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 | Ongoing |
| Venture | Strategic innovation programme for process industrial IT and automation – PiiA |
Purpose and goal
Effective production data analysis for improved process knowledge is a key factor for competitiveness and resource efficiency in manufacturing. Such data is typically large and complex, so manual analysis is virtually infeasible. AI algorithms have proven to capture dependencies in high-complexity data. Also, it benefit from large data. This project will develop AI-based machine models to better understand and control detailed dynamics of critical machines in the value chain. Such improvements has a large effect, and can save large amounts of production resources.
Expected results and effects
Production yield raises with less scrap. Production becomes faster and more reliable, the product quality improves, while delivery times becomes more consistent. For Outokumpu, the estimated process improvement will reduce rework worth 1.7 MSEK/yr. Outokumpu today uses 80-90% of recycled steel as input, but uses fossil gas (LPG) as the primary source of heating so CO2 emissions are reduced. For Ovako, the process improvement reduces scrap and rework worth about 2.0 MSEK/yr for this first line, which is blueprint for other lines so the effects will multiple.
Planned approach and implementation
Highly detailed production data from two steel producers is used for developing an AI-based model that can capture in-production dynamics of critical machines. A prototype based on this data-driven AI model is jointly examined to develop a deeper understanding of detailed dynamic machine behaviour. This prototype is applied in two settings: 1) context-aware prediction of machine preset values for a single critical machine and 2) machine-condition-based presets and prediction of machine and production line setup time.