Review of Common Evaluation Methods for Life Prediction Models

Abstract

As one of the core parts of Prognostics and Health Management (PHM) technologies, residual useful life (RUL) prediction is a very important concept in decision making and contingency mitigation. With life prediction models, researchers could obtain the prediction RUL of different objects. However, since sometimes there will be several available prediction models to be chosen, evaluation methods or selection methods) for life prediction models should be proposed to help choosing models that suit for certain objects. The most important factors that affect the performance of prediction models include prediction accuracy, data fitness, model complexity and parameter sensitivity. This paper presents some common evaluation methods for life prediction models that have already been used in this area.