Automated optimal design techniques for inverse electromagnetic problems

Two methods are considered for the solution of automated digital optimal design problems. The computer procedures based on these methods have been implemented and tested. The first method, based on a deterministic approach, considers a quadratic approximation of the cost function. The second, based on a stochastic approach, is derived from the simulated annealing algorithm. Both methods, implemented as computer codes, have been applied to the solution of a test synthesis problem where the magnetic field is generated by discrete coils. The deterministic method is substantially faster, especially when the calculation of the cost function is time consuming. On the other hand, the stochastic method gives good approximation of the global minimum independently of the initial conditions: as far as CPU time is concerned, the method is more expensive, but can be profitably used when the cost function can be calculated quickly and the number of design variables is large.<>