Adaptive model and threshold algorithm for hot rolling scheduling

Hot rolling batch scheduling problems are NP-hard and have a large number of multi-criteria constraints that do not allow to develop a feasible solution. The goal of this research is to generate plans with minor technological violations quickly and efficiently and to avoid any serious violations. A standardized method of transforming the problem with technological constraints into a constrained optimization problem and a heuristic threshold algorithm are proposed. The algorithm threshold system is determined by penalty constants. An equivalence relation is introduced for threshold systems. The threshold algorithm generates the same plan for any two equivalent threshold systems. An effective algorithm for automatic selection of penalty constants based on real data is also proposed. The model was tested at plate rolling shops of the Magnitogorsk Iron and Steel Works with the purpose of scheduling manufacture, storage and shipment of flat rolled products.