Scheduling High-Grade Steelmaking

• Authors: Jürgen Dorn;Reza Shams
• Venue: IEEE Expert: Intelligent Systems and Their Applications
• Year: 1996

High-grade steelmaking plants produce a great variety of specialized grades of steel for use in the aircraft industry and other exacting applications. Lot sizes in such plants are typically small, and the chemical composition of each lot, or heat, is critical. Chemical infiltration in the plant's electric arc furnace between consecutive heats, for instance, can ruin an entire lot, making production-cycle scheduling quite tricky.Austria's Böhler Uddeholm, a leading European producer of high-grade steel, had earlier attempted to solve its scheduling problems with conventional software approaches. Finding a schedule, however, that observes all compatibility constraints between production runs can theoretically be an NP-hard problem. (Because sequences cannot be excluded a priori, schedulers would need to examine each permutation in a sequence of jobs. For 10 jobs, that would mean 3,628,800 comparisons. But Böhler has over 30 jobs to be scheduled.) These earlier attempts failed because they sometimes simply left difficult-to-schedule jobs unscheduled.In a joint project between the Alcatel Austria-ELIN Research Center (AERC) and the Christian Doppler Laboratory for Expert Systems, we developed an expert system that solves these scheduling problems by first determining the criticality of jobs and then applying a heuristic, or rule, based on the criticality values to guide the search for a feasible schedule. The system also provides local control of the use and relaxation of constraints, enabling engineers to make on-the-spot adjustments.Implemented in Pamela, a rule-based language developed by AERC, this system supports the steelmaking shop's technical staff in generating weekly steel heat schedules. It has operated successfully at Böhler's Kapfenberg plant since 1992, freeing engineers for their regular production supervision duties.