NORMAL CONDITION MODEL FOR MECHANICAL PROGNOSTICS BASED ON GREY MODEL

Abstract

Prognostics is a process of predicting the future conditions of a product based on an assessment of its current state-of-health and its past performance conditions. Actually, in previous research, the prognostics consists of three steps; namely building normal condition model, estimating degradation state, and predicting future condition as well as assessing remaining useful life of the machine. At the first step, the prognostics method has to identify whether a system or a machine is in good condition or not based on current condition. In other words, a normal condition model must be build first. This paper has been proposed two techniques according to the approach of normal condition modeling. These techniques may support developing of a prognostics method based on grey model. Result shows that prediction performance of both techniques is satisfying by root-mean square error (RMSE) and linear correlation (R) values. Although there is little different in result among the proposed methods, experiment and other research, the prediction performance is enough satisfying. However, using the both approach in a prognostics application may give the best result. For validating the proposed method, data from low methane compressor used in petrochemical industry was employ. The data contains information of machine history with respect to time sequence.