Title: T-PdM: a tripartite predictive maintenance framework using machine learning algorithms

Authors: Ozlem Ece Yurek; Derya Birant; Alp Kut

Addresses: Graduate School of Natural and Applied Sciences, Dokuz Eylul University, Izmir, Turkey ' Department of Computer Engineering, Dokuz Eylul University, Izmir, Turkey ' Department of Computer Engineering, Dokuz Eylul University, Izmir, Turkey

Abstract: The purpose of this paper is to propose new predictive maintenance (PdM) framework that has three aims: 1) estimating the remaining useful life (RUL) of a machine; 2) classifying machine health status (failure/non-failure); 3) discovering the relationship between the errors and component failures of machines by using machine learning (ML) techniques. This is the first PdM framework that integrates three ML paradigms (regression, classification and association rule mining) in a single platform. It compares six different ML algorithms. The results indicate that the proposed framework can be successfully used to get valuable knowledge about machines and to build a consistent maintenance strategy to improve machine utilisation in the industry sector. The existing PdM studies usually use only one ML paradigm, remaining insufficient for prediction. To overcome this limitation and improve prediction accuracy, a novel tripartite predictive maintenance (T-PdM) framework is proposed in this study.

Keywords: predictive maintenance; machine learning; classification; regression; association rule mining; ARM.

DOI: 10.1504/IJCSE.2022.123122

International Journal of Computational Science and Engineering, 2022 Vol.25 No.3, pp.325 - 338

Received: 29 Apr 2021
Accepted: 10 Aug 2021
Published online: 30 May 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article