An adaptive gamma process based model for residual useful life prediction

This paper proposes a model to predict the residual useful life of a component by condition monitoring. An adaptive gamma process is used to describe the deteriorating nature of the observed condition indicator but one of the parameters of the gamma model is updated whenever a new observation of the indicator becomes available. The updating is performed by means of a state space model where the parameter is the hidden state variable and the observations are the condition monitoring information. Other unknown model parameters are estimated using the expectation maximization algorithm. We apply the model developed to a case study involving a data set of crack growths and demonstrate the validity of this modeling approach.