Prognostics of Machine Health Condition using an Improved ARIMA-based Prediction method

Abstract

Prognostics is very useful to predict the degradation trend of machinery and to provide an alarm before a fault reaches critical levels. This paper proposes an ARIMA approach to predict the future machine status with accuracy improvement by an improved forecasting strategy and an automatic prediction algorithm. Improved forecasting strategy increases the times of model building and creates datasets for modeling dynamically to avoid using the previous values predicted to forecast and generate the predictions only based on the true observations. Automatic prediction algorithm can satisfy the requirement of real-time prognostics by automates the whole process of ARIMA modeling and forecasting based on the Box-Jenkins's methodology and the improved forecasting strategy. The feasibility and effectiveness of the approach proposed is demonstrated through the prediction of the vibration characteristic in rotating machinery. The experimental results show that the approach can be applied successfully and effectively for prognostics of machine health condition.