电光与控制, 2016, 23 (7): 93, 网络出版: 2021-01-27
基于随机Wiener过程的剩余寿命预测贝叶斯方法
A Bayesian Method for Residual Lifetime Prediction Based on Random Wiener Process
剩余寿命预测 贝叶斯推断 Wiener过程 现场实测数据 常规退化数据 residual lifetime prediction Bayesian inference Wiener process online measured data normal degradation data
摘要
针对性能退化过程服从Wiener过程的产品, 提出了一种融合产品现场实测性能退化数据与同类产品常规退化数据的剩余寿命预测方法。假定分布参数服从联合共轭先验正态-逆伽玛分布, 建立了个体现场实测退化数据下分布参数的贝叶斯估计模型, 给出了超参数后验估计公式; 建立了退化试验数据的完全似然函数, 构建了基于EM算法的超参数先验估计模型; 通过实例验证了方法的正确性。
Abstract
Aiming at the product whose performance degradation process obeys to Wiener process, a residual lifetime prediction method is proposed by combining online measured performance degradation data of a product with normal degradation data of like products. Assume that distribution parameters obey to union conjugate prior normal-inverse gamma distribution, a Bayesian estimation model of distribution parameters is built up under individual online measured degradation data. Posterior estimation formula of hyper parameter is obtained. The complete likelihood function of degradation test data is built up. Prior estimation model of hyper parameter is established based on Expectation Maximization (EM) method. Accuracy of the method is verified by an example.
蔡忠义, 陈云翔, 张亮, 张浪军. 基于随机Wiener过程的剩余寿命预测贝叶斯方法[J]. 电光与控制, 2016, 23(7): 93. CAI Zhong-yi, CHEN Yun-xiang, ZHANG Liang, ZHANG Lang-jun. A Bayesian Method for Residual Lifetime Prediction Based on Random Wiener Process[J]. Electronics Optics & Control, 2016, 23(7): 93.
PDF全文
