Similarity Search of Time Series Data: Evaluation of Search Engine in Industrial Process Data(SIFT)
| Reference number | |
| Coordinator | RISE Research Institutes of Sweden AB - RISE |
| Funding from Vinnova | SEK 3 701 322 |
| Project duration | November 2024 - April 2026 |
| Status | Ongoing |
| Venture | Advanced digitalization - Enabling technologies |
| Call | Test and demo of advanced digitization in a real environment |
Purpose and goal
Many manufacturing companies collect extensive production data, but improved methods for data management, analysis, and visualization are needed to maximize its value. The project aims to test and evaluate a technical solution and method in a real industrial environment to make large multivariate time series more searchable. Participating companies Nexans and Nord-Lock will use enhanced digitalization to improve productivity, reduce waste, and and improve quality.
Expected effects and result
The developed search engine enables searching in time series from numerous sensors and batches. By embedding time series and metadata into vectors stored in a vector database with similarity search functionality, anomalies and similarities can be detected. This simplifies data searches, typically challenging in large datasets. Expected outcomes include supporting industrial digitalization, reducing disruptions, increasing efficiency, enhancing quality, sustainability, and workplace conditions.
Planned approach and implementation
The 18-month project is led by RISE with Chalmers and builds on results from the DFusion project. The method will be adapted, tested, and evaluated in real manufacturing environments at Nexans in Grimsås (cable manufacturing) and Nord-Lock in Mattmar (lock washers). Work methods and contextual usage will also be assessed. Sensor manufacturer IFM will contribute, with their monitoring software, Moneo, serving as a platform for further developing the solution tested in the SIFT project.