Data-Driven Constraint Handling in Multi-Objective Inductor Design

This paper analyses the multi-objective design of an inductor for a DC-DC buck converter. The core volume and total losses are two competing objectives, which should be minimised while satisfying constraints on required differential inductance profile maximum overheating. optimisation problem is solved by means population-based metaheuristic algorithm based Artificial Immune Systems (AIS). Despite its effectiveness in finding Pareto front, requires evaluation many candidate solutions before converging. In case problem, configuration time-consuming. fact, non-linear iterative technique (fixed point) needed to obtain configuration, as it may operate conditions partial saturation. However, configurations evaluated during do not comply with constraint, resulting expensive unnecessary calculations. Therefore, this proposes adoption data-driven surrogate model pre-selection phase optimisation. adopted classify newly generated compliant or constraint. Configurations classified unfeasible disregarded, thus avoiding computational burden their complete evaluation. Interesting results have been obtained, both terms avoided evaluations quality front found procedure.