Reducing the Design Space of Standard Electromagnetic Devices Using Bayesian Response Surfaces

This work proposes a procedure that reduces the volume of the design space of a standard electromagnetic device given a set of existing designs for the same type of device. This is done by building probability models of the device´s design variables using Bayesian response surfaces. Our procedure provides a design space that contains a satisfactory design with a probability of P, where P is specified by the designer to be between 0 and 1. This design space can then be searched with optimization algorithms.