An Experimental Comparison of Supervised Machine Learning Models for Health-based Classification of Remaining Useful of Life in Turbofan Engines

Abstract

Prediction of remaining useful life (RUL) of a component in a manufacturing line is important to predictive maintenance. In this paper, we describe  a data driven approach to using machine learning-based techniques for automating the failure prediction of equipment. The performance of the machine learning models are measured on their precision, recall, F1 scores and predicting the health of the engine from 21 features. This experimental analysis based on a new dataset from NASA on turbofan engines shows that KNN classifier performs the best in modeling a health indicator for this problem.