Digital twin for predictive maintenance in production - Physics-based modelling and simulation to optimize data-driven models Digitaler zwilling zur vorausschauenden instandhaltung in der produktion - Physikbasierte modellierung und simulation zur optimierung datengetriebener modelle

The analysis of production systems and products during operation or usage is a key competitive factor for companies. Digital twins of products and production systems are gaining in importance here since, in addition to the analysis of operating data to estimate remaining useful life, they identify potential failure mechanisms, making it easier to locate the causes of failure. This paper outlines a concept of digital twins for predictive maintenance. The concept serves as a basis for an approach for the development of physics-based and data-driven models and their harmonization. In particular, the generation of data using a physics-based simulation model for the optimization of a data-driven model is highlighted. The concept was realized in a concrete use case for the prediction of filter clogging of a thermal regulator. The procedure is intended to support manufacturing companies in the implementation of predictive maintenance in order to optimize not only the lifetime of machines but also the next product generation.