RCA-ML - Root Cause Analysis of Quality Deviations in Manufacturing Using Machine Learning
|Coordinator||STIFTELSEN FRAUNHOFER-CHALMERS CENTRUM FÖR INDUSTRIMATEMATIK - Fraunhofer Chalmers Centre|
|Funding from Vinnova||SEK 4 000 000|
|Project duration||December 2016 - March 2019|
Purpose and goal
The main objective of the project was to develop methods based on machine learning to identify the root causes of quality deviations in manufacturing.
Expected results and effects
A framework for identifying root causes of quality deviations was developed early in the project. Due to difficulties in gaining access to well-annotated data regarding actual causes of documented deviations, focus in the latter part of the project was placed on subproblems in the manufacturing process. An algorithm has been put into production that detects when the quality after a particular machine deviates and then tries to regulate to counteract the error. Two scientific articles on general methods and a demonstrator software have been produced during the project.
Planned approach and implementation
The project was divided into four work packages, with the common goal to develop and implement methods for root cause analysis. Two work packages were focused on general method development for root cause analysis and were implemented mainly by the FCC and Chalmers. The other two work packages concerned partner specific use cases for SKF and Flexlink. At some occasions the whole project team has met, but the majority of project collaboration has taken place in smaller groups. During part of the project, SKF has had a person co-located one day per week at the FCC premises.