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 are often distorted and difficult to interpret correctly. However, knowledge of the underlying spatially varying material properties allows correction of the expected wave travel times and thus results in more reliable defect reconstructions. In this paper, an approximation to the underlying, locally anisotropic structure of the weld is constructed from ultrasonic time of flight data. A new forward model of wave front propagation in locally anisotropic media is presented and used within the reversible-jump Markov chain Monte Carlo method to invert for the map of effective grain orientations across different regions of the weld. This forward model and estimated map are then used as the basis for an advanced imaging algorithm and the resulting defect reconstructions exhibit a significant improvement across multiple flaw characterization metrics.

Traditional imaging algorithms within the ultrasonic non-destructive testing community typically assume that the material being inspected is primarily homogeneous, with heterogeneities only at sub-wavelength scales. When the medium is of a more generally heterogeneous nature, this assumption can contribute to the poor detection, sizing and characterisation of defects. Prior knowledge of the varying wave speeds within the component would allow more accurate imaging of defects, leading to better decisions about how to treat the damaged component. This work endeavours to reconstruct the inhomogeneous wave speed maps of random media from simulated ultrasonic phased array data. This is achieved via application of the reversible-jump Markov chain Monte Carlo method: a sampling-based approach within a Bayesian framework. The inverted maps are used in conjunction with an imaging algorithm to correct for deviations in the wave speed, and the reconstructed flaw images are then used to quantitatively assess the success of this methodology. Using full matrix capture data arising from a finite element simulation of a phased array inspection of a heterogeneous component, a six-fold improvement in flaw location is achieved by taking into account the reconstructed wave speed map which exploits almost no a priori knowledge

Ultrasonic phased array systems have become increasingly popular in the last 10 years as tools for flaw detection and characterization within the nondestructive testing industry. The existence and location of flaws can often be deduced via images generated from the data captured by these arrays. A factor common to these imaging techniques is the subjective thresholding required to estimate the size of the flaw. This paper puts forward an objective approach which employs a mathematical model. By exploiting the relationship between the width of the central lobe of the scattering matrix and the crack size, an analytical expression for the crack length is reached via the Born approximation. Conclusions are then drawn on the minimum resolvable crack length of the method and it is thus shown that the formula holds for subwavelength defects. An analytical expression for the error that arises from the discrete nature of the array is then derived and it is observed that the method becomes less sensitive to the discretization of the array as the distance between the flaw and array increases. The methodology is then extended and tested on experimental data collected from welded austenitic plates containing a lack-of-fusion crack of 6 mm length. An objective sizing matrix (OSM) is produced by assessing the similarity between the scattering matrices arising from experimentally collected data with those arising from the Born approximation over a range of crack lengths and frequencies. Initially, the global minimum of the OSM is taken as the objective estimation of the crack size, giving a measurement of 7 mm. This is improved upon by the adoption of a multifrequency averaging approach, with which an improved crack size estimation of 6.4 mm is obtained.

The non-destructive testing of austenitic welds using ultrasound plays an important role in the assessment of the structural integrity of safety critical structures. The internal microstructure of these welds is highly scattering and can lead to the obscuration of defects when investigated by traditional imaging algorithms. This paper proposes an alternative objective method for the detection of flaws embedded in austenitic welds based on the singular value decomposition of the time-frequency domain response matrices. The distribution of the singular values is examined in the cases where a flaw exists and where there is no flaw present. A lower threshold on the singular values, specific to austenitic welds, is derived which, when exceeded, indicates the presence of a flaw. The detection criterion is successfully implemented on both synthetic and experimental data. The datasets arising from welds containing a flaw are further interrogated using the decomposition of the time-reversal operator (DORT) method and the total focusing method (TFM), and it is shown that images constructed via the DORT algorithm typically exhibit a higher signal-to-noise ratio than those constructed by the TFM algorithm.

Many safety critical structures, such as those found in nuclear plants, oil pipelines and in the aerospace industry, rely on key components that are constructed from heterogeneous materials. Ultrasonic non-destructive testing (NDT) uses high-frequency mechanical waves to inspect these parts, ensuring they operate reliably without compromising their integrity. It is possible to employ mathematical models to develop a deeper understanding of the acquired ultrasonic data and enhance defect imaging algorithms. In this paper, a model for the scattering of ultrasonic waves by a crack is derived in the time–frequency domain. The fractional Fourier transform (FrFT) is applied to an inhomogeneous wave equation where the forcing function is prescribed as a linear chirp, modulated by a Gaussian envelope. The homogeneous solution is found via the Born approximation which encapsulates information regarding the flaw geometry. The inhomogeneous solution is obtained via the inverse Fourier transform of a Gaussian-windowed linear chirp excitation. It is observed that, although the scattering profile of the flaw does not change, it is amplified. Thus, the theory demonstrates the enhanced signal-to-noise ratio permitted by the use of coded excitation, as well as establishing a time–frequency domain framework to assist in flaw identification and classification.

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.