A lifetime prediction method with hierarchical degradation data

Abstract

Degradation data is an important information source which is usually used to predict products' lifetime, for instance in accelerated degradation testing (ADT) and health management. Degradation data can be easier and cheaper obtained than failure data. As a result, it has been widely applied. However, due to some restrictions of funds and the development cycle, the degradation data of some products might not be adequate enough to predict the lifetime accurately. On the other hand, for the product with hierarchical structure whose sub-systems sometimes have more available degradation data, if these data can be utilized in the product's prediction appropriately, the prediction accuracy of the lifetime would be improved. In order to utilize the degradation data from hierarchical structure appropriately, firstly, feasible degradation data from system and sub-systems are collected and classified. Secondly, the method of support vector machine (SVM) is introduced to model the relationship among hierarchical degradation data, and then all degradation data from sub-systems are integrated and transformed to the system degradation data. Thirdly, with the processed information above, a prediction method based on Bayesian theory is proposed, and the hierarchical product's lifetime is obtained. Finally, an energy system is taken as an example to explain and verify the method in this paper, and the method is also suitable for other products.