An Application of Mixed Integer Programming in a Swedish Steel Mill

The object is to make a tool for a Swedish steel company Fagersta AB to select appropriate ingredients for charges of high-grade steel, specified with regard to weight and percentage of certain chemical elements. The problem is modelled as a traditional blending problem, with additional constraints expressing that some of the variables must be integer valued. The algorithm used was designed at the Contract Research Group for Applied Mathematics, Royal Institute of Technology in Stockholm. The implemented system is available from the time-sharing terminal of the company and has completely replaced the former manual calculations. About two runs are made every day and the results are directly used in the production process. A substantial improvement in profitability has been demonstrated. The cost for raw materials has been decreased by roughly 6%, which corresponds to $200,000 per year, and the planning work has been considerably simplified.