Anomaly Detection and Fault Classification of Printed Circuit Boards Based on Multimodal Features of the Infrared Thermal Imaging

A heightened demand for improved printed circuit board (PCB) fault detection arises with the increasing integration and enhanced functionality of PCBs. Traditional visible light image analysis demonstrates accuracy and safety advantages in non-intrusive detection. However, it exhibits limitations in detecting obscured faults or lack of visible defects. The study proposes a fault detection framework based on infrared thermal imaging, aiming to enhance the practicality and engineering efficiency of PCB fault detection. The paper introduces an infrared thermography-based framework for anomaly detection and fault classification of PCBs. The framework encompasses preprocessing infrared thermal images, extraction of multimodal feature vectors, density-based anomaly detection, and fault classification based on deviation matrix clustering. The framework extracts multi-modal features from the residual temperature scalar fields and residual temperature gradient vector fields, employing multi-scale detection of global and local images for fault classification. Experimental validation on a two-phase drive circuit illustrates a significant enhancement in PCB anomaly detection and fault classification accuracy compared to existing feature extraction techniques. This research provides an innovative and practical PCB manufacturing and maintenance tool, effectively elevating anomaly detection accuracy and operational feasibility in engineering practices.