Chung-Ang University Researchers Develop AI System for Real-Time Defect Detection

Researchers from Chung-Ang University have developed DiffectNet, a diffusion-enabled conditional target generation network designed to enhance non-destructive testing (NDT) by detecting internal material defects with high precision, announced in a press release.

Led by Assistant Professor Sooyoung Lee from the School of Mechanical Engineering, the team’s AI system generates high-fidelity and defect-aware ultrasonic images that outperform traditional inspection methods. The research, published in *Mechanical Systems and Signal Processing* on November 1, 2025, demonstrates how generative AI can reconstruct hidden cracks inside structures in real time.

DiffectNet addresses limitations of conventional NDT techniques, where sensor signals are often distorted by geometry and material properties. By simulating and analyzing internal defects using AI, the system enables accurate and rapid diagnostics without damaging the tested material.

The technology could be applied across sectors such as semiconductor manufacturing, aerospace, power generation, and civil infrastructure. It allows real-time monitoring of structural integrity, potentially preventing accidents and improving maintenance efficiency in safety-critical environments.