针对退化数据的异质性问题, 提出基于随机效应混合回归的贝叶斯退化建模与剩余使用寿命预测方法.通过模型随机参数的混合概率分布描述异质性退化数据的统计特征, 并利用分层先验和Gibbs算法进行贝叶斯参数估计.进而, 将随机参数的贝叶斯估计作为有信息先验, 利用在线观测数据进行贝叶斯参数更新和剩余使用寿命估计.通过涡轮发动机退化数据验证了所提方法的剩余使用寿命预测性能.实验结果表明, 考虑了异质性影响的贝叶斯混合回归退化建模方法, 能更好地刻画设备个体的退化特征, 提高了预测准确度.%A Bayesian degradation modeling and remaining useful life prediction method is proposed for heterogeneous degradation data based on random-effects mixture regression model.The statistical characteristics of heterogeneous degradation data are described by mixture probability distribution of random parameters of degradation model, and the Bayesian parameters estimation is carried out by hierarchical priors and Gibbs algorithm.Furthermore, the Bayesian estimation of random parameters is taken as informative prior toBayesian parameters updating and remaining useful life estimation with online observations.The prediction performance of the proposed method is verified by turbine engines data.The experimental results show that the Bayesian mixture regression model considering the influence of heterogeneity can better describe the degradation characteristics of individual equipment and improve the accuracy of prediction.
展开▼