[2B4] Acoustic velocity mapping of multi-metallic components using laser-induced ultrasonic time-of-flight tomography and neural data interpretation
M Riding
University of Strathclyde, UK
Acoustic inhomogeneity and anisotropy are present in a variety of different engineering components. Multi-material components feature significant acoustic inhomogeneity due to the step-change in acoustic velocity at joints between dissimilar materials. Temperature gradients produce similar effects. Acoustic anisotropy is intrinsic to crystalline materials and can have important effects on ultrasonic signals at both the micro and macro scales, depending on the presence of large-scale crystalline texture. The effects of acoustic inhomogeneity and anisotropy are traditionally ignored in the processing of ultrasonic NDT data, leading to errors when these effects are significant. If, however, spatial variations in acoustic velocity (isotropic or anisotropic) can be mapped experimentally, then this data can be used to apply corrections. Such velocity maps may be acquired with destructive methods, but these are not suitable for in-situ deployment, for example during manufacturing or welding.
This study presents the results of applying non-destructive laser-ultrasonic time-of-flight tomography to the challenge of mapping acoustic velocity in several different inhomogeneous and anisotropic samples. The technique is deployed using laser-induced ultrasonic arrays and the data obtained is rapidly interpreted using a pre-trained neural network that outputs velocity maps in real time[1]. The velocity maps obtained from the technique are shown to be sufficiently accurate for use as stand-alone inspection data and are also highly valuable as corrective inputs to other UT techniques or imaging algorithms. Combining the flexibility innate to laser ultrasonic transduction with the reconstruction speed of neural networks enables robust acoustic velocity mapping for a variety of multi-metallic structures and component geometries with high potential for in-situ deployment during manufacturing. The results motivate further development of the technique towards the reconstruction of polycrystalline grain structures from ultrasonic time-of-flight data.
Reference
1. J Singh et al, ‘Real-time super-resolution mapping of locally anisotropic grain orientations for ultrasonic non-destructive evaluation of crystalline material’, Neural Computing and Applications, pp 1-18, 2021.
This study presents the results of applying non-destructive laser-ultrasonic time-of-flight tomography to the challenge of mapping acoustic velocity in several different inhomogeneous and anisotropic samples. The technique is deployed using laser-induced ultrasonic arrays and the data obtained is rapidly interpreted using a pre-trained neural network that outputs velocity maps in real time[1]. The velocity maps obtained from the technique are shown to be sufficiently accurate for use as stand-alone inspection data and are also highly valuable as corrective inputs to other UT techniques or imaging algorithms. Combining the flexibility innate to laser ultrasonic transduction with the reconstruction speed of neural networks enables robust acoustic velocity mapping for a variety of multi-metallic structures and component geometries with high potential for in-situ deployment during manufacturing. The results motivate further development of the technique towards the reconstruction of polycrystalline grain structures from ultrasonic time-of-flight data.
Reference
1. J Singh et al, ‘Real-time super-resolution mapping of locally anisotropic grain orientations for ultrasonic non-destructive evaluation of crystalline material’, Neural Computing and Applications, pp 1-18, 2021.
