The iron and steel industry needs to be streamlined and the amount of scrap reduced to increase returns. What is required is a faster and more reliable production with stricter tolerances – then better products are created that reach the customer faster. This project, INSITE-X, will develop an AI-based prototype for detailed simulation of dynamic production-behaviour of critical machines, to increase process understanding and control.
There are two industry partners in the project, OVAKO and Outokumpu, that both come from the same industry have its advantages:
- It will develop data collection, data sharing, and analysis research for value chains.
- Joint process challenges will be approached with more powerful analysis techniques, while building proficiency in applying and creating value from these techniques.
Enables deeper process understanding
Process improvements for resource efficiency in steel manufacturing requires analysis of large amounts of data. Having effective analysis tools becomes a key factor for competitiveness. Advanced analysis of detailed production data enables deeper process knowledge, but data is complex in size and heterogeneity, which makes manual analysis virtually incomprehensible. AI algorithms can capture high complexity and benefit from large data sets.
Discover unknown dependencies
In projects Dataflow (with Outokumpu) and Swedish Metal (with SSAB/Sandvik SMT), such algorithms were used to discover unknown complex interdependencies in production variables. A few critical machines were found to significantly influence the entire production chain.
Using AI to model the dynamics of such machines, the current uncertainty originating from unknown dynamics can be better controlled, and large amounts of production resources can be saved for the entire value chain.
Build AI prototype
Machine models are commonly based on the machine design, or on simulations grounded in physical machine properties, and they are typically implemented as digital twins. This is not detailed enough, since the behaviour of operational machines is constantly influenced by operational dynamics and degradations. Even highly detailed simulations are simplifications, still not capturing the full actual machine behaviour in production This project’s prototype will capture such detailed dynamics using AI and Deep Learning.
The effect of implementation
By implementing the AI prototype that will be developed in the project, Ovako and Outokumpu estimate that they will reduce the amount of scrap and rework to a value of approximately SEK 2.0 million per year / production line.
– We see this as a very important project that is completely in line with the digitalisation strategy we have within Ovako, says Marcus Svadling Ovako Sweden AB.
– In the steel industry, we are sitting on a goldmine of data that enables our process development. With the help of AI technology, we will be able to greatly increase our development rate and efficiency, says Joakim Ebervik, Outokumpu Stainless AB.