Developing machine learning-based models for optimal maintenance scheduling within the food industry

Abstract

In the context of the food industry, precisely the company used as a case study, unplanned machine downtimes are mainly caused by ineffective maintenance scheduling that impacts organisational profit. To address this challenge, the study explores predictive maintenance within the food industry and its application to reduce machine inefficiencies and improve overall decision-making. First, a theoretical background on predictive maintenance and machine learning is provided, followed by the development of the random forest and decision tree models. Company data is pre-processed, and the models are trained and tested using scientific methods from academic literature, including cross-validation and hyperparameter tuning. One-year future predictions are made for three retail line machines, aiding in proactive maintenance decision-making to reduce unplanned machine downtime. Subsequently, this study contributes towards academia and industry by providing actionable insights for optimising maintenance scheduling and production processes in the food industry.