DELPHI Diagnosis by ExpLoiting PHysical Insights in neural network models

Purpose and goal:

The purpose of this project is to improve automotive diagnosis as well as improving competence and collaboration between the industry and the academy within this area. The goal is to develop data driven models for fault diagnosis of dynamic systems and isolation of unknown faults by exploiting physical insights of the system. It is also a goal to investigate how this technology can be industrialised.

Expected results and effects:

After execution of the project it is expected that this technology has reached level 5 on the TRL-scale. It has then been investigated whether the hypothesis of being able to use physical insights in data driven models for diagnosis is possible or not. Practical problems such as handling of different vehicle configurations and collection of data are also investigation.

Approach and implementation:

The work is split up into work packages that are either led by the academic part or the industrial part. The work packages led by the academic part are for resolving the theoretical questions. The other work packages are for developing experiment platforms that can be used for case studies in tight collaboration between academy and industry.