Integrated Quality Control System through Advanced sensing tools and AI for Continuous Casting

Important results from the project

The demonstration of real-time visualization of mould temperatures and online defect risk prediction by a machine learning model was realized. A surface imaging system with automatic defect detection was also demonstrated in an industrial environment. However, the underlying machine learning models needs further development in order to produce reliable results. The experience of implementing online AI models in an industrial process is useful beyond the continuous casting application.

Expected long term effects

While the prediction and detection models need further development to give reliable results, the demonstration of an on-line AI model connected to the process is a major achievement and the project has produced important learnings, useful also in other processes, regarding: - Strategies to handle heterogeneous data from different sources - Integration of an AI model in a system running online and in real time - Identification of limiting factors and important considerations

Approach and implementation

The overall progress was as follows: - Preparation and execution of industrial trials - Data processing - Model development - Software update and integration - Demonstration during industrial trials The work with data/models and preparations for trial 2 took longer than planned. One measuring equipment could not be used after the first round. The work with data preparation and with integration of the model for online use was more demanding than anticipated in the planning of the project.