As published in American Metal Market
August 10, 1999

Computer Scrap Analysis

Linear optimization designed for factual results

By J. EDWARD NORRIS
Ferrous Solutions, Inc.

In the past, steelmakers' scrap operations were run much like "mom's kitchen."

Detailed recipes were rare. Melt shop personnel used a pinch of this and a dash of that, all based on their experience and whatever was on hand.

Although the process has become more sophisticated today, many steelmakers have yet to fully capitalize on the enormous cost and quality control methods available through precise scrap management.

One way to achieve cost and quality benefits is through computer modeling. By using computer-modeling techniques, steelmakers can control and optimize the variability of their scrap mixes and achieve a verified return on investment.

What Computer Modeling Can Do to Reduce Cost

Reducing cost through computer modeling is not a new idea. In the past, computer modeling was either too expensive or required a great deal of technical support. All of that has changed. Today's scrap management software reduces bottom line operating costs several ways.

The savings achieved will be directly related to the alloy content of the customer's final product. The opportunity to save money on your scrap/alloy charge increases with alloy content of the product mix. These savings are based on a direct reduction in virgin alloy purchases. An additional reduction in base iron-unit cost occurs from proper scrap allocation. As alloy utilization from home scrap, as well as purchased alloy-bearing scraps such as solids, turnings, flashings, borings, etc., is maximized, alloy costs decrease.

Another opportunity to reduce cost is by direct substitution of premium grades scraps (busheling, bundles, punchings) with less expensive scrap grades. Direct scrap substitutions will be the major source of cost reductions for flat roll and carbon steel producers. This is due to lower alloy content of their customer's product. Computer modeling will also allow for a wider range of materials for melting. Premium grade scraps are typically used only when required to meet a customer's chemical specification or a process imposed constraint.

The melting process affects each scrap type differently based on that scrap's physical characteristics (size, shape, and density), placement in the charge, transformer size, and melt practices being used. The raw materials (scrap, alloy, and fluxes) vary within their own physical and chemical specifications. In spite of this material variation, proper use of linear optimization systems will lower overall raw material costs.

Linear optimization systems are being utilized in melting operations with great success. Many steel producers have maximized their opportunity to reduce cost and variation with linear optimization systems. Reductions in scrap and alloy costs will vary with raw material availability and product mix but overall significant sustainable savings are attainable with a properly tuned linear optimization system.

This is not the end of the computer modeling, cost reduction journey. It is just a starting point. Computer modeling may offer some environmental management opportunities like conversion of a process waste stream material to a process feedstock. Computer modeling may also aid the marketing and purchasing departments in making essential raw material/product pricing decisions based on the operation.

What Computer Modeling Can Do to Improve Quality

Reducing the number of diverted and/or scrapped heats caused by unexpected scrap variation is a serious quality concern faced by all steelmakers. Computer modeling offers several methodologies in which raw material and process variation is reduced.

Tracking scrap by individual supplier, instead of lumping all like material into a single category with a single chemistry is one method. The use of internal safety nets within the software is another. The use of statistical based aim chemistry selection is yet another method of reducing diverted heats. The latter speaks to the process capability of the scrap segregation program in use. Typically, all three methods are employed in the modern computer modeling software.

Additional benefits beyond reduced raw material costs and increased throughput can be realized with a properly tuned linear optimization system. Reducing process variation and standardizing charging practices is required to compete in today's markets.

Computer modeling is a significant management tool. By establishing standard practices, you ensure that all operators follow identical charging standards that have been tuned to your process. Repeatability of a process is an absolute requirement for heat to heat chemistry matching made necessary by sequence casting and tighter alloy grade specifications.

Your linear optimization system's long-term success depends on accurate chemistry and yield information of your raw material base. Some steelmakers incorporate a multiple linear regression program as part of their scrap program. Regression data along with a good sampling program can be invaluable in making the necessary adjustment to the scrap information file of your linear optimization system.

Choosing the Right Linear Optimization Model

Choosing the right linear optimization model for your shop may appear confusing but let's look at your options. There are several types of linear optimization packages on the market today.

At the low end are the spreadsheet models. These models typically run in Microsoft Excel and are PC-based. While the spreadsheet models do solve for a least cost scrap charge they lack the robust user interface and database functionality found in more expensive models. The spreadsheet solvers are not designed for twenty-four hour a day, seven days a week continuous operation and data gathering as required by today's steelmaking shops. One can expect to spend between $45,000 to $75,000 per model.

The second option is a PC-based linear optimization or scrap management system. This type of system provides much of the functionality one would expect to find within the scrap module of an integrated hierarchical shop-wide system. These models are typically written in a higher level language, (Visual Basic, C, etc.) and offer many features not found in simpler spreadsheet packages. A PC-based system can provide most, if not all, of the functionality from the scrap module of a hierarchical, shop wide control systems. What you won't get is the integration and shop-wide database features provided by such systems. One can expect to spend between $200,000 to $400,000.

The third option is a custom written shop wide integrated system. These systems are normally operated on a mainframe computer. In addition, such systems are typically customized to meet defined customer requirements. The added functionality comes at a price with such systems typically starting at $750,000.

Conclusions

How does one determine value? The correct answer is through fact-based analysis. When confronted with determining the relative value of a scrap material, anything other than linear optimization modeling is not the fact-based answer. Linear optimization software compares the costs of the recovered alloy content of every scrap in available inventory and then chooses the least cost charge based on material and operating constraints.

A matrix would typically include Carbon, Phosphorous, Sulfur, and Tin along with Chrome, Nickel, Moly, and Copper. The linear optimization software compares the yield-adjusted cost of each element and then determines the relative value of each scrap material in context with the desired molten chemistry and operating constraints.

People involved with scrap charge design are seeking to predict a future outcome using available information about an inventory of scrap and target chemistry. As the number of scraps in the available inventory increases so, do the difficulties in making an accurate prediction. All the while, the least cost solution has to be kept in mind. The number of calculations required to solve for a scrap charge quickly exceeds the abilities of most people trying to do this manually.

People are always amazed at the computer generated charge recommendations. Today's computer modeling software is capable of solving for hundreds of variables and constraints simultaneously in just seconds.

In the past, steelmakers' scrap operation were run much like "mom's kitchen." Today's tough business conditions leave no room for the "mom's kitchen" technology.

J. Edward Norris is president of Ferrous Solutions, Inc., a Canton, Ohio-based company that uses computer-modeling, sampling and quality control techniques to blend scrap mixes to customers' specifications.