Efficient PCB Fault Detection: A TensorRT-Based Inference Acceleration Approach

The demand for high-quality printed circuit boards (PCBs) necessitates rigorous fault detection to ensure product reliability, particularly considering the susceptibility of PCBs to connection issues in harsh environmental conditions. This paper presents a real-time fault detection system for PCBs utilizing the YOLOv8 object detection framework. A native YOLOv8 implementation is employed for model training, fine-tuning it on a custom PCB fault dataset to achieve precise detection of various defect types. Additionally, the trained model is optimized using NVIDIA's TensorRT framework, significantly enhancing inference speed. This approach enables the integration of high-performance, deep learning-based fault detection into resource-constrained PCB production environments, demonstrating superior accuracy and efficiency compared to traditional inspection methods. This advancement facilitates cost-effective quality control in PCB manufacturing. By showcasing the effectiveness of employing YOLOv8 and TensorRT for real-time fault detection in PCB manufacturing, the research underscores the importance of timely defect detection in ensuring the reliability and quality of electronic devices PCBs, representing a significant step forward in industrial applications.