This study presents an automated ultrasonic pulse velocity (UPV) inspection framework for evaluating inclined and curved concrete surfaces using a UR3 robotic arm. The system utilizes motion planning and surface normal estimation to maintain consistent sensor coupling across complex geometries.

The inspection setup was implemented both experimentally and in simulation, featuring an inclined concrete specimen scanned with a 50 kHz longitudinal transducer. The robotic framework uses 3D point cloud data to estimate local surface normals and dynamically adjust the end-effector orientation, ensuring perpendicular contact during ultrasonic transmission and reception.

The accompanying simulation demonstrates real-time robot motion, tool-path generation, and surface reconstruction, validating the kinematic control pipeline for adaptive inspection of non-planar structures. This setup extends conventional UPV methods to irregular geometries, offering higher repeatability and reduced operator dependency.

A conference paper based on this work has been published, and a journal version is under preparation, expanding on multi-geometry calibration, uncertainty quantification, and comparison against manual inspection benchmarks.