Predictive Maintenance System Integrated with Periodic Maintenance: Machine Learning and Classical Approaches

EasyChair Preprint 5806

Abstract

With the fourth industrial revolution, the preventive and predictive maintenance, replaced the corrective maintenance, have been popular in recent years. Preventive maintenance is a scheduled maintenance strategy applied to reduce failures. On the other hand, predictive maintenance strategy requires to monitor equipment continuously and to analyze the data. The main objective of the predictive maintenance is to predict problems on the equipment that may lead to stops and to maximize utilization of the machine/equipment. It is reasonable to eliminate some failures with preventive maintenance while predictive maintenance can be applicable to eliminate others. For this purpose, we present a few criteria to determine maintenance strategy that will be applied to eliminate failures. Smartly integrating predictive and preventive maintenance will help to improve sustainability of the system. In this study, the preventive maintenance period is determined considering classical approaches such as Weibull analysis. We analyzed the failures of a specific machine for a period. We also collected data about the system, environment and the machine condition during failures. We utilized machine learning algorithms in order to predict the type of possible failure and associations. The proposed decision support system helps to update the maintenance program with respect to results of machine learning methods. We perform a real-life case study and present our results.

Keyphrases: Industry 4.0, Predictive Maintenance, Weibull analysis, machine learning

@booklet{EasyChair:5806,
  author    = {Kamer Pınar Bahar and Gazi Bilal Yıldız and Banu Soylu},
  title     = {Predictive Maintenance System Integrated with Periodic Maintenance: Machine Learning and Classical Approaches},
  howpublished = {EasyChair Preprint 5806},
  year      = {EasyChair, 2021}}