Deep Learning for Ultrasonic Crack Characterization in NDE

On the use of a Transformer Neural Network to deconvolve ultrasonic signals

Pulse-echo ultrasonic techniques play a crucial role in assessing wall thickness deterioration in safety-critical industries. Current approaches face limitations with low signal-to-noise ratios, weak echoes, or vague echo patterns typical of heavily corroded profiles. This study proposes a novel combination of Convolution Neural Networks (CNN) and Transformer Neural Networks (TNN) to improve thickness gauging accuracy for complex geometries and echo patterns. Recognizing the strength of TNN in language processing and speech recognition, the proposed network comprises three modules: 1. pre-processing CNN, 2. a Transformer model and 3. a post-processing CNN. Two datasets, one being simulation-generated, and the other, experimentally gathered from a corroded carbon steel staircase specimen, support the training and testing processes. Results indicate that the proposed model outperforms other AI architectures and traditional methods, providing a 5.45% improvement over CNN architectures from NDE literature, a 1.81% improvement over ResNet-50, and a 17.5% improvement compared to conventional thresholding techniques in accurately detecting depths with a precision under 0.5λ.

Acoustic emission with simulation of simultaneous ultrasonic guided wave propagation & crack propagation

Advancement of computation nondestructive evaluation (CNDE) creates an opportunity to visualize predicted signals received by sensors and may aid the development of artificial intelligence (AI) for NDE 4.0. However, traditional methods face limitations for crack propagation and guided wave propagation simulation, simultaneously. Modeling crack propagation using mesh-based method requires remeshing and implementation of cohesive zone model to name a few alternatives. Multiple meshfree methods have also been implemented for crack propagation but did not immediately translate to simulate the guided waves that are used to interrogate the cracks under nondestructive evaluation (NDE) framework. Ultrasonic CNDE with new era of Machine Learning (ML)/AI requires understanding the signals and its physics-based features when the guided waves propagate to interact with the crack while the crack is simultaneously growing at different time scales. To enable the future of physics to be informed and physics driven ML/AI this article presents a framework of CNDE where guided wave propagation and crack propagation are simultaneously simulated without remeshing and creates an enabling approach for the future AI implementation. A few successful case studies are presented for feasibility demonstration. Detailed flowcharts are presented for easy implementation of the method for the ultrasonic NDE community.

Multi-Harmonic Nonlinear Ultrasonic Fusion with Deep Learning for Subtle Parameter Identification of Micro-Crack Groups

Fatigue crack defects in metallic materials significantly reduce the remaining useful life (RUL) of parts. However, much of the existing research has focused on identifying single-millimeter-scale cracks using individual nonlinear ultrasonic responses. The identification of subtle parameters from complex ultrasonic responses of micro-crack groups remains a significant challenge in the field of nondestructive testing. We propose a novel multi-harmonic nonlinear response fusion identification method integrated with a deep learning (DL) model to identify the subtle parameters of micro-crack groups. First, we trained a one-dimensional convolutional neural network (1D CNN) with various time-domain signals obtained from finite element method (FEM) models and analyzed the sensitivity of different harmonic nonlinear responses to various subtle parameters of micro-crack groups. Then, high harmonics were fused to perform a decoupled identification of multiple subtle parameters. We enhanced the Dempster-Shafer (DS) evidence theory used in decision fusion by accounting for different sensitivities, achieving an identification accuracy of 93.73%. Building on this, we assigned sensor weights based on our proposed new conflict measurement method and further conducted decision fusion on the decision results from multiple ultrasonic sensors. Our proposed method achieves an identification accuracy of 95.68%.

Leveraging Segment Anything Model (SAM) for Weld Defect Detection in Industrial Ultrasonic B-Scan Images

Automated ultrasonic testing (AUT) is a critical tool for infrastructure evaluation in industries such as oil and gas, and, while skilled operators manually analyze complex AUT data, artificial intelligence (AI)-based methods show promise for automating interpretation. However, improving the reliability and effectiveness of these methods remains a significant challenge. This study employs the Segment Anything Model (SAM), a vision foundation model, to design an AI-assisted tool for weld defect detection in real-world ultrasonic B-scan images. It utilizes a proprietary dataset of B-scan images generated from AUT data collected during automated girth weld inspections of oil and gas pipelines, detecting a specific defect type: lack of fusion (LOF). The implementation includes integrating knowledge from the B-scan image context into the natural image-based SAM 1 and SAM 2 through a fully automated, promptable process. As part of designing a practical AI-assistant tool, the experiments involve applying both vanilla and low-rank adaptation (LoRA) fine-tuning techniques to the image encoder and mask decoder of different variants of both models, while keeping the prompt encoder unchanged. The results demonstrate that the utilized method achieves improved performance compared to a previous study on the same dataset.

Ultrasonic Rough Crack Characterization Using Time-of-Flight Diffraction With Self-Attention Neural Network

Time-of-flight diffraction (ToFD) is a widely used ultrasonic nondestructive evaluation (NDE) method for locating and characterizing rough defects, with high accuracy in sizing smooth cracks. However, naturally grown defects often have irregular surfaces, complicating the received tip diffraction waves and affecting the accuracy of defect characterization. This article proposes a self-attention (SA) deep learning method to interpret the ToFD A-scan signals for sizing rough defects. A high-fidelity finite-element (FE) simulation software Pogo is used to generate the synthetic datasets for training and testing the deep learning model. Besides, the transfer learning (TL) method is used to fine-tune the deep learning model trained by the Gaussian rough defects to boost the performance of characterizing realistic thermal fatigue rough defects. An ultrasonic experiment using 2-D rough crack samples made by additive manufacturing is conducted to validate the performance of the developed deep learning model. To demonstrate the accuracy of the proposed method, the crack characterization results are compared with those obtained using the conventional Hilbert peak-to-peak sizing method. The results indicate that the deep learning method achieves significantly reduced uncertainty and error in rough defect characterization, in comparison with traditional sizing approaches used in ToFD measurements.

Advancing Carbon Fiber Composite Inspection: Deep Learning-Enabled Defect Localization and Sizing via 3-D U-Net Segmentation of Ultrasonic Data

In nondestructive evaluation (NDE), accurately characterizing defects within components relies on accurate sizing and localization to evaluate the severity or criticality of defects. This study presents for the first time a deep learning (DL) methodology using 3-D U-Net to localize and size defects in carbon fiber reinforced polymer (CFRP) composites through volumetric segmentation of ultrasonic testing (UT) data. Using a previously developed approach, synthetic training data, closely representative of experimental data, was used for the automatic generation of ground truth segmentation masks. The model's performance was compared to the conventional amplitude 6 dB drop analysis method used in the industry against ultrasonic defect responses from 40 defects fabricated in CFRP components. The results showed good agreement with the 6 dB drop method for in-plane localization and excellent through-thickness localization, with mean absolute errors (MAEs) of 0.57 and 0.08 mm, respectively. Initial sizing results consistently oversized defects with a 55% higher mean average error than the 6 dB drop method. However, when a correction factor was applied to account for variation between the experimental and synthetic domains, the final sizing accuracy resulted in a 35% reduction in MAE compared to the 6 dB drop technique. By working with volumetric ultrasonic data (as opposed to 2-D images), this approach reduces preprocessing (such as signal gating) and allows for the generation of 3-D defect masks which can be used for the generation of computer-aided design files; greatly reducing the qualification reporting burden of NDE operators.

Machine learning modeling for ultrasonic quality attribute assessment of pharmaceutical tablets for continuous manufacturing and real-time release testing

In in-process quality monitoring for Continuous Manufacturing (CM) and Critical Quality Attributes (CQA) assessment for Real-time Release (RTR) testing, ultrasonic characterization is a critical technology for its direct, non-invasive, rapid, and cost-effective nature. In quality evaluation with ultrasound, relating a pharmaceutical tablet's ultrasonic response to its defect state and quality parameters is essential. However, ultrasonic CQA characterization requires a robust mathematical model, which cannot be obtained with traditional first principles-based modeling approaches. Machine Learning (ML) using experimental data is emerging as a critical analytical tool for overcoming such modeling challenges. In this work, a novel Deep Neural Network-based ML-driven Non-Destructive Evaluation (ML-NDE) modeling framework is developed, and its effectiveness for extracting and predicting three CQAs, namely defect states, compression force levels, and amounts of disintegrant, is demonstrated. Using a robotic tablet handling experimental rig, each attribute's distinct waveform dataset was acquired and utilized for training, validating, and testing the respective ML models. This study details an advanced algorithmic quality assessment framework for pharmaceutical CM in which automated RTR testing is expected to be critical in developing cost-effective in-process real-time monitoring systems. The presented ML-NDE approach has demonstrated its effectiveness through evaluations with separate (unused) test datasets.

Machine learning framework for extracting micro-viscoelastic and micro-structural properties of compressed oral solid dosage forms

A compressed pharmaceutical oral solid dosage (OSD) form is a strongly micro-viscoelastic material composite arranged as a network of agglomerated particles due to its constituent powders and their bonding and fractural mechanical properties. An OSD product's Critical Quality Attributes, such as disintegration, drug release (dissolution) profile, and structural strength ("hardness"), are influenced by its micro-scale properties. Ultrasonic evaluation is direct, non-destructive, rapid, and cost-effective. However, for practical process control applications, the simultaneous extraction of the micro-viscoelastic and scattering properties from a tablet's ultrasonic response requires a unique solution to a challenging inverse mathematical wave propagation problem. While the spatial progression of a pulse traveling in a composite medium with known micro-scale properties is a straightforward computational task when its dispersion relation is known, extracting such properties from the experimentally acquired waveforms is often non-trivial. In this work, a novel Machine Learning (ML)-based micro-property extraction technique directly from waveforms, based on Multi-Output Regression models and Neural Networks, is introduced and demonstrated. Synthetic waveforms with a given set of micro-properties of virtual tablets are computationally generated to train, validate, and test the developed ML models for their effectiveness in the inverse problem of recovering specified micro-scale properties. The effectiveness of these ML models is then tested and demonstrated for a set of physical OSD tablets. The micro-viscoelastic and micro-structural properties of physical tablets with known properties have been extracted through experimentally acquired waveforms to exhibit their consistency with the generated ML-based attenuation results.

Shear Wave Ultrasound Inspection of Flaws in Silicon Wafers Using Focused Transducers

Silicon parts can contain micrometer-sized vertical cracks that are challenging to detect. Inspection using high-frequency focused ultrasound has shown promise for detecting defects of this size and geometry. However, implementing focused ultrasound to inspect anisotropic media can prove challenging, given the directional dependence of wave propagation and subsequent focusing behavior. In this work, back surface-breaking defects at various orientations within silicon wafers (0°, 15°, and 45° relative to the [010] crystallographic axis) are experimentally inspected in an immersion tank setup. Using 100 MHz unfocused and focused shear waves, the impact of medium anisotropy on focusing and defect detection is evaluated. The scattering amplitude and defect detection sensitivity results demonstrate orientation-dependent patterns that strongly rely on the use of focused transducers. The defects along the 45° orientation reveal two-lobe scattering patterns with maximum amplitudes less than half that of the defects in the 0° orientation, which in contrast show a one-lobe scattering pattern. The experimental results are further explored using finite element (FE) modeling and ray tracing to visualize the impact of focusing on wave propagation within the silicon. Ray tracing results show that the focused beam profiles for the 45° and 0° orientations form a butterfly wing and elliptical focusing profile, respectively, which correspond directly to experimentally found scattering patterns from defects. Additionally, the FE scattering results from unfocused transducers reveal single lobe scattering for both 0° and 45° orientations, proving the varying scattering patterns to be driven by the anisotropic focusing behavior.

Application of deep neural networks for inferring pressure in polymeric acoustic transponders/sensors

Passive sensor-transponders have raised interest for the last few decades, due to their capability of low-cost remote monitoring without the need for energy storage. Their operating principle includes receiving a signal from a source and then reflecting the signal. While well-established transponders operate through electromagnetic antennas, those with a fully acoustic design have advantages such as lower cost and simplicity. Therefore, detection of pressures using the ultrasound signal that is backscattered from an acoustic resonator has been of interest recently. In order to infer the pressure from the backscattered signal, the established approach has been based upon the principle of detection of the shift to the frequency of resonance. Nevertheless, regression of the pressure from the signal with a small error is challenging and has been subject to research. Here in this paper, we explore an approach that employs deep learning for inferring pressure from the ultrasound reflections of polymeric resonators. We assess if neural network regressors can efficiently infer pressure reflected from a fully acoustic transponder. For this purpose, we compare the performance of several regressors such as a convolutional neural network, a network inspired by the ResNet, and a fully connected neural network. We observe that deep neural networks are advantageous in inferring pressure information with a minimal need for analyzing the signal. Our work suggests that a deep learning approach has the potential to be integrated with or replace other traditional approaches for inferring pressure from an ultrasound signal reflected from fully acoustic transponders or passive sensors.

A review of synthetic and augmented training data for machine learning in ultrasonic non-destructive evaluation

Ultrasonic Testing (UT) has seen increasing application of machine learning (ML) in recent years, promoting higher-level automation and decision-making in flaw detection and classification. Building a generalized training dataset to apply ML in non-destructive evaluation (NDE), and thus UT, is exceptionally difficult since data on pristine and representative flawed specimens are needed. Yet, in most UT test cases flawed specimen data is inherently rare making data coverage the leading problem when applying ML. Common data augmentation (DA) strategies offer limited solutions as they don't increase the dataset variance, which can lead to overfitting of the training data. The virtual defect method and the recent application of generative adversarial neural networks (GANs) in UT are sophisticated DA methods targeting to solve this problem. On the other hand, well-established research in modeling ultrasonic wave propagations allows for the generation of synthetic UT training data. In this context, we present a first thematic review to summarize the progress of the last decades on synthetic and augmented UT training data in NDE. Additionally, an overview of methods for synthetic UT data generation and augmentation is presented. Among numerical methods such as finite element, finite difference, and elastodynamic finite integration methods, semi-analytical methods such as general point source synthesis, superposition of Gaussian beams, and the pencil method as well as other UT modeling software are presented and discussed. Likewise, existing DA methods for one- and multidimensional UT data, feature space augmentation, and GANs for augmentation are presented and discussed. The paper closes with an in-detail discussion of the advantages and limitations of existing methods for both synthetic UT training data generation and DA of UT data to aid the decision-making of the reader for the application to specific test cases.

Interpretable and Explainable Machine Learning for Ultrasonic Defect Sizing

Despite its popularity in literature, there are few examples of machine learning (ML) being used for industrial nondestructive evaluation (NDE) applications. A significant barrier is the "black box" nature of most ML algorithms. This article aims to improve the interpretability and explainability of ML for ultrasonic NDE by presenting a novel dimensionality reduction method: Gaussian feature approximation (GFA). GFA involves fitting a 2-D elliptical Gaussian function in an ultrasonic image and storing the seven parameters that describe each Gaussian. These seven parameters can then be used as inputs to data analysis methods such as the defect-sizing neural network presented in this article. GFA is applied to ultrasonic defect sizing for inline pipe inspection as an example application. This approach is compared to sizing with the same neural network, and two other dimensionality reduction methods [the parameters of 6 dB drop boxes and principal component analysis (PCA)], as well as a convolutional neural network (CNN) applied to raw ultrasonic images. Of the dimensionality reduction methods tested, GFA features produce the closest sizing accuracy to the sizing from the raw images, with only a 23% increase in root mean square error (RMSE), despite a 96.5% reduction in the dimensionality of the input data. Implementing ML with GFA is implicitly more interpretable than doing so with PCA or raw images as inputs, and gives significantly more sizing accuracy than 6 dB drop boxes. Shapley additive explanations (SHAPs) are used to calculate how each feature contributes to the prediction of an individual defect's length. Analysis of SHAP values demonstrates that the GFA-based neural network proposed displays many of the same relationships between defect indications and their predicted size as occur in traditional NDE sizing methods.

Recognition Performance Analysis of a Multimodal Biometric System Based on the Fusion of 3D Ultrasound Hand-Geometry and Palmprint

Multimodal biometric systems are often used in a wide variety of applications where high security is required. Such systems show several merits in terms of universality and recognition rate compared to unimodal systems. Among several acquisition technologies, ultrasound bears great potential in high secure access applications because it allows the acquisition of 3D information about the human body and is able to verify liveness of the sample. In this work, recognition performances of a multimodal system obtained by fusing palmprint and hand-geometry 3D features, which are extracted from the same collected volumetric image, are extensively evaluated. Several fusion techniques based on the weighted score sum rule and on a wide variety of possible combinations of palmprint and hand geometry scores are experimented with. Recognition performances of the various methods are evaluated and compared through verification and identification experiments carried out on a homemade database employed in previous works. Verification results demonstrated that the fusion, in most cases, produces a noticeable improvement compared to unimodal systems: an EER value of 0.06% is achieved in at least five cases against values of 1.18% and 0.63% obtained in the best case for unimodal palmprint and hand geometry, respectively. The analysis also revealed that the best fusion results do not include any combination between the best scores of unimodal characteristics. Identification experiments, carried out for the methods that provided the best verification results, consistently demonstrated an identification rate of 100%, against 98% and 91% obtained in the best case for unimodal palmprint and hand geometry, respectively.

Automated Real-Time Eddy Current Array Inspection of Nuclear Assets

Inspection of components with surface discontinuities is an area that volumetric Non-Destructive Testing (NDT) methods, such as ultrasonic and radiographic, struggle in detection and characterisation. This coupled with the industrial desire to detect surface-breaking defects of components at the point of manufacture and/or maintenance, to increase design lifetime and further embed sustainability in their business models, is driving the increased adoption of Eddy Current Testing (ECT). Moreover, as businesses move toward Industry 4.0, demand for robotic delivery of NDT has grown. In this work, the authors present the novel implementation and use of a flexible robotic cell to deliver an eddy current array to inspect stress corrosion cracking on a nuclear canister made from 1.4404 stainless steel. Three 180-degree scans at different heights on one side of the canister were performed, and the acquired impedance data were vertically stitched together to show the full extent of the cracking. Axial and transversal datasets, corresponding to the transmit/receive coil configurations of the array elements, were simultaneously acquired at transmission frequencies 250, 300, 400, and 450 kHz and allowed for the generation of several impedance C-scan images. The variation in the lift-off of the eddy current array was innovatively minimised through the use of a force-torque sensor, a padded flexible ECT array and a PI control system. Through the use of bespoke software, the impedance data were logged in real-time (≤7 ms), displayed to the user, saved to a binary file, and flexibly post-processed via phase-rotation and mixing of the impedance data of different frequency and coil configuration channels. Phase rotation alone demonstrated an average increase in Signal to Noise Ratio (SNR) of 4.53 decibels across all datasets acquired, while a selective sum and average mixing technique was shown to increase the SNR by an average of 1.19 decibels. The results show how robotic delivery of eddy current arrays, and innovative post-processing, can allow for repeatable and flexible surface inspection, suitable for the challenges faced in many quality-focused industries.

A review of ultrasonic sensing and machine learning methods to monitor industrial processes

Supervised machine learning techniques are increasingly being combined with ultrasonic sensor measurements owing to their strong performance. These techniques also offer advantages over calibration procedures of more complex fitting, improved generalisation, reduced development time, ability for continuous retraining, and the correlation of sensor data to important process information. However, their implementation requires expertise to extract and select appropriate features from the sensor measurements as model inputs, select the type of machine learning algorithm to use, and find a suitable set of model hyperparameters. The aim of this article is to facilitate implementation of machine learning techniques in combination with ultrasonic measurements for in-line and on-line monitoring of industrial processes and other similar applications. The article first reviews the use of ultrasonic sensors for monitoring processes, before reviewing the combination of ultrasonic measurements and machine learning. We include literature from other sectors such as structural health monitoring. This review covers feature extraction, feature selection, algorithm choice, hyperparameter selection, data augmentation, domain adaptation, semi-supervised learning and machine learning interpretability. Finally, recommendations for applying machine learning to the reviewed processes are made.

Uncertainty Quantification for Deep Learning in Ultrasonic Crack Characterization

Deep learning for nondestructive evaluation (NDE) has received a lot of attention in recent years for its potential ability to provide human level data analysis. However, little research into quantifying the uncertainty of its predictions has been done. Uncertainty quantification (UQ) is essential for qualifying NDE inspections and building trust in their predictions. Therefore, this article aims to demonstrate how UQ can best be achieved for deep learning in the context of crack sizing for inline pipe inspection. A convolutional neural network architecture is used to size surface breaking defects from plane wave imaging (PWI) images with two modern UQ methods: deep ensembles and Monte Carlo dropout. The network is trained using PWI images of surface breaking defects simulated with a hybrid finite element / ray-based model. Successful UQ is judged by calibration and anomaly detection, which refer to whether in-domain model error is proportional to uncertainty and if out of training domain data is assigned high uncertainty. Calibration is tested using simulated and experimental images of surface breaking cracks, while anomaly detection is tested using experimental side-drilled holes and simulated embedded cracks. Monte Carlo dropout demonstrates poor uncertainty quantification with little separation between in and out-of-distribution data and a weak linear fit ( R=0.84 ) between experimental root-mean-square-error and uncertainty. Deep ensembles improve upon Monte Carlo dropout in both calibration ( R=0.95 ) and anomaly detection. Adding spectral normalization and residual connections to deep ensembles slightly improves calibration ( R=0.98 ) and significantly improves the reliability of assigning high uncertainty to out-of-distribution samples.

Recent Advances in Machine Learning Applied to Ultrasound Imaging

Machine learning (ML) methods are pervading an increasing number of fields of application because of their capacity to effectively solve a wide variety of challenging problems. The employment of ML techniques in ultrasound imaging applications started several years ago but the scientific interest in this issue has increased exponentially in the last few years. The present work reviews the most recent (2019 onwards) implementations of machine learning techniques for two of the most popular ultrasound imaging fields, medical diagnostics and non-destructive evaluation. The former, which covers the major part of the review, was analyzed by classifying studies according to the human organ investigated and the methodology (e.g., detection, segmentation, and/or classification) adopted, while for the latter, some solutions to the detection/classification of material defects or particular patterns are reported. Finally, the main merits of machine learning that emerged from the study analysis are summarized and discussed.

Artificial Intelligence, Machine Learning and Smart Technologies for Nondestructive Evaluation

Nondestructive evaluation (NDE) techniques are used in many industries to evaluate the properties of components and inspect for flaws and anomalies in structures without altering the part’s integrity or causing damage to the component being tested. This includes monitoring materials’ condition (Material State Awareness (MSA)) and health of structures (Structural Health Monitoring (SHM)). NDE techniques are highly valuable tools to help prevent potential losses and hazards arising from the failure of a component while saving time and cost by not compromising its future usage. On the other hand, Artificial Intelligence (AI) and Machine Learning (ML) techniques are useful tools which can help automating data collection and analyses, providing new insights, and potentially improving detection performance in a quick and low effort manner with great cost savings. This paper presents a survey on state of the art AI-ML techniques for NDE and the application of related smart technologies including Machine Vision (MV) and Digital Twins in NDE.

Exploring 3D elastic-wave scattering at interfaces using high-resolution phased-array system

The elastic-wave scattering at interfaces, such as cracks, is essential for nondestructive inspections, and hence, understanding the phenomenon is crucial. However, the elastic-wave scattering at cracks is very complex in three dimensions since microscopic asperities of crack faces can be multiple scattering sources. We propose a method for exploring 3D elastic-wave scattering based on our previously developed high-resolution 3D phased-array system, the piezoelectric and laser ultrasonic system (PLUS). We describe the principle of PLUS, which combines a piezoelectric transmitter and a 2D mechanical scan of a laser Doppler vibrometer, enabling us to resolve a crack into a collection of scattring sources. Subsequently, we show how the 3D elastic-wave scattering in the vicinity of each response can be extracted. Here, we experimentally applied PLUS to a fatigue-crack specimen. We found that diverse 3D elastic-wave scattering occurred in a manner depending on the responses within the fatigue crack. This is significant because access to such information will be useful for optimizing inspection conditions, designing ultrasonic measurement systems, and characterizing cracks. More importantly, the described methodology is very general and can be applied to not only metals but also other materials such as composites, concrete, and rocks, leading to progress in many fields.

Training neural networks on domain randomized simulations for ultrasonic inspection

To overcome the data scarcity problem of machine learning for nondestructive testing, data augmentation is a commonly used strategy. We propose a method to enable training of neural networks exclusively on simulated data. Simulations not only provide a scalable way to generate and access training data, but also make it possible to cover edge cases which rarely appear in the real world. However, simulating data acquired from complex nondestructive testing methods is still a challenging task. Due to necessary simplifications and a limited accuracy of parameter identification, statistical models trained solely on simulated data often generalize poorly to the real world. Some effort has been made in the field to adapt pre-trained classifiers with a small set of real world data. A different approach for bridging the reality gap is domain randomization which was recently very successfully applied in different fields of autonomous robotics. In this study, we apply this approach for ultrasonic testing of carbon-fiber-reinforced plastics. Phased array captures of virtual specimens are simulated by approximating sound propagation via ray tracing. In addition to a variation of the geometric model of the specimen and its defects, we vary simulation parameters. Results indicate that this approach allows a generalization to the real world without applying any domain adaptation. Further, the trained network distinguishes correctly between ghost artifacts and defects. Although this study is tailored towards evaluation of ultrasound phased array captures, the presented approach generalizes to other nondestructive testing methods.

Domain Adapted Deep-Learning for Improved Ultrasonic Crack Characterization Using Limited Experimental Data

Deep learning is an effective method for ultrasonic crack characterization due to its high level of automation and accuracy. Simulating the training set has been shown to be an effective method of circumventing the lack of experimental data common to nondestructive evaluation (NDE) applications. However, a simulation can neither be completely accurate nor capture all variability present in the real inspection. This means that the experimental and simulated data will be from different (but related) distributions, leading to inaccuracy when a deep learning algorithm trained on simulated data is applied to experimental measurements. This article aims to tackle this problem through the use of domain adaptation (DA). A convolutional neural network (CNN) is used to predict the depth of surface-breaking defects, with in-line pipe inspection as the targeted application. Three DA methods across varying sizes of experimental training data are compared to two non-DA methods as a baseline. The performance of the methods tested is evaluated by sizing 15 experimental notches of length (1-5 mm) and inclined at angles of up to 20° from the vertical. Experimental training sets are formed with between 1 and 15 notches. Of the DA methods investigated, an adversarial approach is found to be the most effective way to use the limited experimental training data. With this method, and only three notches, the resulting network gives a root-mean-square error (RMSE) in sizing of 0.5 ± 0.037 mm, whereas with only experimental data the RMSE is 1.5 ± 0.13 mm and with only simulated data it is 0.64 ± 0.044 mm.

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.

Using Deep Learning to Automate the Detection of Flaws in Nuclear Fuel Channel UT Scans

Nuclear reactor inspections are critical to ensure the safety and reliability of a nuclear facility's operation. In Canada, ultrasonic testing (UT) is used to inspect the health of pressure tubes that are part of Canada's Deuterium Uranium (CANDU) reactor's fuel channels. Currently, analysis of UT scans is performed by manual visualization and measurement to locate, characterize, and disposition flaws. Therefore, there is motivation to develop an automated method that is fast and accurate. In this article, a proof of concept (PoC) that automates the detection of flaws in nuclear fuel channel UT scans using a convolutional neural network (CNN) is presented. The CNN model was trained after constructing a dataset using historical UT scans and the corresponding inspection results. The requirement for this prototype was to identify the location of at least a portion of each flaw in UT scans while minimizing false positives (FPs). The proposed CNN model achieves this target by automatically identifying at least a portion of each flaw where further manual analysis is performed to identify the width, the length, and the type of the flaw.

Towards Explainable Augmented Intelligence (AI) for Crack Characterization

Crack characterization is one of the central tasks of NDT&E (the Non-destructive Testing and Evaluation) of industrial components and structures. These days data necessary for carrying out this task are often collected using ultrasonic phased arrays. Many ultrasonic phased array inspections are automated but interpretation of the data they produce is not. This paper offers an approach to designing an explainable AI (Augmented Intelligence) to meet this challenge. It describes a C code called AutoNDE, which comprises a signal-processing module based on a modified total focusing method that creates a sequence of two-dimensional images of an evaluated specimen; an image-processing module, which filters and enhances these images; and an explainable AI module—a decision tree, which selects images of possible cracks, groups those of them that appear to represent the same crack and produces for each group a possible inspection report for perusal by a human inspector. AutoNDE has been trained on 16 datasets collected in a laboratory by imaging steel specimens with large smooth planar notches, both embedded and surface-breaking. It has been tested on two other similar datasets. The paper presents results of this training and testing and describes in detail an approach to dealing with the main source of error in ultrasonic data—undulations in the specimens’ surfaces.

Ultrasonic Defect Characterization Using the Scattering Matrix: A Performance Comparison Study of Bayesian Inversion and Machine Learning Schemas

Accurate defect characterization is desirable in the ultrasonic nondestructive evaluation as it can provide quantitative information about the defect type and geometry. For defect characterization using ultrasonic arrays, high-resolution images can provide the size and type information if a defect is relatively large. However, the performance of image-based characterization becomes poor for small defects that are comparable to the wavelength. An alternative approach is to extract the far-field scattering coefficient matrix from the array data and use it for characterization. Defect characterization can be performed based on a scattering matrix database that consists of the scattering matrices of idealized defects with varying parameters. In this article, the problem of characterizing small surface-breaking notches is studied using two different approaches. The first approach is based on the introduction of a general coherent noise model, and it performs characterization within the Bayesian framework. The second approach relies on a supervised machine learning (ML) schema based on a scattering matrix database, which is used as the training set to fit the ML model exploited for the characterization task. It is shown that convolutional neural networks (CNNs) can achieve the best characterization accuracy among the considered ML approaches, and they give similar characterization uncertainty to that of the Bayesian approach if a notch is favorably oriented. The performance of both approaches varied for unfavorably oriented notches, and the ML approach tends to give results with higher variance and lower biases.

Automatic Quantification of Subsurface Defects by Analyzing Laser Ultrasonic Signals Using Convolutional Neural Networks and Wavelet Transform

The conventional machine learning algorithm for analyzing ultrasonic signals to detect structural defects necessarily identifies and extracts either time- or frequency-domain features manually, which has problems in reliability and effectiveness. This work proposes a novel approach by combining convolutional neural networks (CNNs) and wavelet transform to analyze the laser-generated ultrasonic signals for detecting the width of subsurface defects accurately. The novelty of this work is to convert the laser ultrasonic signals into the scalograms (images) via wavelet transform, which are subsequently utilized as the image input for the pretrained CNN to extract the defect features automatically to quantify the width of defects, avoiding the necessity and inaccuracy induced by artificial feature selection. The experimentally validated numerical model that simulates the interaction of laser-generated ultrasonic waves with subsurface defects is first established, which is further utilized to generate adequate laser ultrasonic signals for training the CNN model. A total number of 3104 data are obtained from simulation and experiments, with 2480 simulated signals for training the CNN model and the remaining 620 simulated data together with 4 experimental signals for verifying the performance of the proposed algorithm. This approach achieves the prediction accuracy of 98.5% on validation set, particularly with the prediction accuracy of 100% for the four experimental data. This work proves the feasibility and reliability of the proposed method for quantifying the width of subsurface defects and can be further expanded as a universal approach to various other defects detection, such as defect locations and shapes.

Automated Defect Detection From Ultrasonic Images Using Deep Learning

Nondestructive evaluation (NDE) is a set of techniques used for material inspection and defect detection without causing damage to the inspected component. One of the commonly used nondestructive techniques is called ultrasonic inspection. The acquisition of ultrasonic data was mostly automated in recent years, but the analysis of the collected data is still performed manually. This process is thus very expensive, inconsistent, and prone to human errors. An automated system would significantly increase the efficiency of analysis, but the methods presented so far fail to generalize well on new cases and are not used in real-life inspection. Many of the similar data analysis problems were recently tackled by deep learning methods. This approach outperforms classical methods but requires lots of training data, which is difficult to obtain in the NDE domain. In this work, we train a deep learning architecture EfficientDet to automatically detect defects from ultrasonic images. We showed how some of the hyperparameters can be tweaked in order to improve the detection of defects with extreme aspect ratios that are common in ultrasonic images. The proposed object detector was trained on the largest dataset of ultrasonic images that was so far seen in the literature. In order to collect the dataset, six steel blocks containing 68 defects were scanned with a phased-array probe. More than 4000 VC-B-scans were acquired and used for training and evaluation of EfficientDet. The proposed model achieved 89.6% of mean average precision (mAP) during fivefold cross validation, which is a significant improvement compared to some similar methods that were previously used for this task. A detailed performance overview for each of the folds revealed that EfficientDet-D0 successfully detects all of the defects present in the inspected material.

Crack Detection in Images of Masonry Using CNNs

While there is a significant body of research on crack detection by computer vision methods in concrete and asphalt, less attention has been given to masonry. We train a convolutional neural network (CNN) on images of brick walls built in a laboratory environment and test its ability to detect cracks in images of brick-and-mortar structures both in the laboratory and on real-world images taken from the internet. We also compare the performance of the CNN to a variety of simpler classifiers operating on handcrafted features. We find that the CNN performed better on the domain adaptation from laboratory to real-world images than these simple models. However, we also find that performance is significantly better in performing the reverse domain adaptation task, where the simple classifiers are trained on real-world images and tested on the laboratory images. This work demonstrates the ability to detect cracks in images of masonry using a variety of machine learning methods and provides guidance for improving the reliability of such models when performing domain adaptation for crack detection in masonry.