Predicting Remaining Useful Life of Engines Using SVR and CNN

Abstract

Engines’ Remaining Useful Life (RUL) prediction is a considerable issue to realize Prognostics and Health Management (PHM) that is being widely applied in many industrial systems to ensure high system availability over their life cycles. This work presents a data-driven method of RUL prediction based on two Machine Learning (ML) techniques, mainly Support Vector Machine (SVM) for Regression or Support Vector Regression (SVR) and Convolutional Neural Network (CNN). These techniques are applied on the NASA C-MAPSS turbofan engine dataset. To extract the input features, the dataset was analyzed with the help of plots and a filter-based feature selection technique known as Mutual Information (MI). the resulting features are then fed to both models. Although SVM and CNN algorithms are mostly used in classification problems, their effectiveness in estimating the RUL, which is a regression problem, is demonstrated and compared to some state-of-the-art methods. The results show that the SVR and CNN models provide approximately similar performance in predicting the RUL for the used dataset.