Abstract:Imaging defects in austenitic welds presents a significant challenge for the ultrasonic non-destructive testing community. Due to the heating process during their manufacture, a dendritic structure develops, exhibiting large grains with locally anisotropic properties which cause the ultrasonic waves to scatter and refract. When basic imaging algorithms, which typically make constant wave speed assumptions, are applied to datasets arising from the inspection of these welds, the resulting defect reconstructions … Show more
“…In this section, we derive the inverse problems (), () from the more general setting of travel time tomography in anisotropic media. Our physical motivation is the determination of metal microstructure and defects from the travel times of ultrasonic waves 5,8,9,19 …”
Section: Derivation Of the Inverse Problemsmentioning
confidence: 99%
“…This can be done for the isotropic case (where ξ ( x , ν ) is independent of ν ) using for example the supergradient marching algorithm of Benmansour et al, 20 which applies automatic differentiation to a fast‐marching algorithm 21 . One approach to the anisotropic case could be to generalise Benmansour et al 20 to anisotropic metrics, using an anisotropic fast‐marching algorithm such as those presented in other studies, 8,22,23 but we do not pursue this here.…”
Section: Derivation Of the Inverse Problemsmentioning
confidence: 99%
“…In the ultrasonic non‐destructive evaluation (NDE) community, inspections are often carried out using a single transducer (as opposed to the arrays of sensors available in seismic and medical imaging), and so until recently, development of tomographic capabilities has been somewhat neglected. However, with the rapid uptake of ultrasonic phased arrays (which are capable of simultaneously transmitting and receiving ultrasound across multiple elements 5 ), tomographic inversion is now becoming feasible 6‐9 and presents a potential solution to the challenge of imaging defects embedded in complex media. There are many cases in industry where, due to manufacturing conditions or design, heterogeneous and locally anisotropic microstructures can arise 10‐12 and defect detection can prove difficult.…”
Section: Introductionmentioning
confidence: 99%
“…This results in a highly scattering, refractive and locally anisotropic medium where contributions to the scattered field by small defects are often obscured. However, it has been shown that a priori knowledge of such microstructures can be used to compensate for these effects and better focus the defect scattering energy to produce more reliable flaw reconstructions 7‐9,15 …”
Section: Introductionmentioning
confidence: 99%
“…Inspired by this particular problem in the ultrasonic NDE community, we propose a new approach to extracting information on the spatially varying material properties of this class of locally anisotropic media from travel time information collected on the boundary of the object. This specific problem has previously been tackled using Markov chain Monte Carlo (MCMC) methods, 7‐9 where an ensemble of samples estimating the posterior probability density function is produced and its associated moments are used to describe the probability distribution of the model given the observed data. These MCMC methods are generally applicable to most high‐dimensional inverse problems but are usually associated with prohibitive computational costs from the iterative runs of forward models over large and complex parameter spaces, which are in some cases inefficiently sampled 16 .…”
In this paper, we study the inverse problem of recovering the spatially varying material properties of a solid polycrystalline object from ultrasonic travel time measurements taken between pairs of points lying on the domain boundary. We consider a medium of constant density in which the orientation of the material's lattice structure varies in a piecewise constant manner, generating locally anisotropic regions in which the wave speed varies according to the incident wave direction and the material's known slowness curve. This particular problem is inspired by current challenges faced by the ultrasonic non‐destructive testing of polycrystalline solids. We model the geometry of the material using Voronoi tessellations and study two simplified inverse problems where we ignore wave refraction. In the first problem, the Voronoi geometry itself and the orientations associated to each region are unknowns. We solve this nonsmooth, nonconvex optimisation problem using a multistart non‐linear least squares method. Good reconstructions are achieved, but the method is shown to be sensitive to the addition of noise. The second problem considers the reconstruction of the orientations on a fixed square mesh. This is a smooth optimisation problem but with a much larger number of degrees of freedom. We prove that the orientations can be determined uniquely given enough boundary measurements and provide a numerical method that is more stable with respect to the addition of noise.
“…In this section, we derive the inverse problems (), () from the more general setting of travel time tomography in anisotropic media. Our physical motivation is the determination of metal microstructure and defects from the travel times of ultrasonic waves 5,8,9,19 …”
Section: Derivation Of the Inverse Problemsmentioning
confidence: 99%
“…This can be done for the isotropic case (where ξ ( x , ν ) is independent of ν ) using for example the supergradient marching algorithm of Benmansour et al, 20 which applies automatic differentiation to a fast‐marching algorithm 21 . One approach to the anisotropic case could be to generalise Benmansour et al 20 to anisotropic metrics, using an anisotropic fast‐marching algorithm such as those presented in other studies, 8,22,23 but we do not pursue this here.…”
Section: Derivation Of the Inverse Problemsmentioning
confidence: 99%
“…In the ultrasonic non‐destructive evaluation (NDE) community, inspections are often carried out using a single transducer (as opposed to the arrays of sensors available in seismic and medical imaging), and so until recently, development of tomographic capabilities has been somewhat neglected. However, with the rapid uptake of ultrasonic phased arrays (which are capable of simultaneously transmitting and receiving ultrasound across multiple elements 5 ), tomographic inversion is now becoming feasible 6‐9 and presents a potential solution to the challenge of imaging defects embedded in complex media. There are many cases in industry where, due to manufacturing conditions or design, heterogeneous and locally anisotropic microstructures can arise 10‐12 and defect detection can prove difficult.…”
Section: Introductionmentioning
confidence: 99%
“…This results in a highly scattering, refractive and locally anisotropic medium where contributions to the scattered field by small defects are often obscured. However, it has been shown that a priori knowledge of such microstructures can be used to compensate for these effects and better focus the defect scattering energy to produce more reliable flaw reconstructions 7‐9,15 …”
Section: Introductionmentioning
confidence: 99%
“…Inspired by this particular problem in the ultrasonic NDE community, we propose a new approach to extracting information on the spatially varying material properties of this class of locally anisotropic media from travel time information collected on the boundary of the object. This specific problem has previously been tackled using Markov chain Monte Carlo (MCMC) methods, 7‐9 where an ensemble of samples estimating the posterior probability density function is produced and its associated moments are used to describe the probability distribution of the model given the observed data. These MCMC methods are generally applicable to most high‐dimensional inverse problems but are usually associated with prohibitive computational costs from the iterative runs of forward models over large and complex parameter spaces, which are in some cases inefficiently sampled 16 .…”
In this paper, we study the inverse problem of recovering the spatially varying material properties of a solid polycrystalline object from ultrasonic travel time measurements taken between pairs of points lying on the domain boundary. We consider a medium of constant density in which the orientation of the material's lattice structure varies in a piecewise constant manner, generating locally anisotropic regions in which the wave speed varies according to the incident wave direction and the material's known slowness curve. This particular problem is inspired by current challenges faced by the ultrasonic non‐destructive testing of polycrystalline solids. We model the geometry of the material using Voronoi tessellations and study two simplified inverse problems where we ignore wave refraction. In the first problem, the Voronoi geometry itself and the orientations associated to each region are unknowns. We solve this nonsmooth, nonconvex optimisation problem using a multistart non‐linear least squares method. Good reconstructions are achieved, but the method is shown to be sensitive to the addition of noise. The second problem considers the reconstruction of the orientations on a fixed square mesh. This is a smooth optimisation problem but with a much larger number of degrees of freedom. We prove that the orientations can be determined uniquely given enough boundary measurements and provide a numerical method that is more stable with respect to the addition of noise.
Estimating the spatially varying microstructures of heterogeneous and locally anisotropic media non-destructively is necessary for the accurate detection of flaws and reliable monitoring of manufacturing processes. Conventional algorithms used for solving this inverse problem come with significant computational cost, particularly in the case of high-dimensional, nonlinear tomographic problems, and are thus not suitable for near-real-time applications. In this paper, for the first time, we propose a framework which uses deep neural networks (DNNs) with full aperture, pitch-catch and pulse-echo transducer configurations, to reconstruct material maps of crystallographic orientation. We also present the first application of generative adversarial networks (GANs) to achieve super-resolution of ultrasonic tomographic images, providing a factor-four increase in image resolution and up to a 50% increase in structural similarity. The importance of including appropriate prior knowledge in the GAN training data set to increase inversion accuracy is demonstrated: known information about the material’s structure should be represented in the training data. We show that after a computationally expensive training process, the DNNs and GANs can be used in less than 1 second (0.9 s on a standard desktop computer) to provide a high-resolution map of the material’s grain orientations, addressing the challenge of significant computational cost faced by conventional tomography algorithms.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.