Abstract

The paper presents two optimization methods based on the concept of design of experiments: the method by zooms and the method by slidings of designs. Two application variants are presented for each of them. One of the variants only requires access to the values of the objective function at different points in the feasible domain, grouped in designs of experiments, arranged successively, towards the most convenient values. The other variant uses the same technique, but the advance of the designs towards the optimal value is based on the information provided by second-order polynomial models that approximate the objective function on each design, which facilitates the convergence speed of the algorithm. The methods under discussion lend themselves very well to numerical simulations using the finite element method that can provide the values of the objective function at any point in the feasible domain that corresponds to a unique configuration of the simulated device. The paper presents a comparative study of the application of these methods in the two variants, on a 2-D numerical model of an electromagnetic device. Both in the case of the method by zooms and in the case of the method by slidings of designs, the results highlight the simplicity of the first application variant but also the high speed of convergence of the second one, especially for the method by zooms.

Full paper: click here