A parameter set design procedure for the simulated annealing algorithm under the computational time constraint

Abstract

In this research, a design procedure is developed to specify the parameter set for the simulated annealing algorithm under the computational time constraint. The proposed procedure is based on the response surface methodology (RSM). At first, the response is improved by employing the first-order response surface model and steepest descent method. After meeting the constraining surface, the steepest descent path must be modified subject to the constraint. Once further improvement is impossible, a second-order response surface model is fitted to determine the optimal settings of the parameter set. Finally, the optimal settings are adjusted subject to the constraints. The adjusted settings are used as the best parameter set for the simulated annealing algorithm. A job shop scheduling problem solved by the proposed procedure developed above is used to illustrate the design procedure. Scope and purpose The simulated annealing algorithm is a local search algorithm for obtaining good solutions to difficult combinatorial optimization problems. To get a good solution, the implementation of the simulated annealing algorithm must specify the parameter set for the cooling schedule. This paper proposes a six-step design procedure to determine the parameter set. The proposed procedure takes the interaction effects among parameters and the restriction of computational time into account.