Towards prediction of machine failures: overview and first attempt on specific automotive industry application

Abstract

The automotive sector is facing new challenges and increased competition nowadays. Customer satisfaction depends on products and parts quality, as well as possible customizations. To reach these objectives, productivity is key, meaning machines availability needs to be maxed and not impacted by unplanned breakdowns, which cost a lot of money and time, and possibly quality issues on parts produced during the deteriorating phase of the machine. Industry 4.0 will play an essential role, as it comes with new digital tools to improve productivity through real-time interactions from the digital world to the physical world. It is especially true with the maintenance policies, which are changing from corrective to planned ones from predictions of machine failures. We use the terms Condition-Based Maintenance (CBM) or Predictive Maintenance (PdM) in these cases; they are based on data analysis to propose a health assessment of critical components, to predict future issues. The Prognostics and Health Management (PHM) framework proposes methodologies to deal with such problems. In the recent years, many researches focused on these topics; however, few of them deal with the full scope of implementing practically this strategy in the industry, particularly in the automotive sector. Thus, this paper aims at reviewing current approaches, and presenting the strategy employed as well as the first use case investigated in the Clean Energy Systems division of Plastic Omnium.