The steel industry produces 750 million tons of steel every year, most of it through a process called continuous casting. The metal is formed and solidified in large sheets, used to make a variety of everyday products like automobiles, soup cans, and household appliances.

In an industry with such diverse end-products in which safety is critical, quality control is an important issue. During the molding and solidification process, the intricate physics and geometry cause gas bubbles, trapped sand particles, and other imperfections in the steel sheets. The problems caused by such quality variations can be as harmless as surface defects, like uneven spots in the painted hood of a car. They can also be more serious structural defects, such as fatigue failure caused by trapped particles that affect safety in products that need stability. And some quality variations can even cause costly industrial accidents, which may mean molten metal breaking through the shell of solidifying steel and halting the casting process.

Due to the complexity of continuous casting and the range of imperfections that can occur, steel companies are researching ways to minimize problems caused by industrial processing. In the past, research meant gathering empirical data from casting experiments, possibly changing nozzle sizes or varying the amount of gas injected into the mold with the metal. Today, supercomputing supplements empirical data with numerical modeling of the process.

Brian G. Thomas, a professor of mechanical engineering at the University of Illinois at Urbana-Champaign, and a team of graduate students are working with NCSA's Origin2000 computing system to simulate the continuous casting process. They take new information forged from the models directly to the private sector, working with a group of steel-producing companies, called the Continuous Casting Consortium, to implement numerical results in real-world technology. Since its formation in 1991, the consortium has provided a way for the steel companies to share the expenses and results of computational research and a vehicle for the direct transfer of computational information to industrial implementation.

"Since forming the consortium as a cooperative research effort over a decade ago, we have been developing comprehensive models of continuous casting and using them to improve understanding, optimize the process, and solve practical problems for the steel industry," says Thomas.

Pierre Dauby, a former consortium member of LTV Steel and currently vice president of process technology at Danieli Rotelec, contends that the consortium, coupled with supercomputing resources, has been an effective collaboration. "In contrast with other academic and industrial research centers that exist in the United States, the Continuous Casting Consortium gathers only a small group of industry representatives, but all of them are truly experts in their fields. The result is extraordinary feedback and exchange of information and experience between the university and the member companies."

Access Online | Posted 11-19-2002