一种相似性框架下基于非线性扩散过程的剩余寿命估计模型

A model for residual life estimation based on the nonlinear diffusion process under the framework of similarity

剩余寿命(residual life,RL)估计是预测与健康管理(prognostics and health management,PHM)的关键环节。目前,传统的基于相似性的RL估计模型仅利用失效设备的退化数据,忽略截断设备和运行设备的退化数据,难以保证RL的估计精度。针对该问题,在相似性框架下提出一种新的基于扩散过程的RL估计模型。首先基于扩散过程对截断设备进行退化建模和RL估计,然后通过比较参考设备(失效设备与截断设备)和运行设备间的相似性,同时利用参考设备与运行设备的退化数据实现运行设备的RL估计。最后仿真实验验证该文模型优于传统的基于相似性的模型。研究结果表明:该文模型能够显著提高运行设备RL的估计精度,具有潜在的工程应用价值。

Residual life（RL） estimation is a key part in prognostics and health management （PHM）. In current literature, traditional similarity-based RL estimation model can only utilize the degradation data of failed devices, while ignoring the degradation data of suspended devices and operating device, which cannot guarantee the accuracy of the RL estimation. Aiming at this issue, this paper develops a new diffusion process driven RL estimation model in the framework of similarity. Firstly, the degradation model is constructed and the RL of the suspended devices is estimated. And then the observed degradation data from the reference devices （i. e. failed and suspended devices） and the operating device are utilized to estimate the RL of the operating device by comparing the similarity between the operating device and reference devices. Finally, a numerical simulation is provided to substantiate the superiority of the proposed model over the traditional similarity - based approach. The research result shows that the proposed model can remarkably improve the accuracy of the RL estimation for the operating device, which can be potentially applied in practice.