A Comparison of Crossover Operators in Genetic Algorithms for Steel Domain

This chapter deals with the hot rolling scheduling problem in steelmaking. In the factories, the slabs are first assigned to order and sent to the heating furnaces and rolling mills. After the slabs arrive at the rolling mills, they need to be modelled mathematically to minimize the attribute difference between rolling slab parameters such as width, thickness, hardness, tapping temperature, finishing temperature and coiling temperature. The problem is modelled by reducing it to the asymmetric Traveling Salesman Problem (TSP) and simulated with a genetic algorithm. 7 different crossover operators in the literature namely; Partially-Mapped Crossover (PMX), Cycle Crossover (CX), Order Crossover (OX1), Order Based Crossover (OX2), Position Based Crossover (POS), Maximal Preservative Crossover (MPX), Extended Partially-Mapped Crossover (EPMX) are analyzed both on the production data from the steel mill and 10 symmetric and 5 asymmetric TSPLIB data sets. The performance of the applied operators has been compared in terms of error rate and computation times. While the MPX method was more efficient than other operators in steel data sets, the PMX operator performed better in the TSPLIB data set. In terms of computation time, the PMX method has been the least time-consuming.