Machine Learning-based Component Figures of Merit and Models for DC-DC Converter Design

Abstract

This paper is focused on a data-driven approach to capturing figures of merit and features of semiconductor switches and passive components used in switched-mode power converters. Extensive amounts of component data available on commercial distributor sites are gathered and processed to provide insights into relationships among component characteristics beyond what is commonly available in physics-based models. The data is used to train supervised regression machine learning (ML) models that can be used to predict component parameters. One practical use of these ML-based models is in an optimization tool that advises power converter designers on component selection to achieve an optimal specified objective function.