中文摘要：

针对产品性能退化数据样本个数少、退化轨迹存在非线性与随机性的特点,提出了一种灰色时序组合模型对产品的退化轨迹进行建模并实时预测个体寿命。首先,采用灰色系统GM（1,1）模型和时间序列AR（p）模型分别对同类产品退化数据中的趋势项与随机项进行预测,构造灰色时序组合预测模型来建立同类产品的退化轨迹。然后,根据K均值聚类理论计算特定个体与同类产品退化轨迹的相似度权值,通过加权同类产品的退化轨迹来获得特定个体的退化轨迹;最后,通过个体实测退化数据更新退化模型并实时预测寿命。将本文方法用于某电子产品的寿命预测中,试验结果验证了该方法的准确性与有效性。

英文摘要：

To solve the problem that the products have few performance degradation data and nonlinear stochastic degradation paths, a grey time series combined forecasting model is proposed to build the model of degra- dation paths and predict the product lifetime in realtime. Firstly, in order to establish degradation paths for the same kind of products, the GM（1,1） grey model is used to forecast the trend term and the AR（p） time series model is used to forecast the stochastic term. Then, similarity weights of degradation paths between the specific individuals and the same kind of products are evaluated based on the K-means clustering theory, and the specific individual degradation path model is built by weighting models of the same kind of products. Lastly, the specific individual degradation path model is updated with real-time degradation data and the lifetime is predicted. This method is used to predict lifetime of certain electronic products, and the experiment results verify its accuracy and validity.