BlueScope Steel is honing a series of machine learning models that are being put to work improving quality and reducing waste in its manufacturing operations.
The company said in its sustainability report [pdf] that the use of machine learning and artificial intelligence is occurring within the context of a broader internal push to drive up its digital capabilities.
It said current digital pilots also cover areas such as robotic process automation, the internet of things, and building information modelling.
“This year we have tested a range of opportunities to drive manufacturing efficiency, including digital simulations to help optimise operations and advanced analytics and robotic process automation (RPA) to reduce waste and improve manufacturing costs,” BlueScope said.
“These projects, and many others, are regularly shared through our Manufacturing Excellence Network and are now scaling globally.
“We continue to test value-adding opportunities for our supply chain and to support customer engagement.”
On its use of advanced analytics, BlueScope said one early use case is in “reducing defects and downgrades” to quality, which it said “can have [a] significant impact on delivering customer satisfaction whilst resulting in real savings from reduced quality claims and inefficient and wasteful rework.”
“In producing our next generation coated products, we identified metal spot marks as one key area we could improve, initially focused in our Australian manufacturing facilities,” BlueScope said.
“Using advanced analytics techniques including machine learning [and] advanced visualisation tools combined with investments in new surface inspection systems (SIS) and improved processes have reduced the quality claims and allowed significant savings to be generated.”
Another use case identified for machine learning is to “optimise” the amount of zinc coating “applied [to steel products] during the production process”.
BlueScope said its aim was to “minimise resource consumption and waste whilst maintaining our high quality promise to customers.”
“We are developing machine learning models to predict coating mass more accurately by applying longer term learning algorithms that automatically adapt with the latest data, resulting in significant metal coating savings,” it said.
“Leveraging and scaling the development of these advanced analytics tools and capabilities to the global network of coating lines present significant opportunities to further reduce our waste footprint whilst maintaining our service to customers and optimising our production assets.”