ARISE - Analytical Root-cause Identification in data Streams for detection of Emerging quality issues
| Reference number | |
| Coordinator | Volvo Technology AB - Advanced Technology & Research |
| Funding from Vinnova | SEK 6 000 000 |
| Project duration | September 2016 - June 2019 |
| Status | Completed |
| End-of-project report | 2016-02543.pdf (pdf, 4326 kB) |
Purpose and goal
Product quality is a top priority for a modern heavy vehicle manufacturer. One way to achieve higher quality is to identify quality problems more quickly by combining the vehicle data with already available knowledge such as warranty cases, technical specifications and technical experts. The ARISE project exploited the existing data for the early detection of arising quality problems in vehicles already on the market. In essence, it developed algorithms and models to detect quality problems and visualise them as support tools for experts in quality and customer satisfaction.
Expected results and effects
This project shows and quantifies the business benefit with Big Data in quality analysis. Quality issues can be detected earlier by identifying the pattern and trend changes and taking advantage of logged vehicle data, diagnostic trouble codes and warranty claims. ARISE developed incremental algorithms for detection of the warranty claim ratios, which can be used to improve early detection of quality issues. ARISE also developed algorithms for analyzing the resolution of quality journals. The results are disseminated as presentations, software and publications.
Planned approach and implementation
ARISE developed ML methods to find useful patterns in warranty-related data. Various regression and classification approaches are used for the early detection of quality issues. The efficiency gains of the approaches are analysed and corrective actions are planned. A software for exploration of the quality journals was developed which uses periodical data to provide the current status of warranty operations. E.g. quality journal length and effectiveness can be used to monitor on-going journals. Finally, some of the approaches are integrated into Volvo existing analysis tools.