A degradation evaluation method based on data from multi-conditions

Product´s degradation data is a major data source for inferring its failure time distribution model, which is usually used in application for Accelerated degradation testing (ADT) and Prognostic and Health Management (PHM), and has become a common approach to the reliability prediction for those highly reliable products. However, sometimes in the actual application, there is only one product and it is successively used in different stress conditions, like different tests, different service environment and so on. As the stresses´ type and amount are different in these conditions, the product´s degradation process is complex and the data size is small in each condition, so it is hard to utilize them for reliability extrapolation with the previously modeling and evaluation method. To solve the problems above, firstly, the general log-linear model is introduced to take into the consideration of the relationship between different stress conditions and their effects on the degradation process, in which the stress variables and various conditions are all described as parameters in an entirety model. Secondly, to understand the relationship and to estimate the reliability accurately, Bayesian approach is introduced to fuse this complex and high dimensional model into the reliability extrapolation and integrate all available information to infer unknown parameters. Then, with the obtained parameters, the accumulated effects from different conditions are quantified, and the reliability of the product can be estimated or predicted. Finally, the proposed method is demonstrated by one example and relevant simulation analysis.