Silicon, in the form of a chip shortage, has dominated the news for the past 3-4 years. It’s unsurprising given the role of computers in our lives. 

But even though the modern world revolves around computers, they’ve only been around since the mid 70’s. Steel has been around much longer: nine small beads made of meteoritic iron were found in Egypt and date back to 3200 BC.  

Steel forms much of the foundation of modern life. Literally. Roads use steel reinforcing rods for stability and strength, while skyscrapers are held up with steel frames. Steelmakers are always looking for ways to produce more steel. 

Computer Vision’ Effect on Inclusion Analysis

Part of the process of steelmaking is controlling inclusions, or nonmetallic compounds. Too many can affect the steelmaking process itself like clogging nozzles. Too little, and you can get “clean steel” which is harder to machine into useful shapes like rolls. 

Analyzing inclusions is a routine part of process and quality control. A trained metallurgist, through specialized equipment like an industrial panel PC, looks for inclusions in a piece of polished steel. Their sizes and chemical makeup are also examined. The entire process is both time- and labor-intensive. 

Researchers are turning to computer vision (CV) to speed up the inclusion analysis process. Simply, CVs work by examining an image, then interpreting and processing it via software algorithms. 

Results have been promising. First, CV has been able to figure out what’s an inclusion versus things like dust, holes, scratches, and other with 98 percent accuracy. Contrast this to a metallurgist, who is accurate around 72 percent of the time.

Second, the CV was 14 times faster in making the determination. Third, it did so through using one of the specialized equipment. A metallurgist, on the other hand, had to use several. The outcome from using CV for inclusion analysis is less equipment use and labor costs without sacrificing quality. (Side note: Computer vision is sometimes mistaken for machine vision. Machine vision in manufacturing must use a camera to view an object for its computer algorithms to process it. CV does not: the image can simply be uploaded into its rugged mini PC.)

Solving Complex Processes through Quantum Computing

Today’s steelmaking can be divided into three processes: primary, secondary, and tertiary. Primary steelmaking smelts iron ore and scrap (recycled steel) into new steel. Secondary processes add or remove elements like other metals and dissolved gasses to create requested alloys. Finally, tertiary casts these finished products into sheets, rolls, rails, bars, pipes, tubes, and wheels. 

Each process uses multiple techniques which can affect steel quality. And everything is running under strict deadlines. 

Steelmakers turned to computers to enhance these processes while cutting costs at the same time. They quickly discovered even the most powerful ones money could buy couldn’t keep up.

So some, like Nippon Steel, have turned to quantum computers. These are machines that use quantum mechanics called qubits to perform their calculations. They are millions to even septillions times faster than today’s most advanced supercomputers.

Nippon Steel teamed up with Cambridge Quantum Computing and Honeywell to see how such a computer could improve supply chain efficiency. Nippon and Cambridge did so by first creating a formula of a typical problem found in steel making. It then ran it through Honeywell’s quantum computer to optimize. According to company representatives, it didn’t take long for the computer to find a solution.

The team spent over a year in testing with the computer and new algorithms. Afterwards they announced that while it’s still too early to use quantum computers in steel manufacturing regularly, the results look “promising.” In fact, they continued, the original formula for optimizing can be used in many industries like transportation. 

Bringing Artificial Intelligence One Plant at a Time

As discussed above, steel manufacturing is an extremely complex process. Another issue is lack of recorded data. Many of the recipes for particular kinds of steel were developed through worker experience and trial-and-error. Plants that have been running for decades usually didn’t have sensors or other forms of data-capturing devices. And new equipment brought in to do so may not be compatible, be the wrong kind, or both. 

It was probably these and other issues that led to the Big River Steel Mill (BRS), the world’s first steel making plant built with artificial intelligence from the drawing board, according to Ron Ashburn, executive director of the Association for Iron & Steel Technology. Located in Osceola, Arkansas, BRS has nearly 50,000 sensors located throughout the new plant. These are used to collect data on plant function and equipment. The AI collects this data and analyzes it for wear and tear, maintenance requirements, and power usage just to name a few. 

Owners of the plant, which went into operations in 2017; and Noodle Analytics, which developed the AI; hope it’ll pave a way for other steelmakers when building out their steel mills. 

Closing Thoughts

Steel manufacturing forms the foundation for much of modern civilization. Many on-the-edge computer technologies like computer vision, quantum computing, and artificial intelligence, can be used in the trillion-dollar industry to both increase production and lower costs. Contact an expert at Cybernet if you’re interested in more details for your steel mill. Also follow Cybernet on Facebook, Twitter, and Linkedin to stay up to date on this and other relevant topics. 

AI industry industry PCs rugged industrial PCs workplace efficiency