As one example, he cites improved grade intermix control at the No. 2 caster of U.S. Steel's Gary, Indiana plant: "One day our Gary works called. They wanted us to investigate if there was a way to predict the chemical composition of slabs produced during periods of steel grade change." (During a grade change, steel at the end of one heat mixes with steel of different chemistry from the next heat.) At Gary, because of the diversity of steel chemistries required by customer orders, many slabs produced during grade change had to be stored until they could be tested. By the time they were tested, even if the chemistry was good enough to meet customer specs, they lost their place in the processing line and ended up clogging inventory.
A U.S. Steel team devised a computational model based in a mixture of physical modeling, plant experiments and Y-MP computations. "We came up with a rather simplified model," says Vassilicos, "that can be used in a process computer, real time, to predict how the chemistry in the cast product changes as a function of time." Plant and research personnel implemented the model on the Gary No. 2 caster. It has proven highly reliable. Several months of trial run showed that the model predicted chemistry as accurately as laboratory testing of slab samples and it wasn't susceptible to errors inherent in the testing procedure.
With the new method, the Gary plant has significantly reduced the need to test for slab chemistry. "We have the confidence to supply our customers with quality steel grades on the basis of the model," says Vassilicos. "This is a cost savings because you save on the handling, you save on testing the chemistry, and you reduce the inventory."
go back to the main screen