Lifetime models for remaining useful life estimation with randomly distributed failure thresholds

Abstract

In order to predict in advance and with the smallest possible uncertainty when a component needs to be fixed or replaced, lifetime models are developed based on the information of the component deterioration trend and its failure threshold to estimate the stochastic distribution of the hitting time (the first time the deterioration exceeds the failure threshold) and the remaining useful life. A primary issue is how to effectively handle the uncertainties related to the component deterioration trend and failure threshold. This problem is here investigated considering a non-stationary gamma process to model the component deterioration and a gamma-distributed failure threshold. Two lifetime models are proposed for comparison on an application concerning deterioration of choke valves used in offshore oil platforms.