Abstract: There is an implicit assumption that two adjacent moments have the same state quantity of the traditional adaptive Wiener process-based remaining life prediction method.For this problem,a novel remaining life prediction method was proposed by considering the state increment in the adaptive Wiener process.Wiener process was used to characterize the performance degradation process of product and the corresponding incrementing state space equation was established.Then the analytical formulas about the online update parameters in the degradation model were derived.To fully develop and utilize the historical degradation data of similar products,an information fusion method based on Expectation-Maximization (EM) algorithm was proposed to estimate the initial parameter values in the state space equation.Furthermore,the distribution function and the point estimation of the remaining life could be obtained based on the concept of the first hitting.Finally,two engineering examples involving IRLED and spherical plain bearing were used to verify the proposed method.The results showed that compared with the existing methods in the literature,the prediction accuracy of the proposed method was improved by about 40.07% and 101.23%,respectively.

Key words: remaining useful life prediction, state increments, adaptive Wiener process, expectation-maximization algorithm, performance degradation

摘要： 针对传统自适应Wiener过程剩余寿命预测方法具有相邻两个时刻状态量相同隐含假设的问题,提出一种考虑状态增量的自适应Wiener过程剩余寿命预测方法。首先利用Wiener过程来表征产品性能退化过程,建立具有状态增量的Wiener过程状态空间方程,推导出退化模型参数在线更新解析式;为了充分开发利用同类产品的历史退化数据,提出基于期望最大化(EM)算法的信息融合方法,用以估计状态空间方程参数初始值;其次,利用首达时概念,得到产品剩余寿命的分布函数和点估计。最后,结合红外发光二极管IRLED和关节轴承工程实例对所提方法进行验证,与传统方法相比,所提方法的预测精度分别提高了约40.07%和101.23%。

关键词: 剩余寿命预测, 状态增量, 自适应Wiener过程, 期望最大化算法, 性能退化