This paper presents a hybrid deep learning model that combines Convolutional Neural Network (CNN) and Transformer to enable efficient prediction of far-field scattered signals of S0 mode Lamb waves from defects of thin plates. The proposed model combines CNN for local spatial feature extraction with Transformer to model global temporal dependencies, enhancing the ability to predict scattering from irregularly-shaped defects beyond the limitations of traditional methods. A three-dimensional (3D) finite element model of an aluminum plate with irregularly-shaped defects was developed to generate scattering fields with diverse morphologies and parameters for model training and testing. CNN-Transformer model successfully predicted the scattering behavior of S0 mode Lamb wave, demonstrating high accuracy in scenarios with irregularly-shaped defects. The model's performance was further validated through laser Doppler experiments, demonstrating strong consistency with the predicted scattering characteristics. Furthermore, the model was extended to solve the scattering matrix, enabling accurate prediction of scattered signals across multiple incident angles. This study introduces a new approach to defect scattering in ultrasonic guided wave detection. It provides both theoretical insights and practical support for engineering applications.

Ultrasonic guided waves are widely used for structural health monitoring, while traditional stress detection methods based on weak nonlinear elasticity theory suffer from limited sensitivity. This study presents a numerical investigation using the highly sensitive Sideband Peak Count-index (SPC-I) technique for improved stress assessment in plates. A finite element (FE) model is developed to analyze the transient evolution of higher-order harmonics under various uniaxial stress states. This study explores the influence of both stress magnitude and its orientation relative to the wave propagation direction, establishing a quantitative link to the acoustic nonlinear parameter, β. The results demonstrate that SPC-I is a robust indicator, sensitive not only to the stress magnitude but also to its orientation. Notably, the proposed method significantly enhances measurement sensitivity. Experimental validation confirms that SPC-I values exhibit a pronounced change with stress variations, representing a marked improvement over conventional ultrasonic techniques. The findings establish a theoretical framework for ultrasonic stress detection and provide essential technical guidance for structural health monitoring (SHM) applications.

Significance Fringe projection profilometry (FPP) is a high-precision optical measurement technology characterized by its rapid acquisition speed and non-contact nature. This technique has found extensive applications in various domains including three-dimensional (3D) measurement, system calibration, and range image segmentation. Within the FPP framework, the analysis of deformed fringe patterns modulated by object surface morphology enables the extraction of corresponding phase information. The phase map's high-frequency components encompass both phase singular features-resulting from the modulation of object geometry and mathematical properties-and environmental noise-induced phase artifacts. According to its formation mechanism, phase singularity features can be systematically classified into two different types: type I is a regular phase discontinuity caused by the inherent discontinuity of the tangent function; type II is an irregular phase jump caused by a sudden change in the surface morphology of an object Current research predominantly focuses on type I phase singular features, leaving type II features relatively underexplored. To address this gap and further investigate the characteristics and potential applications of phase singular features, this paper presents a comprehensive review specifically dedicated to these phenomena. We systematically examine the generation mechanisms, current research progress, and prospective application scenarios for different classes of phase singular features. Progress This paper presents a comprehensive review of phase singular features in FPP, focusing on their generation mechanisms, characteristic properties, and practical applications. The study systematically classifies these features into two distinct types based on their formation mechanisms: type I features resulting from the inherent discontinuity of the arc-tangent function, and type II features arising from abrupt surface topography variations. The extraction methodologies for these features are illustrated in Fig. 2, while their comparative characteristics are summarized in Table 2. For type I phase singular features, which exhibit distinct properties, including distribution regularity, depth variation trend characterization, and phase shift variability, we examine their crucial applications in three key areas: 1) phase unwrapping, where their characteristic pi to-pi jumps enable robust absolute phase recovery through various unwrapping approaches (spatial, temporal, geometric constraint, and deep learning-based methods, as detailed in Table 3); 2) system calibration, leveraging their anti-blur robustness derived from mathematical properties (Fig. 9); and 3) camouflage detection, utilizing their feature saliency (Fig. 11). Regarding type II phase singular features, characterized by phase shift invariance and high-frequency correlation with 3D point cloud data, we discuss their emerging applications in computational range image segmentation (Fig. 12 and Table 7), and phase error correction (Fig. 15). These features demonstrate strong dependence on surface gradient variations and measurement system sensitivity. The review concludes with prospective research directions for phase singular feature utilization in FPP, including the development of unified theoretical frameworks, real-time processing algorithms, and multimodal measurement integration. This systematic analysis aims to provide valuable insights for advancing high-precision 3D measurement technologies based on fringe projection principles. Conclusions and Prospects As a key bridge connecting optical measurement and data processing, the research on phase singular features in fringe projection profilometry has gradually expanded from the early phase unwrapping algorithm focusing on type I phase has singular features to the geometrical characterization and application innovation of type II phase singular features. With the continuous deepening of phase singular features research, current research has broken through the traditional "feature removal" thinking limitations, and began to explore the potential value of phase singular features in the data acquisition, analysis, especially in the field of ultra-high-speed measurements, micro-deformation measurements and micro-vibration analysis. Future development directions include the following three perspectives: intelligent algorithm-driven feature decoupling, multimodal fusion and hardware co-measurement, and interdisciplinary application outreach.

This short communication proposes a joint wavenumber- and spatial-domain compressed sensing (JWSCS) framework for guided wavefield reconstruction. Existing sparse reconstruction methods typically operate in either the wavenumber domain or the spatial domain, each exploiting only one type of sparsity prior and exhibiting a lack of robustness against different damage scales. The proposed JWSCS framework simultaneously leverages wavenumber-domain modal sparsity and spatial-domain damage-induced sparsity. A joint sparse representation model is first established to integrate the spatial and wavenumber characteristics of the wavefield signals. A unified compressed sensing (CS) problem is then formulated, where a weighting coefficient is introduced to balance the sparsity constraints between the two domains. The non-convex problem is relaxed into a weighted ℓ1-minimization problem and efficiently solved via convex optimization. Simulation and experimental results validate that the proposed JWSCS framework enables more accurate wavefield reconstruction with significantly fewer measurements than conventional single-domain CS methods. On real measured data at 70% sample compressive ratio, JWSCS achieves an average Pearson correlation coefficient of 0.88 ± 0.02, improving by approximately 7% and 17% compared with wavenumber-domain and spatial-domain CS, respectively.

Delamination quantification using guided wavefields has gained attention due to the high sensitivity of ultrasonic guided waves and advances in wavefield scanning. Although supervised deep learning enhances wavefield analysis, its application is limited by the high cost of large-scale labeling. Semi-supervised approaches alleviate this issue but still depend on curated, defect-free reference data. To overcome these challenges, this study proposes an unsupervised framework that integrates Convolutional Autoencoders (CAE) with guided wavefields, requiring neither labels nor historical data. The framework enables delamination quantification from a single wavefield snapshot captured when direct waves fully cover the inspection region. It leverages neural networks’ tendency to overfit dominant healthy patterns, causing poor reconstruction in delaminated areas. The snapshot is divided into uniform patches and symmetrically flipped to enhance data diversity. Each patch is independently reconstructed by the CAE, and a pixel-wise reconstruction error map highlights delamination regions as high-error zone. A two-stage training strategy further reinforces healthy-region reconstruction. Experiments on scanning laser Doppler vibrometer data validate the effectiveness of the augmentation scheme, the two-stage strategy, and the framework’s robustness to varying defect shapes and locations. Overall, the proposed method provides a practical and robust solution for wavefield-based delamination detection without pre-collected or labeled data.

To address low data acquisition efficiency in full-wavefield data collection for guided-wave-based structural damage detection, the limited adaptability of conventional single-domain sparse reconstruction methods, and insufficient imaging accuracy, this paper proposes a damage detection framework that integrates joint wavenumber-space domain sparse reconstruction with damage-sensitive mode gradient enhancement (JWSSR-SMGE). A joint wavenumber-space domain sparse reconstruction (JWSSR) model is established, which incorporates two weighting strategies, namely offline-calibrated ADMM (OC-ADMM) and iteratively adaptive weighted ADMM (IAW-ADMM). By simultaneously exploiting the modal sparsity in the wavenumber domain and the damage-induced sparsity in the spatial domain, the proposed model achieves high-accuracy and robust reconstruction under various damage characteristics. To address the issues of noise amplification and weak-component attenuation in the reconstructed signals, a damage-sensitive mode gradient enhancement (SMGE) imaging method is further designed. This method employs annular filtering and gradient-based weighting strategies to achieve balanced enhancement of damage-related responses across different wavenumber components, thereby effectively suppressing background noise and highlighting damage features. Both numerical simulations and experimental results demonstrate that under sparse sampling conditions with a sampling compression ratio as low as 30%, the total inspection time is reduced by more than 58% compared with full sampling. The reconstruction accuracy of the proposed method is improved by 6%--21% over conventional single-domain approaches. The imaging error of the SMGE-based method is generally lower than that of the conventional sensitive mode enhancement (SME) method, thereby achieving integrated detection from high-efficiency sparse measurement to high-precision damage imaging.

In non-destructive testing applications, conventional ultrasonic nondestructive testing methods still face significant challenges in the effective feature extraction and high-contrast imaging of micro-defects. This is primarily due to the severe geometric attenuation and material absorption of laser-generated ultrasonic waves during propagation, as well as the strong background noise induced by random energy fluctuations in the laser excitation pulses. To address these issues, this paper proposes a multidimensional laser-ultrasonic C-scan imaging algorithm that synergistically integrates spatial compensation with Gaussian Mixture Model (GMM)-driven adaptive thresholding. The proposed method incorporates an adaptive compensation mechanism that accounts for the spatial evolution of three acoustic features: energy, cross-correlation coefficient, and time delay. Furthermore, it employs GMM-based statistical discrimination alongside multi-feature weighted voting to cross-validate the detection results across independent channels. The Experimental results demonstrate the effectiveness of the algorithm in precisely identifying both surface and internal micro-cracks with dimensions of 1 mm × 1 mm × 10 mm. Compared to conventional amplitude imaging methods relying on hard-threshold segmentation, the proposed algorithm boosts Precision by more than five times and increases both F1-score and IoU by more than three times, thereby significantly enhancing imaging quality and reliability.

Fingerprint authentication, as a mainstream biometric method, has become an indispensable component in consumer electronics and public safety. However, optical and capacitive scans are susceptible to deception by fake textures and interference in environmental contaminations. The scanning domain of ultrasonic methods is constrained by under-screen pulse-echo devices. Here we report a biometric texture imaging and authentication method using an ultrasonic array that occupies no space under the screen, with fingerprint authentication as an example. The method characterizes fingerprints with sub-millimeter minutiae by elaborating sound-field variations within the array induced by fingertip press. Although the sound field propagates along the plane, it can be used to reconstruct the 3D information of fingerprint ridges. Anti-counterfeit and highly robustness enables excellent texture imaging and authentication despite water or dirt on the fingertips. Demonstrations of any position on the developed fingerprint acquisition device prove that our method promises to advance large-screen identity authentication.

The barely visible damage in aircraft composite structures can be a significant concern, making impact monitoring crucial for ensuring their integrity and reliability. This paper proposes a composite impact localization method based on artificial neural networks and Bayesian updating, utilizing time-difference feature vectors from impact events. The method determines the impact location of composite materials with unknown material properties or complex characteristics. This study uses finite element simulation data and experimental data as training sets, extracts feature vectors from the impact signal obtained by the sensor, and then compares and constructs the best neural network model for localization training. Finally, the Bayesian updating algorithm is used to enhance the accuracy and reliability of the location results. Specifically, the best average location error for Hsu-Nielsen source events is 0.429 cm, with a variance of 0.300 cm2, while for falling ball simulation events, the best average location error is 1.122 cm, with a variance of 0.954 cm2.

When an object experiences external impacts or internal damage such as cracks, it generates a transient elastic wave known as an acoustic emission (AE) signal. According to the characteristics of AE signals, they can be classified into burst and continuous signals. Current acoustic localization methods mostly target single-type signals. When both types of signals coexist, they interfere with each other, increasing the difficulty of localization. To address the impact of signal aliasing on localization, this article introduces a beamforming localization method with dispersion curve compensation. First, the signal types are identified by using kurtosis thresholds. When both types exist simultaneously, an adaptive energy threshold method is employed to determine the arrival and end times of burst signals, separating them from continuous signals to reconstruct the signals. Post reconstruction, filter bands are selected based on energy distribution, and multiple beamforming is conducted on the filtered acoustic signals with a velocity change of 10 m/s, combining energies to achieve source localization. Experimental results demonstrate that when both types coexist, the maximum localization error for burst sources is 3.6 cm with an average error of 2.04 cm, while for continuous sources, the maximum localization error is 2 cm with an average error of 1.32 cm. This shows that the beamforming localization method with dispersion curves compensation can effectively localize both burst-type and continuous-type acoustic sources simultaneously.

To avoid catastrophic structural failures resulting from the hidden accumulation of delamination in carbon-fiber reinforced polymer, timely and accurate detection is crucial. Several deep learning-based methods have been developed for full wavefield segmentation to image delamination. However, the cost of experimentally acquiring extensive damage scenarios for dataset construction is prohibitively high. In addition, accurately segmenting delamination remains challenging due to the complex superposition of guided wave components in the full wavefield. This paper proposes a data preprocessing strategy that combines wavenumber filtering with a hybrid noise-flipping augmentation, to enhance the performance of deep learning models in full wavefield segmentation for delamination imaging. By isolating the derived guided wave modes introduced by delamination in the frequency domain, the deep learning models are guided to concentrate more effectively on delamination-relevant features. Noise augmentation and flipping augmentation are employed to improve the generalization of the delamination imaging models, enabling them to better handle real-world measurement conditions, which may include various external interference factors such as structural vibration and transducer noise. Seven distinct deep learning models were employed and evaluated in both simulated and experimental settings to examine the effectiveness of the data preprocessing strategy. The results demonstrate that models trained solely on simulated data can be effectively applied to experimental measurements, achieving a highest intersection over union score of 0.8634 and producing artifact-free delamination imaging.

The branched topology of thermal pipeline networks creates multiple propagation paths for leak-induced negative pressure waves (NPWs), causing leak localization algorithms to potentially output “multiple matching solutions,” resulting in erroneous localization. To address this challenge, this paper proposes a novel method that combines shortest path planning (SPP) and Monte Carlo tree search (MCTS) to optimize pressure sensor deployment. Unlike conventional approaches relying on historical network information or simulation software, this method optimizes sensor placement based on actual NPW transmission paths. First, the method discretizes the pipeline network and employs the sum of inter-sensor shortest path lengths as the optimization objective. Then, it utilizes MCTS to iteratively update sensor deployment schemes, ultimately improving the uniqueness of leak localization results obtained through NPW arrival delay matching. In a 12 km×12 km network with 10 sensors, the optimization method increased the total SPP length from 26.1 km to 68.6 km. Across 1,000 simulated leak scenarios, points with unique NPW arrival delay signatures increased from 54.8 to 79.0

This study presents the development of a flexible ultrasonic transducer array with automatic phase calibration (FUT-APC) for high-resolution carotid artery imaging and continuous monitoring of vascular mechanical parameters. The transducer integrates element position sensing with real-time phase compensation, utilizing five flexible pressure sensors to reconstruct the curvature profile of the attached surface with a reconstruction error as low as 0.34  mm. The system maintains stable operation on complex curved surfaces, significantly improving imaging quality compared to conventional probes on bent or irregular geometries. A contact pressure visualization mechanism enables real-time monitoring of applied pressure, ensuring data consistency during long-term measurements. The FUT-APC has a center frequency of 4.6  MHz, a -6 dB focal width of 0.74  mm, and a bandwidth of 53 %. In phantom tests, the system achieved axial resolution better than 0.75  mm and lateral resolution better than 0.93  mm. In vivo testing successfully captured the dynamic diameter variations of the carotid arterial wall and, together with wall thickness measurements, enabled extraction of key mechanical parameters, including circumferential stress (83.1  kPa), strain (7.09 %), and static elastic modulus (1.17  MPa). The FUT-APC offers a wearable and accurate solution for early screening, risk assessment, and dynamic tracking of atherosclerosis.

In the long-range pipeline monitoring application of Phase-sensitive Optical Time Domain Reflectometer (q -OTDR), instantaneous high-energy vibration event, e.g., PIG, generates a pair of invert-V pattern in the spacetemporal graph (ST-graph), providing clear signature for event recognition. This paper proposed a sample learning free q -OTDR event recognition method, where a ST-graph is treated as a series of one-dimensional (1D) signals and a grayscale image separately in the preprocessing step and pattern recognition step. A homemade q -OTDR system is employed as the distributed sensor to collect real-field space-temporal graph data containing invert-V pattern. In the preprocessing step, the Stockwell transform is applied to each 1D temporal signal as filter, and normalization is then performed to mitigate the effects of interference fading. In the pattern recognition step, a rho-summation Hough transform is proposed to detect the two edges of invert-V pattern, considering that lines in ST-graph possess a certain width. The invert-V pattern is recognized by detecting pattern shaped like "X", the lower half of which is extracted as the recognition result. Experimental result proves the feasibility of the proposed method of accurately capturing small invert-V patterns in the space-temporal graph.

Lamb wave scattering offers valuable insights into material properties, defect characteristics, and wave propagation behaviors, making it a prominent research topic in the fields of non-destructive testing and structural health monitoring. Analytical methods, while offering theoretical analysis, are often limited to simple defect geometries and homogeneous media. Numerical methods such as the boundary element method and hybrid techniques can handle complex structures but suffer from computational costs, particularly at high frequencies due to the need for fine discretization. To overcome these challenges, this study proposes a physics-enhanced TransUNet (PTUNet) for solving the scattered wavefield of S0 Lamb waves in plate structures with irregular defects. PTUNet combines the local feature extraction capability with the global modeling capacity while employing finite difference approach to incorporate the Kirchhoff-Love plate theory as a physical constraint. The effectiveness of the proposed method is validated through numerical simulations on random defects and experimental measurements using a scanning laser Doppler vibrometer on a 1.5 mm-thick aluminum plate. The results demonstrate that PTUNet accurately predicts wave propagation, mode conversion, and intricate scattering characteristics, achieving reasonable agreement with both finite element simulations and experimental observations. PTUNet presents a promising approach for wavefield modeling and defect characterization, with potential applications in acoustic scattering problems.

This study focuses on the localization of delamination damage in composite materials. Due to the attenuation of the air-coupled ultrasound at the gas-solid coupling interface, the signal-to-noise ratio of the air-coupled ultrasonic signal is low, which results in low accuracy of damage detection. This paper proposes a method to locate delamination damage in composite materials. This method uses non-orthogonal Air-coupled Ultrasonic Guided Waves for B-scan; an artifact-reduction damage index (ARDI) is designed with the proportion of the scan paths containing damage information to the entire scan paths. These two exquisite designs improve detection accuracy. Finally, the probability of damage occurrence is calculated, and the detection of the damage location is achieved, with the positioning error of the damage being less than 1.5%.

Guided wavefields encode rich information on wave-damage interactions, making them widely utilized for damage mapping. While deep learning-based supervised approaches have recently been developed for guided wavefield analysis, their practical application is limited by the high cost of data acquisition, potential labeling errors, and poor transferability-models trained for specific regions often require retraining when applied to new inspection areas. Semi-supervised approaches offer a promising alternative, relying solely on signals collected from healthy structures for model training. However, existing semi-supervised frameworks primarily utilize pitchcatch signals, enabling only qualitative detection or damage localization. To date, there's no semi-supervised framework that leverages guided wavefields to achieve damage imaging. Therefore, this study proposes a novel semi-supervised framework based on Convolutional Autoencoder (CAE) and guided wavefields for delamination imaging. This approach employs a CAE to learn intrinsic patterns of healthy wavefields, enabling detection of wavefield anomalies caused by delamination. Unlike conventional semi-supervised methods that rely on pre-collected historical data, the training data in this framework are directly extracted from a healthy portion of the inspected structure. Furthermore, this framework does not rely on long-term wavefield sequences, but operates solely on a single-frame snapshot captured when the direct waves fully cover the inspection region. To facilitate efficient learning, the wavefield snapshot is segmented into uniformly sized patches. Additionally, symmetric flipping-based augmentation was applied to the training data to further enhance model performance. In this framework, each patch within the inspection area is individually reconstructed using the CAE, and a pixel-wise reconstruction error map is computed, wherein regions influenced by delamination exhibit elevated errors. Finally, this error map is thresholded to identify wavefield anomalies and generate delamination images. The effectiveness of flipping-based augmentation, the influence of patch size and sliding stride used for patch sampling, as well as the model's noise robustness, are systematically investigated. To validate these findings, the proposed approach was experimentally evaluated using guided wavefield data acquired with a scanning laser Doppler vibrometer.

Understanding the dispersion characteristics of guided waves is crucial for the application of ultrasonic guided wave method in fluid-saturated porous media. In this paper, the Semi-Analytical Finite Element (SAFE) method is applied to analyze the dispersion characteristics in fluid-saturated porous media. We derive the SAFE equations for fluid-saturated porous media and reformulate them into the standard Finite Element (FE) eigenvalue form that can be solved using COMSOL Multiphysics. Unlike the simple boundaries of elastic waveguides (such as free or rigid boundaries), the boundary conditions of fluid-saturated porous media present an additional challenge in the analysis of guided wave problems. We also derive the open-pore and closed-pore boundary conditions for fluid-saturated porous media and reformulate them into standard FE form. On this basis, we analyzed the causes of high-attenuation modes, wave structures, energy distribution, and the effects of surface treatment methods. The proposed method is first validated on waveguides with regular cross sections. Furthermore, we apply the SAFE method to L-shaped fluid-saturated porous bar for which there is no analytical solution, demonstrating the unique advantages of the SAFE method in solving the dispersion characteristics of irregular fluid-saturated porous media.

The timely detection of delamination is essential for preventing catastrophic failures and extending the service life of carbon fiber-reinforced polymers (CFRP). Full wavefields in CFRP encapsulate extensive information on the interaction between guided waves and structural damage, making them a widely utilized tool for damage mapping. However, due to the multimodal and dispersive nature of guided waves, interpreting full wavefields remains a significant challenge. This study proposes an end-to-end delamination imaging approach based on UNet++ using 2D frequency domain spectra (FDS) derived from full wavefield data. The proposed method is validated through a self-constructed simulation dataset, experimental data collected using Scanning Laser Doppler Vibrometry, and a publicly available dataset created by Kudela and Ijjeh. The results on the simulated data show that UNet++, trained with multi-frequency FDS, can accurately predict the location, shape, and size of delamination while effectively handling frequency offsets and noise interference in the input FDS. Experimental results further indicate that the model, trained exclusively on simulated data, can be directly applied to real-world scenarios, delivering artifact-free delamination imaging.