Data analysis in maintenance

Conférence plénièreInternational audiencechallenging applications in the new era of “Factory of the Future” or “Industry 4.0” paradigms. Indeed, predictive maintenance allows to take pro-active measure to predict and avoid system failure from asset abnormality control. That represents a major advantage to increase the asset uptime and reduce overall maintenance costs while mastering the upfront investment necessary. This anticipation feature is also very useful to ease logistics by maintaining system at convenient time impacting positively the logistics support organization, for example by predicting which resources (e.g. operator, spare parts, tools) will be needed and when. It means that predictive maintenance can have an impact both on the targeted system but also on its support system for making the targeted system as maintained before failure but operating as long as possible without interruption. This philosophy is those defended by the PHM (Prognostics and Health Management) community promoting an integration of the maintenance activities/processes as a whole (a global added value chain).In the predictive vision, these processes are encompassing a variety of data-based transformation allowing to implement condition monitoring, failure/degradation detection, failure/degradation diagnostics (root cause analysis), failure prediction, … and recommendation/development of dynamic maintenance actions to keep the (degraded) asset in its nominal operating space. Thus data processing is a very important tool for predictive maintenance leading to consider the technologies and methodologies referred to advanced data analytics (e.g. internet of things, big data, artificial intelligence) as a main pillar to offer predictive capabilities and deploy a new way of organizing and implementing proactive maintenance on an industrial scale.In that way, the aim of this presentation is to survey with illustrations and case-studies the key developments in data analytics-based predictive maintenance, and its emerging challenges. It will address issues such as:-How are recent developments in IoT (Internet of Things), Big Data, Deep Learning … Predictive Analytics are impacting on Preventive Maintenance and more precisely Predictive Maintenance activities? -What are the benefits proposed by such developments but also their constraints and limits in link to industrial requirements?-How the maintenance organization should evolve to take advantage of the opportunities that these developments enable