INSITE-X - AI-based analysis of machine dynamics
Reference number | |
Coordinator | Högskolan i Skövde - Institutionen för informationsteknologi |
Funding from Vinnova | SEK 3 248 407 |
Project duration | March 2021 - March 2024 |
Status | Completed |
Venture | Strategic innovation programme for process industrial IT and automation – PiiA |
Important results from the project
The project INSITE-X develops data-driven deep AI models to predict production quality for key operations in a production chain, from detailed production data measurement. Through a detailed production model that is based on data, the production output be predicted for alternative production parameters, which allows process developers to try alternative production settings and see the outcome for these. With deep AI models, complex relationships can be captured where input data and measurement during production are combined in a common AI model.
Expected long term effects
For strip rolling, there is now a deep AI model that can predict eight shape-parameters for resulting strips, with a 90% certainty, as well as a demonstrator where a user can test parameters and see the predicted output from this. One can try new settings not possible to test in production, which gives a deeper understanding of various dependencies between parameters. Less experienced operators can examine the process by trying different settings. The effects consist of better process understanding and process development, with less scrap and thus a more sustainable process.
Approach and implementation
The project has been developed in stages, using two rolling cases at two steel companies. The work has integrated the competences of all partners, with joint meetings every two weeks, where joint discussions have enabled project progress. This working method has given great competence and knowledge development among partners and is a prerequisite for now being able to take on new AI-related projects in process development. The working method has been highly appreciated by partners and will be a prerequisite in the upcoming project since it has proven to be an effective working method.