Prediction of the Remaining Useful Life Distribution Based on the Gamma Process and FC-SVR Method



Remaining Useful Life (RUL) prediction is a key issue in equipment Prognostics and Health Management (PHM). It plays an important role in condition-based maintenance. Several previous studies have used the degradation process model to estimate the RUL distribution, but the majority of existing methods consider only a single characteristic variable, ignoring the effects of multi-characteristic variables on the RUL prediction. This study proposes the fusion character SVR (FC-SVR), which considers the influence of multi-characteristic variables. The degradation process of the characteristic variable is modelled by the gamma process. The regression model between the degradation process of the characteristic variables and the RUL distribution is established by applying the FC-SVR method. This regression model can be used to predict the RUL distribution change process and the rules governing the time-varying reliability. The feasibility of this model is verified using detailed examples.