Robotic Inclined and Curved Surface Inspection
about.
We present a robotic ultrasonic pulse velocity (UPV) framework for through-thickness estimation and surface condition assessment of concrete, using a UR3 arm carrying a pair of 54 kHz transducers at fixed spacing. A RealSense depth camera provides point clouds for surface-normal tracking so the end-effector stays perpendicular on inclined/curved slabs, enabling reliable coupling and spatially indexed ToF measurements. Validation includes Abaqus/Explicit wave simulations and experiments on inclined and defect-containing specimens.
challenge.
UPV on non-planar geometry suffers from variable coupling, misalignment, and wave directivity errors that distort ToF and echo detection.
Manual scanning lacks repeatability and spatial indexing; probe separation causes oblique echo paths that must be modeled.
Detecting near-surface delaminations and small thickness variations requires robust first-arrival/echo picking under changing surface conditions.
results.
Adaptive alignment: Real-time point-cloud normals keep the probes normal to the surface; signals are synchronized with robot poses.
Thickness accuracy: Echo-based ToF (with oblique-path correction) yields 5.3–7.6 cm estimates with MAE ≈ 0.61 cm (~1.8%) and >90% valid rate over the scan area.
Signal processing: AIC first-arrival and Hilbert-envelope echo analysis improve robustness across coupling variations; shear vs. longitudinal waves (≈ 2727 vs. 3508.77 m/s) are compared for back-wall echo clarity.
Defect sensitivity: Robotic UPV velocity maps highlight embedded delaminations as low-velocity zones; FE shows ~4 µs arrival delay over a delaminated region, matching experiments.
testimonial.
“This robotic UPV system finally makes thickness checks reliable on real, non-flat concrete. The arm auto-aligns to curved and inclined surfaces, picks first arrivals cleanly, and turns scans into thickness and velocity maps that reveal near-surface delaminations—with accuracy on par with, or better than, manual methods.”
Ali Ghadimzadeh
Ph.D. Candidate, Mechanical Engineering
