Deep Learning-Based Open-Circuit Fault Diagnosis in Dual-Active-Bridge Converters for Integrated BMS with DC-DC Converter Applications
Year: 2026
Type: conference-paper
Abstract: Reliable operation of high-voltage battery systems requires effective cell balancing and robust fault detection in integrated battery management systems (BMSs). Converter-based architectures using dual-active-bridge (DAB) DC-DC converters enable efficient energy transfer but introduce reliability challenges due to open-circuit faults (OCFs) in power electronic switches. This paper addresses the problem of accurate and fast OCF diagnosis in an integrated BMS composed of multiple bidi-rectional DAB converters. A lightweight data-driven diagnostic method based on a one-dimensional convolutional neural network (1D CNN) is proposed, which directly processes multichannel time-domain signals from all DAB modules while maintaining low computational complexity for edge-device deployment. The approach is validated using data generated from a high-fidelity MATLAB-PLECS co-simulation framework covering one healthy case and 24 single-switch OCF scenarios. The trained model is converted to TensorFlow Lite and deployed on a Raspberry Pi 5, achieving an average computational inference time of$177 \mu \mathrm{s}$and a maximum below$400 \mu \mathrm{s}$per diagnostic window. The results demonstrate accurate simulation-based fault classification and embedded execution feasibility, while full hardware-based validation is identified as future work.
Institution: Vrije Universiteit Brussel
Language: English
Topic: Advanced DC-DC Converters
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7224747885