Multi-step Ahead Forecasting for Fault Prognosis Using Hidden Markov Model

With the development of the system scale and complexity, the demand for reliability and safety of dynamic system grows rapidly. Fault prognosis, aimed at predicting the future health condition of the system, plays an important role in system safety and reliability. In this paper, we present two multi-step ahead forecasting methods for fault prognosis using Hidden Markov Model (HMM). One predicts each state applying a one-step ahead forecasting model recursively and uses the predicted value to determine the next time step value, and the other one predicts each state independently from the others just using the previous known observations. The proposed methods are exploited to estimate the next long-term system states and predict the development trends of a fault in Tennessee Eastman (TE) chemical process. We perform a comparative study on the prediction performance of these two approaches. And our methods are proved to be effective for fault prognosis from the experimental application.