Modelling the slab stack shuffling problem in developing steel rolling schedules and its solution using improved Parallel Genetic Algorithms

Abstract An improved Parallel Genetic Algorithm (iPGA) is proposed to resolve the complexities of the slab stack shuffling problem of the rolling mill. Two new operators namely the modified crossover operator and the kin selection operator have been proposed. These operators not only make the resulting iPGA more efficient (in terms of exploration as well as exploitation of various schemata) but also act as an insurance agent against the loss of certain genes, which may turn out to be useful in later stages of evolution as well as against premature convergence. Genetic codes and operators are specially designed to ensure the solution feasibility as well as to speed up the solution convergence. Exhaustive experimentation carried out on 512 randomly generated test problems show that the proposed algorithm offers an improvement of 6% over the conventional GA-based optimization algorithm. Application of test run on real production data of the rolling mill gave results consistent with those obtained from randomly generated set of representative test problems.