The research group Condition Monitoring investigates how slowly developing damages or creeping degradation processes in safety-critical system components can be detected, assessed, and predicted. To enable reliable forecasts of the remaining useful life and its probability distribution at any given time, historical usage data is translated into a probable damage state. At the same time, current measurement data is used to estimate the present health status. A weighted fusion of both complementary approaches is carried out based on their respective reliability.
The prediction of the health state up to failure is performed using statistical degradation models. Training data for the algorithms employed in the individual subprocesses are generated through lifetime testing and supplemented by model-based data synthesis, ensuring sufficient information is available even for the training of complex AI (Aritifical Intelligence) methods.
