Mitigating RGB-D camera errors for robust ultrasonic inspections using a force-torque sensor

Tabatabaeipour, Morteza and Jackson, William and Gilmour, Adam and Zhang, Dayi and Poole, Alastair and Tzaferis, Konstantinos and Dobie, Gordon and Gachagan, Anthony (2024) Mitigating RGB-D camera errors for robust ultrasonic inspections using a force-torque sensor. Nondestructive Testing and Evaluation. ISSN 1058-9759 (In Press)

[thumbnail of Tabatabaeipour-etal-NTE-2024-Mitigating-RGB-D-camera-errors-for-robust-ultrasonic-inspections-using-a-force-torque-sensor] Text. Filename: Tabatabaeipour-etal-NTE-2024-Mitigating-RGB-D-camera-errors-for-robust-ultrasonic-inspections-using-a-force-torque-sensor.pdf
Accepted Author Manuscript
Restricted to Repository staff only until 1 January 2099.
License: Creative Commons Attribution-NonCommercial 4.0 logo
Download (7MB) | Request a copy

Abstract

Robot-based phased array ultrasonic testing is widely used for precise defect detection, particularly in complex geometries and various materials. To inspect constrained areas, a compact robot with a miniature arm can be employed, but payload limitations are present - limiting the choice of sensors. RGB-D cameras offer a solution due to their small size, light weight, and ability to rapidly capture RGB colour and depth data, enabling the generation of colorised 3D point clouds for scene representation. These point clouds aid in estimating surface normals used to align the ultrasound transducer on complex surfaces. However, relying solely on RGB-D cameras for transducer positioning can lead to inaccuracies in the captured point clouds, which impacts the direction of the ultrasonic beam and causes test result inaccuracies. This paper explores the impact of transducer pose and RGB-D camera limitations on ultrasonic inspections and proposes a novel method that utilises force-torque sensors to mitigate errors caused by inaccurately estimated normals from the camera. The force-torque sensor, integrated on the robot end effector, provides tactile feedback to the controller. This feedback enables the adjustment of joint angles, correcting for errors in the estimated normal. The practical experimental results demonstrate the successful application of ultrasound transducers using the force-torque sensor-based method, even in cases of significant misalignment. The adjustments took approximately 4 seconds to correct deviations from 12.55° to the angle, where the transducer can obtain meaningful results. An additional 4 seconds were used to ensure the probe was perfectly parallel to the surface, ultimately enhancing the accuracy of ultrasonic inspections in complex and constrained environments.

CORE (COnnecting REpositories)