Prediction of remaining useful life for mech equipment using regression

Abstract

Maintenance has always been an important function in any manufacturing operations. Recently, Condition Based Maintenance (CBM) prognosis approach is gaining popularity in prediction of mechanical equipment remaining useful life (RUL). However, there is a need to improve the existing RUL prediction approach for mechanical equipment with multiple components. In this paper regression tree is used for developing the RUL prediction model of multiple components. A widely investigated dataset, PHM 2008 from NASA Prognostic Center was used in this study. Seven out of 21 sensors data were selected and used for prediction modeling. Before the data can be used, it must be filtered, clustered and normalized. Then the regression tree approach was used to develop the prediction model. The proposed regression tree model gave almost comparable results to other prediction methods such as Dempster-Shafer Regression (DSR), Support Vector machine (SVM) and Recurrent Neural Network (RNN). Besides, regression tree provides simplicity and the ability to manage large dataset.