem,so as to predict the remaining useful life with improved particle filter.Based on the 6312/C3 bearing run-to-failure experimental data and the XJTU-SY public bearing data set,the comparative experiment proves that the proposed cumulative KL divergence degradation index combined with the improved particle filter prediction method has higher prediction accuracy than the conventional one.

关 键 词：滚动轴承 剩余使用寿命 累积KL散度 粒子滤波 

分 类 号：TH133.33[机械工程—机械制造及自动化]