Date of Award

5-1-2024

Language

English

Document Type

Master's Thesis

Degree Name

Master of Science (MS)

College/School/Department

Department of Electrical and Computer Engineering

Dissertation/Thesis Chair

Mohammed Agamy

Committee Members

Hany Elgala, Nathan Dahlin

Keywords

Condition Monitoring, Digital Twin, Neural Networks, Parameter Identification, Power Electronics, Resonant Converter

Subject Categories

Electrical and Electronics

Abstract

This paper introduces a Neural Network based approach for creating a Digital Twin of a Half-Bridge LLC Resonant Converter. By using a set of measured inputs, the Digital Twin system can characterize the component values, and hard to measure switch characteristics such as the Junction Temperature or Gate-Source Capacitance. This capability is useful for reliability or health applications, and may also be useful for adaptive control. A case study is done for the LLC Resonant converter designed around an operating point. Data sweeps around that operating point are simulated to form a dataset. An introduction to Neural Networks is given before determining the input vector for each output parameter. The results and final best design of each Neural Network will be discussed with experiments demonstrating their performance. The range of the Mean Absolute Errors for all parameters was from 0.3% to 7%, with an average over all parameters of 2.3%. Predictions on the boundaries of the dataset proved difficult, while the midpoints were fit without issue.

Download

Share

COinS