Automatic Detection of Concrete Defects Using Laser Ultrasonic Visualization Technique Based on Deep Learning

Common defects in concrete structures include cracks, spalling, blistering, delamination, pitting, and strain, which are typically detected visually and/or with a hammering test. However, the conventional inspection approaches are time-consuming and expensive, and are limited by the inspector’s judgement skills. In this research, an automatic nondestructive evaluation system for detecting concrete damage is developed to combine a Laser Ultrasonic Visualization Technique (LUVT) with deep learning and numerical simulation. The AI system created in this study accurately determines whether the concrete test specimens contain a defect.