Predictive maintenance
by Prof. dr. Geert-Jan van Houtum
29th April 2020
About Geert-Jan van Houtum
Geert-Jan van Houtum is Professor of Maintenance, Reliability, and Quality at the Department Industrial Engineering and Innovation Sciences (IE&IS) of Eindhoven University of Technology since 2008. Prior to that he filled positions as assistant and associate professor at the same department (1999-2007) and the University of Twente (1994-1998) and as visiting professor at Carnegie Mellon University (2001). He obtained his M.Sc and Ph.D. degree in Applied Mathematics from Eindhoven University of Technology in 1990 and 1995, respectively.
His research is focused on the maintenance and reliability of capital goods, and in particular on: (i) Design and control of service supply chains; (ii) Maintenance concepts, in particular predictive maintenance; (iii) Design for availability. He has over 80 publications in international refereed journals such as Operations Research, Manufacturing and Service Operations Management, IIE Transactions, and European Journal of Operational Research. He is area editor at Service Science and associate editor at Manufacturing and Service Operations. Much of his research is in cooperation with the industry. He works with companies such as ASML, Canon, Dutch Railways, Philips, Marel, the Royal Dutch Airforce, the Royal Dutch Navy, Thales, and Vanderlande. He is vice-dean IE of the Department IE&IS since September 2017. Further, he is a board member of the Service Logistics Forum.
You can find here a list of his publications.
Abstract
About 10 years ago, we started with predictive maintenance research in the Netherlands. In my projects, we studied systems in the high-tech, maritime, and chemical industry. In this presentation, I present a general predictive maintenance approach that works for systems where a limited set of components causes most of the failures. This approach builds on stochastic processes, data mining, Bayesian learning, machine learning, and Operations Research techniques. We will also discuss what we can investigate in the coming 10 years. An important learning point of the past 10 years is that data analysis methods often lead to predictions with a certain percentage of false positives. That is often not good enough for users of systems to replace a component or module preventively. But these predictions can still be useful to be better prepared when a failure occurs.
The slides are available here.