Hybridization of simulated annealing with genetic algorithm for cell formation problem

The grouping of parts and machines into manufacturing cells with the objective to improve grouping efficacy is the prime focus of cell formation problem. This paper presents the hybridization of simulated annealing (SA) with genetic algorithm (GA), where exploration power of GA, to broaden the search space, is combined with the intensification power of SA. A certain percentage of best solution in each generation of GA is sent, for further intensification and trajectory search, to SA. In intensification phase (with SA), Boltzmann probability distribution functions are used to allow a downhill move, with variable temperature. Starting temperature is chosen to be a small value with rapid cooling rate to keep search in a narrow region. If same solution is repeated for a certain number of generations, then shake-off is given to escape local maxima. During this shake-off, the SA parameters are relaxed to broaden the search for best neighbourhood solution. The effectiveness of proposed hybrid heuristic algorithm is evaluated through 35 benchmark problems from the literature. The GA with intensification through SA along with shake-off performed well in terms of solution quality, i.e. produced 24 best results with overall mean of 66.20. The overall comparative study shows that the proposed approach not only achieves the best solutions consistently, with minimum computational time, but also improves the results of two problems (problems 29 and 33), reporting it for the first time.