Real-Time Solar Radio Burst Detection with Machine Learning Trained on Physics-Based Synthetic Data

This presentation outlines a real-time solar radio burst detection system developed for the OVRO-LWA array. It leverages machine learning trained on physics-based synthetic data to automatically identify solar radio bursts in dynamic spectra within seconds. The system integrates fast beamformed data streaming, HDF-based data handling, and YOLO-based event detection, achieving sub-second latency for data delivery and rapid burst classification. Preliminary results demonstrate high efficiency on both simulated and observed datasets, with future improvements focusing on human-in-the-loop validation and continuous model feedback.