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

Abstract

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.

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

“…For each of the transducer array configurations shown in Figure 3, there are 32 transmitting elements and 32 receiving elements, therefore there are 32 × 32 = 1024 ToF measurements and 1024 nodes in the input layer of the DNN. We follow the approach of [25,40], where a separate network is trained for each output model sub-region, and so a total of 91 networks are trained to predict the grain orientations of the sub-regions within the weld. An additional set of constant hyperparameters need to be assigned before the learning process can begin: such as the learning rate and decay rate used for the Adam optimisation, the number of hidden layers, the number of nodes in each layer and the non-linear activation function.…”

Research on Super-Resolution Relationship Extraction and Reconstruction Methods for Images Based on Multimodal Graph Convolutional Networks

“…e interaction enhancement between local information includes the interaction between the internal elements of local target information and local image information and the interaction between local target information and local image information. e principle of internal element interaction enhancement is that a subset of elements that are relatively important or create a common theme can be calculated using the interrelationship between the interior features [24].…”

Deep learning in automated ultrasonic NDE -- developments, axioms and opportunities

“…The end goal of DL is to enable the estimation of a property difficult to obtain, which typically relies on structural noise distributions. Singh et al [67] proposed two DL architectures for creating ultrasonic tomographic images with grain orientation information. The first model consists of multiple DNNs (one per pixel) that take as input the ToF matrices from ultrasonic arrays and provide the grain orientation information as output.…”