COMPAS aims to research and develop software tools that enable both the efficient collaborative development of high-tech systems across the value chain using innovative compact modelling (CM) techniques, and the provision of AI-based ‘predictive health management’ for these systems using ‘compact digital twins’ (CDT). These are capable of capturing non-linear, transient and coupled (i.e. simultaneously dependent on multiple physical fields) situations, and of making decisions autonomously and ultimately in real time. They will require so few computing resources that they can be integrated directly into high-tech systems such as motor control in machines in fully automated factories, smart infrastructure (street lighting, power grids, etc.) or autonomous vehicles. In all these (and many other) application scenarios, the operational behaviour of the systems is also influenced by several factors that go beyond those covered by the deterministic CM. Therefore, the CDTs are supplemented by data-driven AI algorithms. These minimise the risk of decision errors due to incomplete coverage of important effects.
This very high level of complexity requires several groundbreaking innovations in reducing the model order (MOR) for CMs, in AI techniques for CDTs, and in programming. COMPAS will develop these using the example of the thermo-mechanical reliability of high-tech systems. This is as complex as all the scenarios mentioned above. The consortium has comprehensive test results and full knowledge of the relevant mechanisms. The project work can therefore focus from day one on the non-linear MOR for CM and – combined with the AI routines – CDT, with only minimal time required for the training and validation data.
