Published May 19, 2026 | Version v1

AI-Powered PCB Defect Detection System with Explainable AI, Real-Time Feedback, and Severity Analysis

  • 1. Vishwakarma Institute of Technology

Description

Printed circuit board defects significantly impact the reliability and cost of electronic products. Traditional manual inspection methods are time-consuming, error-prone and unsuitable for modern high-speed manufacturing environments. This paper presents a smart AI based PCB defect detection system that integrates deep learning based classification, real time feedback, defect severity analysis, and explainable artificial intelligence i.e. XAI. The proposed system employs a convolutional neural network with transfer learning (EfficientNetB0) to detect and classify common PCB defects such as missing hole, mouse bite, open circuit, short, spur, and spurious copper. Severity levels are estimated based on defect size, location and class specific risk enabling prioritized decision making in manufacturing lines. To improve trust and interpretability gradient weighted class activation mapping (Grand CAM) is used to visualize regions influencing model prediction. Experimental results demonstrate and accuracy exceeding 95% on benchmark PCB datasets, highlighting the effectiveness and practicality of the proposed approach for industrial deployment.

Files

ai-powered-pcb-defect-detection-system-with-explainable-ai-real-time-feedback-and-severity-analysis-IJERTV15IS041934.pdf

Additional details