Exploring ML for Predictive Maintenance Using Imbalance Correction techniques and SHAP

Abstract

This paper focuses on an application of machine learning in industries - predicting machine failure from sensor data for maintenance purposes. The primary purpose of this paper is to use machine learning models to predict whether a machine is going to fail or function normally. Various machine learning techniques are implemented and evaluated on a synthetic dataset and then a real-world dataset. This method allows a comparison to be drawn between a theoretical approach and real-world application. This paper first introduces the various machine learning (ML) models and describes the datasets. Then, the best performing models are further developed and discussed. Lastly, the models are evaluated quantitatively (using performance metrics) and qualitatively (by Shapley Additive Explanations, SHAP values).