Simulated Annealing with Restarts for the Optimization of Electromagnetic Devices

The simulated annealing (SA) algorithm is often used in the optimization of electromagnetic devices, but in multi-modal continuous problems it may be affected by premature convergence. This paper proposes an improvement to prevent the algorithm from becoming trapped in local minima, called SA with restarts. This algorithm allows an enhancement of performance and a reduction in computing effort, appreciated especially in the optimization of electromagnetic devices, where the single evaluation of an objective function calls for the numerical solution of an electromagnetic field problem.