The series: We look at the manufacturing industry which is using technology to create “smart factories” fit for long-term global competitiveness.
Angela Pappin, vice-president for technology at Dofasco, explains how molten metal goes through the process of becoming steel in the Hamilton plant. She says the advanced manufacturing supercluster could potentially deliver 'transformational' change in her industry.Glen Lowson
Steel-making is a science but still relies on experienced human operators at key stages of production.
But what if “smart” technology replaced manual tasks with digital sensors that consistently update information and reveal insights impossible to detect with the trained eye – or brain?
Answering the “what if” question is a top priority for Canadian steel giant ArcelorMittal Dofasco as it strives to compete globally. To speed the hunt for answers – when Canadian steel already is under siege from U.S. tariffs in a burgeoning global trade war – the Hamilton-based manufacturer hopes to join forces with innovative startups, university researchers and even a competitor, Essar Steel Algoma Inc., to promote a new generation of manufacturing.
Dofasco wants to improve its ladle metallurgy process, a key step in steel-making, while Algoma eyes the potential to automate product development. In both cases, the companies look to artificial intelligence and machine learning to help them set new, world-beating standards of efficiency, quality, energy savings and generate as-yet unimagined innovations.
A full ladle of liquid steel, seen here at the ArcelorMittal Dofasco plant in Hamilton, weighs about 318 tonnes, and can be heated to about 1,600 C. Dofasco wants to fully automate its ladle metallurgy process – the stage when trained operators manually add the 'secret sauce' to liquid steel before it is cast into numerous grades of steel slabs for construction, car-making and packaged goods. Instead, digital sensors could consistently determine the precise temperature at which to add the right ingredients to produce the desired grade of steel.Glen Lowson
The steel makers’ projects are the kind of game-changing breakthroughs imagined for a new $950-million fund announced this year by the federal government. The Innovation Superclusters Initiative, with federal dollars matched by industry, identified advanced manufacturing as one of five economic “superclusters” to spark collaboration among large and small Canadian businesses, postsecondary institutions and non-profit groups to solve industry-identified problems. Inspired by playbooks in Silicon Valley and Germany, Ottawa predicts that these new relationships will boost Canada’s lagging performance in innovation, expand the economy by $50-billion over the next decade and add 50,000 jobs.
“Usually our innovation policies have spent a lot of money supporting university and college researchers and it is usually up to the researcher to find their industry partner,” says Jayson Myers, chief executive officer of Next Generation Manufacturing Canada, a new network of private and public sector organizations created to reverse the trend of Canadian manufacturers trailing on innovation, global competitiveness and growth. “This [supercluster arrangement] turns it [the previous funding model] on its head: Now the money is going to industry.”
With government funding and project selection expected this fall, Mr. Myers hopes his organization will be able to disperse up to $1-billion over four years (up to $250-million from Ottawa and the remainder pledged by industry) to mobilize transformational projects of the kind proposed by Dofasco and Algoma.
The glow of molten metal is everywhere in the ArcelorMittal Dofasco plant in Hamilton.Glen Lowson
Dofasco, for example, is exploring the potential to fully automate its ladle metallurgy process – the stage when trained operators manually add the “secret sauce” to liquid steel before it is cast into numerous grades of steel slabs for construction, car-making and packaged goods. Assuming supercluster funding, the company would draw on the applied steel research by university professors and tap the expertise of startups already building artificial intelligence platforms to modernize manufacturing operations.
If successful, the collective effort would mean that Dofasco technicians no longer physically rake the surface of the liquid steel to remove imperfections and assess when to add alloys and chemicals for the exact grade of steel ordered by a customer. In future, it is hoped, digital sensors could determine exactly what inputs to add, and when, guaranteeing consistent quality every time and reducing energy required for the process.
“This would be the first [steel maker] in the world to have a fully automated ladle metallurgy facility,” says Angela Pappin, vice-president of technology for Dofasco, who hopes the project rolls out over the next four years. “Now I will know that when I made this vat [of liquid steel] six months ago it resulted in this type of performance at the customer’s plant. The machine will learn from that and I can feed [the information] back in.”
Ms. Pappin says the workers now performing the manual tasks would not lose their jobs, but would be deployed in different capacities in monitoring the metallurgy process. “We still need the workers, we still need the brains and we need to teach the system what the workers know and it never stops,” she says. “Steel just keeps advancing.”
Embracing the latest technology to generate consistent, high-quality data also holds true for Algoma, though its challenge – how to respond fast to ever-changing demands of customers – is different from the one at Dofasco.
“Traditionally, the research and development of a product in steel-making is a slow process,” says Pramod Shukla, chief operating officer of Algoma, in Sault Ste Marie, Ont., estimating his company needs an average of six months to test new steel products to customer specifications.
In future, he says, Algoma could apply self-learning algorithmic models to assess multiple variables at once – beyond the capacity of any experienced technician. As a result, the product development cycle could take just 15 days to two or three months.
An ArcelorMittal Dofasco steel worker monitors a ladle of molten metal as it makes its way through the Hamilton plant.Glen Lowson
“When you have a machine learning model, it will tell us exactly the best way to achieve our objective,” he says. “That is what smart manufacturing is all about.” He also sees a role for artificial intelligence in probing “the hidden capability in our [Algoma] assets,” further expanding the company’s ability to respond quickly to customer demand.
An important feature of the supercluster model is that companies like Dofasco and Algoma no longer have to go it alone on innovation. As a facilitator for supercluster projects, Next Generation Manufacturing Canada hopes to attract private and public sector players, big and small, (including some whom might not have met otherwise) to bring their collective expertise to bear in applying digital and other technological tools to modernize manufacturing. In the process, say supercluster advocates, small and medium-sized Canadian companies at the forefront of artificial intelligence and machine learning will be able to raise their global profile.
“We are not supporting manufacturers to do the same [as before] to acquire equipment and we are not supporting technology companies just to scale up,” says Next Generation’s Mr. Myers. “We are really trying to leverage the strengths of both manufacturing and technology.”
That approach wins praise from Humera Malik, chief executive officer and founder of Toronto-based startup Canvass Analytics Inc., already attracting global attention for its work in developing artificial intelligence platforms for industrial operations.
“The supercluster is allowing companies like Dofasco to create an ecosystem of partners and providers to help them in all areas,” she says. “They need a digital strategy, a data transformation strategy and they need folks like us, where AI now can be applied on top of all the data.” Her company, a member of the advanced manufacturing supercluster, is exploring a possible relationship with Dofasco.
Meanwhile researchers at McMaster University’s Steel Research Centre in Hamilton already have a relationship with Dofasco, creating a computer model that could predict outcomes for the ladle metallurgy furnace. Assuming supercluster funding, the McMaster mode could become a “digital twin” for Dofasco’s ladle metallurgy process, enabling sensors and robotic samplers to gather large amounts of data from the plant, learn from the “experience” and identify ways to ratchet up consistency in steel making.
“This is based on a concept generally known as Industry 4.0, a buzzword that refers to the idea that each of your processes would have a virtual computer twin that would communicate with the real process through information from the sensors,” says Ken Coley, director of McMaster’s Steel Research Centre in the department of materials science and engineering, and an ArcelorMittal Dofasco chair in ferrous metallurgy. “It would predict what happens next and the predictions and the real data would be stored in the cloud and be available for subsequent analysis … or even immediate analysis.”
Both Dofasco’s Ms. Pappin and Mr. Shulka, of Algoma, use the same word, “transformational,” to describe the potential of the advanced manufacturing supercluster.
Mr. Shukla says “the beauty of the supercluster concept is that it is bringing all of us together and giving us an opportunity to innovate together – and create a platform that can really accelerate the entire process of innovation.”