A method for determining design parameters of a silicon carbide switch

Abstract

The application provides a design parameter determination method of a silicon carbide switch, comprising the following steps: firstly, collecting and processing multi-source data related to the silicon carbide switch, extracting efficiency, stress and reliability evaluation elements, establishing a design parameter space and generating a candidate design parameter set; then, identifying manufacturing deviation factors and constructing a manufacturing deviation description model to form manufacturing deviation constraint conditions corresponding to the candidate design parameter set; further, establishing a robust evaluation model to evaluate the robust performance of the candidate design parameter and form an evaluation record, and updating the evaluation credibility information; on this basis, performing robust multi-objective solving to obtain an initial robust parameter window; subsequently, constructing an active sampling iteration environment, training a sampling decision strategy by using reinforcement learning, and iteratively triggering simulation evaluation and actual measurement evaluation to update the model and gradually shrink the parameter window; when the convergence condition is met, outputting the final robust parameter window as the design parameter determination result.

Images