US Steel aims to build a high-tech future in both old and new factories

When U.S. Steel acquired full ownership of Big River last year, it also acquired the factory’s artificial intelligence know-how, signaling the commitment of the 120-year-old manufacturing giant to advance the factory’s technology. did. But according to the company’s chief information officer, bringing the type of technology used at Big River to other steel mills, some of which are over 100 years old, is proving to be a daunting task. Proven.

Competitors like Nucor Ltd.

According to KeyBanc Capital Markets analyst Philip Gibbs, one of the reasons US Steel has an edge over US Steel is, like Big River, a new plant with cost-effective electric arc furnaces and new technology. because it mainly operates U.S. Steel has struggled to turn a profit in recent years, he said.

According to Gibbs, when U.S. Steel announced its acquisition of Big River, some saw it as a Hail Mary pass to regain its footing. He said the strategy worked well in the short term as steel demand surged temporarily after the pandemic, but he added that it remains to be seen how it will play out in the longer term as markets tighten. .

US Steel said the Big River acquisition was already profitable, with record first and second quarter earnings this year before weaker demand hampered the third quarter. Told.

Big River’s contract hasn’t revolutionized the way U.S. Steel works at its older facility, according to Kevin Burns, process excellence engineer at the company’s Gary Works facility in Gary, Indiana. It inspired me on what is possible.

According to Cody Hore, a third-generation steelworker and one of the factory’s first employees, some of the main ideas when Big River was built involved advanced technology. I was. “The next thing I knew, US Steel was knocking on my door,” he said. “They wanted to get their foot in the door with up-and-coming technology.”

“When someone first suggested that we automate one of our cranes, I was like, ‘No, it’s never going to happen,’” says Burns. “Then you go there and you realize, ‘Oh, it’s not impossible.’ You can do it. You can do it safely. Oh, that’s a really good idea. Maybe I we should do that.

Once the largest steel mill in North America, the 110-year-old Gary Works on the shores of Lake Michigan contains some equipment originally commissioned in the 1950s and 1960s, but has been updated over the years. I came.

Gary Works is still one of the largest steel mills in the United States, employing 4,000 people and producing 7.5 million tons of crude steel annually. Big River has an annual crude steel production capacity of 3.3 million tons and employs approximately 750 people.

Big River uses advanced technology to make basic steel mill functions, such as cooling hot steel coils, more efficient. If the coils are too close together, they will take longer to cool, which is why Big River’s machine learning automated cranes are so important. During the summer, temperatures in parts of the factory can reach 150 degrees, so maintaining cooling can be a challenge. Big River Steel recently installed a slushy machine to help employees cool off.

Big River also uses cameras to feed input into machine learning algorithms to detect defects in coil slabs and to determine if someone has gotten too close to a particular machine and caused a safety hazard. increase.

Big River built some of its algorithms in-house and implemented some through third parties. But according to Christian Holliday, senior director of his Digital Studio and Big River Steel Integration at US Steel, the algorithms deployed in Big River are not plug-and-play at other factories. Each factory typically needs to build and train its own model based on its own environment. Older factories may face more challenges because they don’t always produce the data they need, he said.

According to US Steel Chief Information Officer Steve Bugajski, transforming old factories means installing data-generating devices such as sensors and cameras into existing equipment.

According to Burns, it can be difficult to find the best places to place sensors on equipment because the equipment needs to be protected from steam and heat. Newer equipment is often equipped with built-in sensors that are safely placed inside the machine, which can’t be done during the retrofit of older equipment, Barnes said.

Additionally, sensors designed to detect vibrations may need to be more sensitive, Barnes said, because older equipment tends to be heavier and bulkier and less prone to vibration. On average, Burns says, he may need to perform a sensor installation once a year during a planned equipment outage.

Even older factories like Gary Works may lack wireless network capacity to connect sensors, Bujasky said. Environments like steel mills full of concrete and steel are notoriously bad for bandwidth. Newer and more advanced factories can also struggle with network capacity, but connectivity planning is often considered at the design stage, according to Forrester analyst Paul Miller.

The introduction of new technologies also creates challenges in teaching employees new skills and helping them adapt to change.

US Steel recently began offering digital training to its non-IT employees, such as machine operators who spend time on the factory floor. Holliday said U.S. Steel is on track to meet its goal of training 100 employees as “digital agents” by the end of 2022.

“Not everyone is open to change, but we try to work with people who are open to change,” he said.

Gary Works’ Burns said when he got into the steel industry 22 years ago, he made decisions based on who was the loudest in the room.

“Now that we have all this data, instead of relying on someone who has lived 30 or 40 years and what he does, You can look it up and see what happened and make the right decisions.I remembered what happened,” Barnes said.

Despite the challenges, Gary Works employs machine learning algorithms designed to make operations more efficient.

According to Process Innovation Engineer Todd Hardesty, an area called the hot strip mill, where steel slabs are converted into coils, can become a production bottleneck. Machine learning systems can now analyze why a particular slab took longer than it should. Problems such as slow human operators and faulty equipment can be addressed, he said. According to Hardesty, factories are expanding their time-keeping systems to other bottlenecks.

Gary Works is also working on creating a digital twin, a live virtual representation of what the primary device is doing at any given time, according to U.S. application architect Veeraiah Katta, the ultimate goal is done. It’s about predicting time and optimizing output. steel.

US Steel hopes to speed up the creation of such analytical models by creating a single repository for all data. This project, the so-called data lake, has been underway since 2018 and will ultimately help factories build complex models faster.

Bugajski, CIO of US Steel, said that despite the challenges, US Steel continues its modernization efforts.

The Big River acquisition does not provide a one-size-fits-all template for modernizing US Steel’s old factory. But Burns of Gary Works said it helped move the company in the right direction.

“It really takes you from that kind of ‘I don’t know’ to ‘Wow, they do that all the time,'” he said.

Write to Isabelle Bousquette at isabelle.bousquette@wsj.com