Using Yolo Object Detection with PINN Integration for Thermal-Based Fault Detection in Integrated Circuits

This paper presents a novel approach combining YOLO (You Only Look Once) object detection with Physics-Informed Neural Networks (PINN) for identifying faulty integrated circuits (ICs) on printed circuit boards (PCBs) through thermal imaging analysis. The method leverages thermal signatures to detect anomalous components while using PINN to model heat propagation patterns, significantly reducing false positives. Our integrated approach achieves improved accuracy in fault detection compared to traditional methods, offering a more efficient and cost-effective solution for electronics remanufacturing and repair processes. The system demonstrates effectiveness in distinguishing between normally operating high-power components and genuinely faulty ICs through physics-based validation of thermal patterns.