Root cause analysis by machine learning

Important results from the project

The project was a collaboration between industry and academia in the field of system diagnostics that will be of great importance for improved root fault diagnosis for vehicle systems. The aim was to find methods for root fault diagnosis for vehicle systems by means of machine learning between complex vehicle systems (on-board) and underlying calculation systems (back-end). This aim is considered fulfilled. The project will also increase research capacity and international competitiveness in Sweden. This is also considered fulfilled.

Expected long term effects

The result of the project should increase the ability to diagnose systems by also including error information from surrounding systems and using multivariate techniques to model error patterns and then classify these patterns when the cause is known (supervised machine learning). When a fault pattern and its cause (s) are known, the data model must be used at the time of maintenance to be able to directly point out the correct part when a new vehicle shows a known pattern. These results are considered to be fulfilled

Approach and implementation

For this next generation diagnostic system, the idea was that an industrial doctoral student, connected to Chalmers, would be responsible for the method development and Volvo Cars would integrate it into his IT system. Unfortunately, the doctoral student left after one year, but the project continued, but now with Linköpings Universitet as his new partner. That collaboration worked well as LiU has extensive experience in the field and has delivered outstanding results and published these.