Deep Learning Based Inversion of Locally Anisotropic Weld Properties from Ultrasonic Array Data

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.

Real-time super-resolution mapping of locally anisotropic grain orientations for ultrasonic non-destructive evaluation of crystalline material

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.

Microstructural differences between naturally-deposited and laboratory beach sands

The orientation of, and contacts between, grains of sand reflect the processes that deposit the sands. Grain orientation and contact geometry also influence mechanical properties. Quantifying and understanding sand microstructure thus provide an opportunity to understand depositional processes better and connect microstructure and macroscopic properties. Using x-ray computed microtomography, we compare the microstructure of naturally-deposited beach sands and laboratory sands created by air pluviation in which samples are formed by raining sand grains into a container. We find that naturally-deposited sands have a narrower distribution of coordination number (i.e., the number of grains in contact) and a broader distribution of grain orientations than pluviated sands. The naturally-deposited sand grains orient inclined to the horizontal, and the pluviated sand grains orient horizontally. We explain the microstructural differences between the two different depositional methods by flowing water at beaches that re-positions and reorients grains initially deposited in unstable grain configurations.

Physiochemical effect on the microgroove microstructure in microcutting of pure copper

The physiochemical effect on the microcutting of pure copper is studied through characterization on the cross-sectioned microgroove. An obvious reduction in cutting force and thrust force can be obtained with the application of ink surfactant. The surface roughness of microgroove with physiochemical effect is 12 nm, and that without physiochemical effect is 17 nm. The average grain size of the ink-affected sample is 67.9 µm within the microgroove zone, and that of the ink-free sample is 48.3 µm within the microgroove zone, moreover, the grain size of ink-free microgroove near the microgroove surface is larger than that far away from the microgroove surface. Additionally, the grain orientations of ink-affected cross-sectioned surface present anisotropy, while that of ink-free cross-sectioned surface are towards {101} direction. Based on the calculation and analysis of geometrically necessary dislocation, it can be inferred that the induced stress and temperature in the sample with physiochemical effect are higher than that without physiochemical effect, which can provide enough driving energy for recrystallization.

Physiochemical effect on the microgroove microstructure in microcutting of pure copper

Physiochemical effect on the machining of pure copper is studied via microstructure characterization on the cross-sectioned microgroove. An obvious decreased cutting force and thrust force were obtained with the application of surfactant. The surface roughness of microgroove with physiochemical effect is 12 nm, and that without physiochemical effect is 17 nm. The average grain size of the medium-affected sample is 67.9 µm within the microgroove zone, and that of the medium-free sample is 48.3 µm within the microgroove zone, moreover, the grain size of medium-free microgroove near the microgroove surface is larger than that far away from the microgroove surface. Additionally, the grain orientations of medium-affected cross-sectioned surface present anisotropy, while that of medium-free cross-sectioned surface are towards {101} direction. Based on the calculation and analysis of geometrically necessary dislocation, it can be inferred that the induced stress and temperature in the sample with physiochemical effect are higher than that without physiochemical effect, which can provide enough driving energy for recrystallization.

Qualitative Investigation of Damage Initiation at Meso-Scale in Spheroidized C45EC Steels by Using Crystal Plasticity-Based Numerical Simulations

This research uses EBSD data of two thermo-mechanically processed medium carbon (C45EC) steel samples to simulate micromechanical deformation and damage behavior. Two samples with 83% and 97% spheroidization degrees are subjected to virtual monotonic quasi-static tensile loading. The ferrite phase is assigned already reported elastic and plastic parameters, while the cementite particles are assigned elastic properties. A phenomenological constitutive material model with critical plastic strain-based ductile damage criterion is implemented in the DAMASK framework for the ferrite matrix. At the global level, the calibrated material model response matches well with experimental results, with up to ~97% accuracy. The simulation results provide essential insight into damage initiation and propagation based on the stress and strain localization due to cementite particle size, distribution, and ferrite grain orientations. In general, it is observed that the ferrite–cementite interface is prone to damage initiation at earlier stages triggered by the cementite particle clustering. Furthermore, it is observed that the crystallographic orientation strongly affects the stress and stress localization and consequently nucleating initial damage.

Effect of Crystallographic Orientation and Grain Boundaries on Martensitic Transformation and Superelastic Response of Oligocrystalline Fe–Mn–Al–Ni Shape Memory Alloys

In situ tensile tests employing digital image correlation were conducted to study the martensitic transformation of oligocrystalline Fe–Mn–Al–Ni shape memory alloys in depth. The influence of different grain orientations, i.e., near-〈001〉 and near-〈101〉, as well as the influence of different grain boundary misorientations are in focus of the present work. The results reveal that the reversibility of the martensite strongly depends on the type of martensitic evolving, i.e., twinned or detwinned. Furthermore, it is shown that grain boundaries lead to stress concentrations and, thus, to formation of unfavored martensite variants. Moreover, some martensite plates seem to penetrate the grain boundaries resulting in a high degree of irreversibility in this area. However, after a stable microstructural configuration is established in direct vicinity of the grain boundary, the transformation begins inside the neighboring grains eventually leading to a sequential transformation of all grains involved.

On the multiscale analysis of a two phase material: crystal plasticity versus mean field

In this paper, a comparison is made between two multiscale methods, namely crystal plasticity finite element and mean field on a material composed of two phases. Both methods are used to homogenize a given microstructure. In order to obtain macroscopic behavior, in the mean field approach, a Self-Consistent scheme is used to evaluate stress and strain partitioning among the phases. In this method, an average of the fields is estimated and local distributions cannot be captured. In parallel, crystal plasticity simulations on Representative Volume Elements (RVEs) composed of hexagonal grains are performed. In these simulations, grain orientations are attributed randomly respecting Mackenzie's distribution function in order to achieve isotropic behavior and macroscopic hardening is extracted from the simulations. The results on macroscopic hardening of both methods are compared to distinguish the extents of validity of mean field homogenization. In addition to Self- Consistent, other mean field schemes such as Voigt, Reuss and Bound-Interpolation are compared in terms of efficiency and accuracy. The comparison manifests that Self-Consistent scheme is capable of predicting material behavior well.