随着先进传感与监测技术的快速发展，实时获取随机退化设备的多源传感监测数据已成为现实，如何有效融合多源传感监测数据以实现随机退化设备剩余寿命的精准预测成为剩余寿命预测领域的研究前沿。针对多源传感监测的线性随机退化设备，提出了一种考虑随机失效阈值的数模联动剩余寿命预测新方法。该方法在离线训练过程中，基于多源传感历史数据提取的复合健康指标及据此线性随机退化建模预测的寿命，构建综合寿命预测值与设备实际寿命的均方误差及寿命预测方差的优化目标函数，形成复合健康指标提取与随机退化建模的反馈闭环，对多源传感器融合系数和复合健康指标对应的随机失效阈值分布参数进行优化调整，以实现复合健康指标提取与随机退化建模的自动匹配，即数模联动。在线预测时，根据提出的数模联动方法，融合实际运行设备的多源传感监测数据以获取复合健康指标，然后采用随机模型对其演变过程进行建模。同时，为使模型实时反映设备当前状况，提出了一种退化模型参数的贝叶斯更新方法，在此基础上基于首达时间得到了考虑设备失效阈值随机性的剩余寿命概率分布。最后，基于航空发动机的多源传感监测数据，验证了所提方法在改善复合健康指标特性和提高剩余寿命预测准确性方面的有效性和优势。

Advances in sensing and monitoring techniques enable real time acquisition of multi-sensor monitoring data of stochastic degrading devices. The effective fusion of these multi-sensor monitoring data to precisely predict the RUL in the prognostics field has drawn extensive attention. This paper presents a novel data-model interactive RUL prediction method for multi-sensor monitored linear stochastic degrading devices with a random failure threshold. In the offline training part， an optimization objective function synthesizing the mean squared error between the predicted and the actual life as well as the variance of the predicted life is constructed based on the composite health index extracted from historical multi-sensor data and the associated lifetime prediction via stochastic degradation modeling. A closed-loop feedback mechanism is thus established for composite health index constructing and stochastic degradation modeling. Based on this feedback mechanism， the fusion coefficients for multi-sensor data and the distribution parameters of the random failure threshold of the associated composite health index are optimized to enable automatic matching of the constructed composite health index with the adopted stochastic degradation model. In the online prediction part， the composite health index is first constructed based on the multi-sensor data of the device in service and then the linear stochastic degradation model is applied to track the degradation progression. To adapt the degradation model to the current status of the device， the Bayesian method is proposed to update the model parameters， and the RUL distribution considering the randomness of the failure threshold is subsequently derived according to the concept of the first passage time. Finally， we validate the proposed method by the multi-sensor data of aircraft gas turbine engines and the results indicate its merits in improving the properties of the composite health index and the accuracy of the prognosis.