부산대학교 도서관

The design of inductors for non-isolated DC-DC converters aims to obtain a required differential inductance value to limit the current ripple, together with low power losses and reduced component size, desired for highly efficient and power-dense converters. However, small size and low losses are often contrasting objectives. In addition, the design solution feasibility must be evaluated by verifying saturation and thermal constraints. This multi-objective optimisation problem of the inductor design can be effectively tackled through population-based algorithms, such as Artificial Immune Systems. As these approaches require the evaluation of many designs through time-consuming procedures, a classifier system trained in advance to recognise non-admissible solutions can support the search for candidate solutions. The adoption of the Support Vector Classifier for the constraints handling of the inductor design problem is here presented and discussed.