Predictive Maintenance of ATM Machines by Modelling Remaining Useful Life with Machine Learning Techniques

One of the main relevant topics of Industry 4.0 is related to the prediction of Remaining Useful Life (RUL) of machines. In this context, the “Smart Manufacturing Machine with Predictive Lifetime Electronic maintenance” (SIMPLE) project aims to promote collaborations among different companies in the scenario of predictive maintenance. One of the topics of the SIMPLE project is related to the prediction of RUL of automated teller machines (ATMs). This represents a key task as these machines are subject to different types of failure. However the main challenges in this field lie in: i) collecting a representative dataset, ii) correctly annotating the observations and iii) handling the imbalanced nature of the dataset. To overcome this problem, in this work we present a feature extraction strategy and a machine learning (ML) based solution for solving RUL estimation for ATM devices. We prove the effectiveness of our approach with respect to other state-of-the-art ML approaches widely employed for solving the RUL task. In addition, we propose the design of a predictive maintenance platform to integrate our ML model for the SIMPLE project.