Project Details
Description
TUG will provide research work on diagnosis, predictive
maintenance, and lifetime estimation. In any case, obtaining
knowledge from available data, which includes real-world and
simulation data, is of importance. In this project, we will extend
previous work in diagnosis utilizing models substantially. Instead of considering hand-crafted models for diagnosis and the prediction of faults, we are considering learning models from available data. In contrast to existing work, where there is often a need for having data that corresponds to the correct and the faulty behaviour, we want to obtain a model solely from correct behaviour that can be used in a similar way for model-based diagnosis avoiding the use of known faulty behaviour. In this way, classical restrictions regarding the use of model learning and other types of machine learning can be avoided. In particular, we expect to use the learned model of the correct behaviour for fault detection and localization directly, not
needing information about the faulty behaviour represented in the data. In the project, TUG will provide the foundations behind the model-learning approach and also apply it to SC2, where the focus is on diagnostics during operation considering powertrains of vehicles.
The research will be carried out until reaching a technology readiness level of 4 comprising experimental proofs of the concepts using parts of the powertrain and technology to be validated in the lab.
Status | Active |
---|---|
Effective start/end date | 1/05/23 → 30/04/26 |
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.