In many geoscientific, material science, and engineering applications it is of importance to estimate a representative bulk seismic velocity of materials or to locate the source of recorded seismic or acoustic waves. Such estimates are necessary in order to interpret industrial seismic and earthquake seismological data, for example, in nondestructive evaluation and monitoring of structural materials, and as an input to rock physics models that predict other parameters of interest. Bulk velocity is commonly estimated in laboratories from the time of flight of the first‐arriving wave between a source and a receiver, assuming a linear raypath. In heterogeneous media, that method provides biased estimates of the bulk velocity, and of derived parameters such as temporal velocity changes or the locations of acoustic emissions. We show that coda wave interferometry (CWI) characterizes changes in the bulk properties of scattering media far more effectively on the scale of laboratory rock samples. Compared to conventional methods, CWI provides significant improvements in both accuracy and precision of estimates of velocity changes, and distances between pairs of acoustic sources, remaining accurate in the presence of background noise, and when source location and velocity perturbations occur simultaneously. CWI also allows 3‐D relative locations of clusters of acoustic emissions to be estimated using only a single sensor. We present a method to use CWI to infer changes in both P and S wave velocities individually. These innovations represent significant improvements in our ability to characterize the evolution of properties of media for a variety of applications.

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.

SUMMARY: Porosity exerts a strong control on the mechanical and hydraulic properties of rocks, but can often only be imaged indirectly from the surface using geophysical measurements, such as seismic velocity. Understanding and quantifying the relationship between seismic velocity and porosity is therefore a fundamental goal of many rock physics models. Simulating the geological processes that control porosity to generate digital rocks, and numerically modelling wave propagation to estimate their elastic properties, allows for flexible and rapid calibration of velocity–porosity trends. Here, the initial deposition of two digital carbonate sediments are simulated: grainstone (near spherical grains) and coquina (anisotropic shell fragments). The gradual precipitation of cement is then simulated, resulting in a suite of 3-D volumes of varying porosity with otherwise constant and known mineral and grain phases. These models are then used as input to a 3-D acoustic staggered-grid finite difference simulation of wavefield propagation, from which we estimate bulk seismic velocity and calculate the estimated bulk modulus. The resulting bulk modulus varies systematically with respect to porosity within the physical limits imposed by the Hashin–Shtrikman bounds. The samples exhibit anisotropy in the measured velocity consistent with structural anisotropy due to the settling of elongate grains under gravity. We use the resulting bulk velocity–porosity trends to test competing rock physics models, including one that accounts for varying effective pore-aspect ratio with porosity. The results validate the hypothesis that there is a power-law relationship between effective pore aspect ratio and porosity. This relationship is consistent with similar results obtained from a suite of natural carbonate grainstones examined in the laboratory. The results show the optimal rock physics model to be relatively insensitive to the degree of anisotropy in the fabric of the starting material, and may now be used with more confidence to link observed changes in effective pore aspect ratio to changes in porosity due to a range of geological processes, for example fracturing, dissolution and compaction, where other process-based models are available.

The ability to reliably detect and characterise defects embedded in austenitic steel welds depends on prior knowledge of microstructural descriptors, such as the orientations of the weld’s locally anisotropic grain structure. These orientations are usually unknown but it has been shown recently that they can be estimated from ultrasonic scattered wave data. However, conventional algorithms used for solving this inverse problem incur a significant computational cost. In this paper, we propose a framework which uses deep neural networks (DNNs) to reconstruct crystallographic orientations in a welded material from ultrasonic travel time data, in real-time. Acquiring the large amount of training data required for DNNs experimentally is practically infeasible for this problem, therefore a model based training approach is investigated instead, where a simple and efficient analytical method for modelling ultrasonic wave travel times through given weld geometries is implemented. The proposed method is validated by testing the trained networks on data arising from sophisticated finite element simulations of wave propagation through weld microstructures. The trained deep neural network predicts grain orientations to within 3° and in near real-time (0.04 s), presenting a significant step towards realising real-time, accurate characterisation of weld microstructures from ultrasonic non-destructive measurements. The subsequent improvement in defect imaging is then demonstrated via use of the DNN predicted crystallographic orientations to correct the delay laws on which the total focusing method imaging algorithm is based. An improvement of up to 5.3 dB in the signal-to-noise ratio is achieved.

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 non-linear tomographic problems. In this paper, 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 ever 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 dataset to increase inversion accuracy is highlighted; known information about the material's structure should be present in the training data. We show that after a computationally expensive training process, the DNNs and GANs can be used in less that one second (0.9 seconds on a standard desktop computer) to provide a high resolution map of the material's grain orientations.