Cavendish Lecture Series: Scaling AI Across Multiple Steel Sites: From Pilot to Practice
Scaling AI Across Multiple Steel Sites: From Pilot to Practice
Please login or create your My IOM3 Account to access this on-demand webinar.
We heard from
Qian Xiang, Technical Specialist – Enterprise Data & AI Architecture at Baoshan Iron & Steel Co., Ltd.
What to expect
Steel is not short of AI pilots. It is short of AI that survives contact with a second location. At Baoshan Iron and Steel, models now run blast furnace operations predicting hot metal temperature and silicon content two hours ahead and issuing the control adjustments themselves alongside surface inspection and virtual chief operator tools. Since the furnace model came into service, fuel rate has fallen by 2 kg per tonne of hot metal and carbon emissions by roughly 5 kg per tonne. The harder achievement is not any one of those: it is that a model developed at one location can be moved to the next and trusted there, rather than rebuilt from scratch.
Getting from isolated successes to a group-wide capability has depended less on model architecture than on what sits underneath it: describing the same piece of equipment and the same process step the same way at every plant; connecting the control systems that run the line to the business systems that plan the work; shared computing on site, with agreed rules on who may use which data; and a catalogue that records what data exists and who is answerable for it, so that governance is an operational discipline rather than a policy document. This talk describes that foundation and how it was specified, what it targets, and what it takes to move data across domains and across locations without losing trust in it.
The talk will share what this work has brought Baosteel, and an honest account of the fundamentals: data quality, governance coverage, and organisational talent, that still set the ceiling on industrial AI in steel
Speaker Biography
Qian Xiang is a Technical Specialist in Enterprise Data & AI Architecture at Baoshan Iron & Steel (Baosteel), where he leads work on the enterprise data foundation. He heads the company's Knowledge Graph Working Group and authored the Enterprise Knowledge Graph Foundation — Access and Topology Protocol Standard. His current work spans IT/OT data governance, a unified equipment data model, edge data platforms and the enterprise data asset catalogue, alongside the build-out and proof-of-concept of a trusted data space covering connectors, privacy-preserving computation and data sandboxes. He is also responsible for upgrades to the big data centre and management platform. His interest is in the unglamorous layer that determines whether industrial AI scales: data that is governed, portable and trusted.
This webinar was organised by the IOM3 Iron & Steel Group.
Previous webinars in this series
