CODA
Project leader:
- Tony Lindgren, Senior Lecturer, DSV, SU
Researchers
- Ram Gurung, PhD student, DSV, SU
Partners
- Scania Vehicle Service Information
- Linköping University, Department of Computer Science and Engineering (ISY), department vehicle
Project period: 2017-11-01 to 2020-10-31
Funding: Vinnova and Scania
Budget: Approx 18.2M SEK
Description
Data-driven methods based on machine learning have shown their strength in many applications. In a previous research project, IRIS funded by FFI and Scania, it has been demonstrated how data-driven methods could be used to plan optimal maintenance. The project has highlighted two characteristics that are important and need further research, the ability to handle concept drift in data and the possibilities for interpretability. Concept drift arises when the fleet of products that the models have been created to predict the health of changes over time. This occurs for example naturally in the automotive industry when new modules are available for sale or the launch of updated features. Strongly linked to concept drift is interpretability. In order for engineers to understand if their design changes affect the models, it would be a strength if they could understand how different variables and properties affect the models’ outcomes. The project will strengthen the field of research in machine learning and prognostics, and especially in the field of concept drift and interpretability of models.
The main work packages of the project are:
- Data drift and dynamic data
Project manager: Erik Frisk, Linköping University. - Interpretability of models
Project Manager: Tony Lindgren, DSV. - Deployment of data driven models
Project Manager: Mona Elofsson, Scania. - Effective data collection
Project leader: Karin Avatare, Scania. - Data warehouse
Project leader: Henrik Brandin, Scania.
The main applicant for the project is Scania CV AB (project leader: Jonas Biteus) and the other parties are Linköping University, Department of Systems Engineering, Stockholm University, Department of Computer Science, and Royal Institute of Technology, School of Information and Communication Technology.
People
