A Data-driven Prognostic Model for Industrial Equipment using Time Series Prediction Methods

Abstract

Condition-based maintenance strategy is considered popular and received high demand in industry to ensure high availability and reliability of equipment in the plant. Prognostic is one of an important functions in condition-based maintenance strategy which is used to predict the future condition of the observed and estimate the remaining useful lifetime (RUL) based on the current and historical condition data. Due to the fact that most of the current automated equipment in industry has the capability to capture and store the condition and process data during operation, the research aimed to formulate a prognostic model based on the integration of the data and predict the series of future condition. This paper presents a data-driven prognostic model to predict the estimated RUL by using condition and process data which are taken from a single unit of equipment. The structure of prognostic model is presented and two time series methods are employed namely Artifical Neural Network and Double Exponential Smoothing in prognostic process. The feasibility of this prognostic model was demonstrated with applying real data from industrial equipment. The result from the model shows that both of the methods are able to extrapolate the extimated  RUL  and  give  useful  information to the maintenance department to take an appropriate decision.