Abstract: To comprehensively consider the nonlinearity of the stochastic error term,a generalized Wiener process based remaining life prediction method was proposed for rolling bearings.To determine the degradation starting point,3σ criteria was adopted for recognizing the change point.Aiming at the evolution characteristics of the degradation stage,a generalized Wiener process-based degradation model was constructed.The approximate analytical expressions of the distribution and the point estimation of bearing life were derived under the concept of first hitting time.By fusing the historical degradation data,a maximum likelihood estimation method was established to estimate the unknown model parameters.Then,an online updating method by combining the online monitoring data was constructed via Bayesian theory to predict remaining life.Finally,the effectiveness of the proposed method was verified by a real application.The results showed that the prediction accuracy of the proposed method could be improved more than 50% compared with current approaches.

Key words: rolling bearing, remaining life prediction, generalized Wiener process, Bayesian theory, change point recognition

摘要： 为解决传统滚动轴承剩余寿命预测方法忽略随机误差项非线性的问题,提出一种基于广义Wiener过程的滚动轴承剩余寿命预测方法。为确定轴承初始退化点,利用3σ准则进行变点识别;针对退化阶段演化特征,建立基于广义Wiener过程的退化模型,根据首达时概念,推导出轴承寿命分布模型和点估计近似解析式。融合同类型轴承历史性能退化数据,提出退化模型未知参数极大似然估计方法;其次,结合轴承在线监测数据,建立基于贝叶斯理论的退化模型随机参数在线更新方法,实现滚动轴承剩余寿命在线预测。最后,通过滚动轴承工程实例分析,验证了所提方法的有效性和适用性,与现有文献方法相比,所提方法的预测精度提高了50%以上。

关键词: 滚动轴承, 剩余寿命预测, 广义Wiener过程, 贝叶斯理论, 变点识别