Compressed Sensing: from big data to relevant data. - In: Handbook of Nondestructive Evaluation 4.0, (2022), S. 329-352
Though the ever-increasing availability of digital data in the context of NDE 4.0 is mostly considered a blessing, it can turn to a curse quite rapidly: managing large amounts of data puts a burden on the sensor devices in terms of sampling and transmission, the networks, as well as the server infrastructure in terms of storing, maintaining, and accessing the data. Yet, NDE data can be highly redundant so the storage of massive amounts of data may indeed be wasteful. This is the main reason why focusing on relevant data as early as possible in the NDE process is highly advocated in the context of NDE 4.0. This chapter introduces Compressed Sensing as a potential approach to put this vision to practice. Compressed Sensing theory has shown that sampling signals with sampling rates that are significantly below the Shannon-Nyquist rate is possible without loss of information, provided that prior knowledge about the signals to be acquired is available. In fact, we may sample as low as the actual information rate if our prior knowledge is sufficiently accurate. In the NDE 4.0 context, prior knowledge can stem from the known inspection task and geometry but it can also include previous recordings of the same piece (such as in Structural Health Monitoring), information stored in the digital product memory along the products’ life cycle, or predictions generated through the products’ digital twins. In addition to data reduction, reconstruction algorithms developed in the Compressed Sensing community can be applied for enhanced processing of NDE data, providing added value in terms of accuracy or reliability. The chapter introduces Compressed Sensing basics and gives some concrete examples of its application in the NDE 4.0 context, in particular for ultrasound.
Deep learning aided interpolation of spatio-temporal nonstationary data. - In: 30th European Signal Processing Conference (EUSIPCO 2022), (2022), S. 2221-2225
Despite the growing interest in many fields, spatio-temporal (ST) interpolation remains challenging. Given ST nonstationary data distributed sparsely and irregularly over space, our objective is to obtain an equidistant representation of the region of interest (ROI). For this reason, an equidistant grid is defined within the ROI, where the available time series data are arranged, and the time series of the unobserved points are interpolated. Aiming to maintain the interpretability of the whole process while offering flexibility and fast execution, this work presents a ST interpolation frame-work which combines a statistical technique with deep learning. Our framework is generic and not confined to a specific application, which also provides the prediction confidence. To evaluate its validity, this framework is applied to ultrasound nondestructive testing (UT) data as an example. After the training with synthetic UT data sets, our framework is shown to yield accurate predictions when applied to measured UT data.
Preprocessing of freehand ultrasound synthetic aperture measurements using DNN. - In: 29th European Signal Processing Conference (EUSIPCO 2021), (2021), S. 1401-1405
Manual ultrasonic inspection is a widely used Nondestructive Testing (NDT) technique due to its simplicity and compatibility with complex structures. However, in contrast to the data acquired using a robotic positioner, manual measurements suffer from perturbations caused by a variable coupling and a varying scanning density. Imaging techniques like the synthetic aperture focusing technique rely on an unperturbed dense measurement from an equidistant measurement grid. Consequently, imaging based on freehand measurements leads to artifacts. This work aims at reducing such artifacts by preprocessing the manual measurements using Deep Neural Networks (DNN). The training of a DNN requires a large set of labeled measurements which is difficult to obtain in NDT. In this work, we present a technique to train the DNN using only synthetic data. We show that the resulting DNN generalizes well on real measurements. We present an improvement in Generalized Contrast to Noise Ratio by a factor of 20 and 3 compared to omitting the preprocessing for synthetic and measurement data, respectively.
Compressed ultrasound computed tomography in NDT. - In: IEEE IUS 2021, (2021), insges. 4 S.
Ultrasound Computed Tomography (UCT) is challenging due to phenomena such as strong refraction, multiple scattering, and mode conversion. In NDT, large speed of sound contrasts lead to strong artifacts if such phenomena are not modeled correctly; however, enhanced models are computationally expensive. In this work, a two-step framework for Compressed UCT based on the integral approach to the solution of the Helmholtz equation is presented. It comprises a physically motivated forward step and an imaging step that solves a suitable inverse problem. Multiple scattering is accounted for through the use of Neumann series. Convergence problems of Neumann series in high contrast settings are addressed via Padé approximants. Compressed sensing is employed to reduce the computational complexity of the reconstruction procedure by reducing data volumes directly at the measurement step, avoiding redundancy in the data and allowing the ability to steer the admissible computational effort at the expense of reconstruction quality. The proposed method is shown to yield high quality reconstructions under heavy subsampling in the frequency and spatial domains.
Frequency subsampling of ultrasound nondestructive measurements: acquisition, reconstruction, and performance. - In: IEEE transactions on ultrasonics, ferroelectrics, and frequency control, ISSN 1525-8955, Bd. 68 (2021), 10, S. 3174-3191
In ultrasound nondestructive testing (NDT), a widespread approach is to take synthetic aperture measurements from the surface of a specimen to detect and locate defects within it. Based on these measurements, imaging is usually performed using the synthetic aperture focusing technique (SAFT). However, SAFT is suboptimal in terms of resolution and requires oversampling in the time domain to obtain a fine grid for the delay-and-sum (DAS). On the other hand, parametric reconstruction algorithms give better resolution, but their usage for imaging becomes computationally expensive due to the size of the parameter space and a large amount of measurement data in realistic 3-D scenarios when using oversampling. In the literature, the remedies to this are twofold. First, the amount of measurement data can be reduced using state-of-the-art sub-Nyquist sampling approaches to measure Fourier coefficients instead of time-domain samples. Second, parametric reconstruction algorithms mostly rely on matrix-vector operations that can be implemented efficiently by exploiting the underlying structure of the model. In this article, we propose and compare different strategies to choose the Fourier coefficients to be measured. Their asymptotic performance is compared by numerically evaluating the Cramér-Rao bound (CRB) for the localizability of the defect coordinates. These subsampling strategies are then combined with an l1-minimization scheme to compute 3-D reconstructions from the low-rate measurements. Compared to conventional DAS, this allows us to formulate a fully physically motivated forward model matrix. To enable this, the projection operations of the forward model matrix are implemented matrix-free by exploiting the underlying two-level Toeplitz structure. Finally, we show that high-resolution reconstructions from as low as a single Fourier coefficient per A-scan are possible based on simulated data and measurements from a steel specimen.
Locally optimal subsampling strategies for full matrix capture measurements in pipe inspection. - In: Applied Sciences, ISSN 2076-3417, Bd. 11 (2021), 9, 4291, insges. 14 S.
In ultrasonic non-destructive testing, array and matrix transducers are being employed for applications that require in-field steerability or which benefit from a higher number of insonification angles. Having many transmit channels, on the other hand, increases the measurement time and renders the use of array transducers unfeasible for many applications. In the literature, methods for reducing the number of required channels compared to the full matrix capture scheme have been proposed. Conventionally, these are based on choosing the aperture that is as wide as possible. In this publication, we investigate a scenario from the field of pipe inspection, where cracks have to be detected in specific areas near the weld. Consequently, the width of the aperture has to be chosen according to the region of interest at hand. On the basis of ray-tracing simulations which incorporate a model of the transducer directivity and beam spread at the interface, we derive application specific measures of the energy distribution over the array configuration for given regions of interest. These are used to determine feasible subsampling schemes. For the given scenario, the validity/quality of the derived subsampling schemes are compared on the basis of reconstructions using the conventional total focusing method as well as sparsity driven-reconstructions using the Fast Iterative Shrinkage-Thresholding Algorithm. The results can be used to effectively improve the measurement time for the given application without notable loss in defect detectability.
Subsampling approaches for compressed sensing with ultrasound arrays in non-destructive testing. - In: Sensors, ISSN 1424-8220, Bd. 20 (2020), 23, 6734, insges. 23 S.
Full Matrix Capture is a multi-channel data acquisition method which enables flexible, high resolution imaging using ultrasound arrays. However, the measurement time and data volume are increased considerably. Both of these costs can be circumvented via compressed sensing, which exploits prior knowledge of the underlying model and its sparsity to reduce the amount of data needed to produce a high resolution image. In order to design compression matrices that are physically realizable without sophisticated hardware constraints, structured subsampling patterns are designed and evaluated in this work. The design is based on the analysis of the Cramér–Rao Bound of a single scatterer in a homogeneous, isotropic medium. A numerical comparison of the point spread functions obtained with different compression matrices and the Fast Iterative Shrinkage/Thresholding Algorithm shows that the best performance is achieved when each transmit event can use a different subset of receiving elements and each receiving element uses a different section of the echo signal spectrum. Such a design has the advantage of outperforming other structured patterns to the extent that suboptimal selection matrices provide a good performance and can be efficiently computed with greedy approaches.
Hardware architecture for ultra-wideband channel impulse response measurements using compressed sensing. - In: 28th European Signal Processing Conference (EUSIPCO 2020), (2020), S. 1663-1667
We propose a compact hardware architecture for measuring sparse channel impulse responses (IR) by extending the M-Sequence ultra-wideband (UWB) measurement principle with the concept of compressed sensing. A channel is excited with a periodic M-sequence and its response signal is observed using a Random Demodulator (RD), which observes pseudo-random linear combinations of the response signal at a rate significantly lower than the measurement bandwidth. The excitation signal and the RD mixing signal are generated from compactly implementable Linear Feedback Shift registers (LFSR) and operated from a common clock. A linear model is derived that allows retrieving an IR from a set of observations using Sparse-Signal-Recovery (SSR). A Matrix-free model implementation is possible due to the choice of synchronous LFSRs as signal generators, resulting in low computational complexity. For validation, real measurement data of a time-variant channel containing multipath components is processed by simulation models of our proposed architecture and the classic M-Sequence method. We show successful IR recovery using our architecture and SSR, outperforming the classic method significantly in terms of IR measurement rate. Compared to the classic method, the proposed architecture allows faster measurements of sparse time-varying channels, resulting in higher Doppler tolerance without increasing hardware or data stream complexity.
Acoustic process monitoring in laser beam welding. - In: 11th CIRP Conference on Photonic Technologies [LANE 2020], (2020), S. 763-768
Structure-borne acoustic emission (AE) measurement shows major advantages regarding quality assurance and process control in industrial applications. In this paper, laser beam welding of steel and aluminum was carried out under varying process parameters (welding speed, focal position) in order to provide data by means of structure-borne AE and simultaneously high-speed video recordings. The analysis is based on conventionally (e.g. filtering, autocorrelation, spectrograms) as well as machine learning methods (convolutional neural nets) and showed promising results with respect to the use of structure-borne AE for process monitoring using the example of spatter formation.
Cramér-Rao bounds for flaw localization in subsampled multistatic multichannel ultrasound NDT data. - In: 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, (2020), S. 4960-4964