Session: Structural Health Monitoring I & II, Material Characterization by Ultrasonic wave

Paper Number: 170848

170848 - Ultrasonic Characterisation of Thick Section Welds Assisted by Ai and Ray-Based Methods. 

Abstract: 

ABSTRACT

 

Data interpretation of wave propagation in inhomogeneous materials remains a challenging task. This rises as a topic of interest in the nuclear industry, where the inspection of inhomogeneous thick section welds is vital to meet safety standards. Such welds are characterised by columnar grains with varying preferred orientations resultant of their complex solidification process. Consequently, conventional post-processing techniques of ultrasound data are imprecise, as the ultrasonic beam is distorted and attenuated, no longer following a straight path.

 

Conventional algorithms utilised for defect detection operates on the assumption of a homogeneous and isotropic medium, where the time-of-flight depends solely on the distance between the transmitter and receiver pair, which is known since the wave travels along a straight path. In heterogeneous anysotropic media, however, the same assumption does not hold. The spacial variation of the material properties and consequent velocity dispersion are responsible for distorting the ultrasonic wavepaths. Under these circumstances, employing a homogenised post-processing model to inspect heterogeneous media will lead to incorrect conclusions regarding, for example, defect size and location. However, the ultrasonic image can be corrected if the detailed material information is coupled with an appropriate propagation model [1].

 

We propose a workflow to extract the crystallographic orientation from time-of-flight maps implemented through the development of a Deep Neural Network metamodel. The workflow makes use of numerically generated welds, and the fast ray tracing solver based on the shortest ray path (SRP) principle to determine propagation times. The weld configurations are generated according to the generalized Ogilvy map, sampled in a parameter space with up to ten variables such as the dominant grain orientation, the rate of change of the grain orientations, chamfer angle, array position, and weld bead lengths. Therefore, the characterisation considers both geometric and material properties, encompassing a broader range of welds. The AI-enabled process is also compared with ray-based methods, providing a better understanding of the relationship between microstructure and wave propagation, given the relevant parameters, and enhancing ultrasound data interpretation for defect localisation [2].

 

Keywords: ultrasound, material characterisation, inversion, optimisation, deep learning.

 

REFERENCES

[1] Michał K. Kalkowski, Michael J. S. Lowe, Vykintas Samaitis, Fabian Schreyer, and Sébastien Robert. Weld map tomography for determining local grain orientations from ultrasound. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 479, 20230236 (2023).

[2] Jonathan Singh, Katherine Tant, Andrew Curtis, and Anthony Mulholland. Real-time super-resolution mapping of locally anisotropic grain orientations for ultrasonic non-destructive evaluation of crystalline material. Neural Computing and Applications 34, 4993–5010 (2022).

Presenting Author: Lucas Queiroz Machado University of Southampton

Presenting Author Biography: Lucas is a Research Fellow at the Institute of Sound and Vibration Research (ISVR), University of Southampton. He holds a PhD in Mechanical Engineering from Heriot-Watt University in Edinburgh, UK, with full funding from CNPq. His current research focuses on the ultrasonic data interpretation and characterisation of inhomogeneous anisotropic materials—particularly complex austenitic steel welds commonly found in the nuclear industry.

Authors:

Lucas Queiroz Machado University of Southampton
Michal Kalkowski University of Southampton
Thomas Blumensath University of Southampton
Michael Lowe Imperial College London