Performance-based Reliability Prediction of Power Supply Considering Degradation Uncertainties

Abstract

Reliability prediction is an effective approach to find weak links and evaluate whether the product satisfies its reliability requirement during the design phase. For the highly reliable product, it is far difficult to predict its reliability level according to the traditional method that largely relies on historical data from reliability handbooks. This paper proposes a performance-based reliability prediction method, which makes use of degradation to reflect product's health conditions. Additionally, a degradation model considering multivariate uncertainties such as operation conditions and parameter randomness is proposed to describe soft failures induced by degradation. A case study of power supply is illustrated to show the effectiveness and feasibility of the proposed method and degradation model for reliability prediction.