Dialog, UA Faculty and Staff News

UA Data Mining Team Develops Model to Improve Productivity in Aluminum Processing

Wise Alloys, in Muscle Shoals, manufactures aluminum coils.

by Bill Gerdes

The next time you open a can of Campbell's tomato soup or a cold cola, you may be the beneficiary of a new technology that UA's data mining program helped develop.

About a year ago, Wise Alloys, a Muscle Shoals company, and representatives from the UA department of information systems, statistics, and management science, applied a new technology to improve quality control at Wise Alloys. The company is the third leading U.S. producer of aluminum can stock for the beverage and food industries, and two of its largest customers are Ball Metal Corporation and Crown Cork and Seal, which, between them, produce a large number of the world's aluminum cans.

To make these cans and other products, the 4-million-square-foot Wise Alloys Plant casts molten aluminum into sheet ingot. The ingot is then rolled into coils of aluminum sheet approximately 1/100 of an inch thick. Wise rolls out 150 of these coils per day, each of which weigh between 25,000 and 30,000 pounds when shipped. However, not every roll comes out perfect. Being able to predict when an imperfection was due could improve the bottom line dramatically, and that is where UA came in.

Chip Flournoy, information technology manager at Wise, explained.

"We had a need to predict when we would have problems with quality issues," Flournoy said. "We run scrap aluminum into the plant, melt it and put alloy agents into it to make it the right chemistry. We then cast it into ingots. When we cast ingots, we cast six at a time, each weighing about 35,000 pounds. We then heat it back up and process it through rolling mills where it formed into a coil. Each time it goes through a mechanical or thermal event, metallurgical characteristics of the metal are affected that may be beneficial or detrimental to our customers.

"At any given time, there can be a large number of variables in the process and all are connected in some form or fashion with a data collector that puts the data into our datawarehouse," Flournoy said.

"What it comes down to is, we have a lot of information about the metal we process but have not made the best use of the information to identify what causes the good or the bad properties. That's where we brought in Mike and his team."

Under the tutelage of Drs. Michael Hardin and Michael Connerly and department head Dr. Edd Mansfield, three UA students spent last summer at Wise Alloys studying the process and using "data mining" to develop a process model that allows Wise to predict which coils will be defective.

Flournoy said Wise brought in the UA team for two primary reasons.

"We needed them to help us get a handle on what kinds of tools we could use to process the information we had (data mining) and use their expertise to perform different kinds of analyses, depending on what we were looking for," Flournoy said.

"If defective coils can be detected early, the aluminum can be re-melted and re-stretched, at less cost than if the coils had completed the process," Hardin said. "We developed a model to help detect early in the process which coils would be defective."

The process developed by the UA students turned out to be 92 percent correct, which translates into a huge cost savings for Wise. "To me it was a win-win situation," Hardin said. "Our students had the opportunity to study a real problem and provide a real solution. They could see how the theory learned in the classroom was applied in the real world. And the company benefited from the model."

A third party involved in the project was SAS, whose software is used in the data mining process.

"I was very pleased with the attitude of both the faculty and students." Flournoy said. "They were very serious about trying to help us. They saw the need and went right after it. As a matter of fact, the original project scope was extended after we saw what they could do."

Founded in 1941 to manufacture aluminum sheets for airplanes during World War II, the original 2,000-acre facility in Muscle Shoals was owned by Reynolds Aluminum until it was sold to Wise Alloys in March 1999. Its sheet products are used in the manufacture of beverage cans, ends, tabs, food containers and other products including trailer roof and fin stock. The facility has a current capacity for finished products of 950 million pounds per year.