Luc van Nerom, Innovation Manager at PSI Metals & Managing Director at PSI Metals Belgium
Luc Van Nerom studied mathematics and computer science and started as a researcher at the AI Lab of the Vrije Universiteit Brussel. In 1986 he created a spin-off company “Artificial Intelligence Systems” brining AI and optimization technology to the metals industry. After a merger, the products of this company have been embedded in the production management software of PSI Metals. Today, Luc is focusing on product innovation management and industrial intelligence at PSI and PSI Metals.
Industrial AI, beyond the stars or down to earth?
Since the very beginning of AI, industrial applications were driving and inspiring research. The first expert systems were based on an extensive knowledge model mimicking domain experts’ reasoning and resulted in very successful industrial applications, also in the steel industry.
The internet revolution at the beginning of this millennium, cheap data storage and fast processors along with the sponsorship by the internet technology giants, made data driven AI a reality. This led to a revival of AI, the focus switching to data-centric technology vs model driven applications. Today, neural networks and machine learning (ML) are embedded into our daily lives, and are used in navigation systems, e-commerce applications, smart cars, social media …
With Industry 4.0 initiatives around the world, also the steel industry has been looking intensively at data-centric AI technology. We see many successful use cases and potential of ML in the metals industry: quality prediction, quality prescription, defect detection, maintenance prediction, forecasting of energy demand or replenishment or order delivery dates… just to name a few.
Unfortunately besides the successes, we also see up to 70% of failures of applying this data driven AI technology in an industrial context. These model accuracy failures are mostly related to lack of data or data consistency, lack of prediction power in the data, and lack of business and process understanding by the (IT and data focused) project team. Key for a successful industrial AI project is that both production experts and data scientists are collaborating closely. Advances in explanatory AI, model interpretability and easy-to-use AutoML will be able to close the gap between production experts and data scientists. So the way forward is that ML becomes a commodity like Excel became a commodity for reporting and calculation technology.