Key points are not available for this paper at this time.
Determination of the right time for machine maintenance is a major challenge for many industrial companies. Currently, most companies react on occurring breakdowns (reactive maintenance) or maintenance is carried out in scheduled time intervals (preventive maintenance). These results in either unexpected production stops, or a waste of machine working hours, because components are switched too early. Consequently, predictive maintenance strategies offer a big potential. An essential part of predictive maintenance is the estimation of the Remaining Useful Life (RUL) of machine assets. RUL estimation approaches are based on statistical methods and derived algorithms. Thus, a lot of data is needed for a good estimation. Additionally, data can be generated by means of simulation to improve the RUL estimation. However, companies hardly have an overview of available data and according modules, which are needed for a holistic predictive maintenance strategy. This paper shows an approach for a predictive maintenance strategy dealing with acquisition, processing, and analysis of historical field data as well as the generation of respective simulation data. A structured process map with a derived systematic strategy will give companies an idea of how they can integrate predictive maintenance into existing processes. By incorporating the concept of a digital twin of a production machine, the interaction of measured and estimated as well as generated data by means of simulation, are shown. The digital twin could deliver results to retrofit data-driven prediction models, in order to improve the estimation of the RUL.
Werner et al. (Tue,) studied this question.