“Flying-Spot” laser infrared thermography (FST) is a non-destructive testing technique capable of detecting small defects by scanning surfaces with a laser heat source. Defects such as cracks in metallic parts are revealed through disruptions in heat propagation, which are captured by an infrared camera. Combining this technique with visible-spectrum inspection, providing information about surface textures and geometries that are less discernible in the infrared domain, can enhance both the robustness and performance of defect detection. This work proposes the deployment of state-of-the-art visible-infrared fusion neural models for multi-spectrum crack detection on data combining visible and FST imaging. The limited amount of data available is mitigated through the deep image generation techniques for visible-FST image pairs, through the models Stable Diffusion and Control-Net. These synthetic data are used for the pre-training of multi-spectral object detection architectures for surface crack detection in metallic materials. Specifically, we assess the benefits of multi-spectral fusion compared to single-spectrum FST detection and evaluate the robustness of several state-of-the-art fusion models when handling image pairs with poor multi-spectral registration.
Water seepage accelerates tunnel lining deterioration, yet automated detection from mobile LiDAR remains challenging because low back-scattered intensity is shared by seepage, joints, and appendages. We propose a multi-stage, instance-level seepage detection framework for on-vehicle LiDAR point clouds. Intensity is normalized via angular-binning Z-score statistics, and non-seepage appendages are removed using point-wise curvature filtering followed by cross-sectional convex-hull refinement. The remaining low-intensity regions are unfolded into a cylindrical (θ,y) map and grouped into line and blob instances using prompt-guided candidate-region grouping with Segment Anything Model 3 (SAM3). Seepage is identified using a two-stage strategy: high-confidence seeds are obtained from an intensity-contrast metric, and ambiguous instances are classified by a lightweight SVM using instance-level features. On the 30 m development segment, evaluated within the denoised low-intensity candidate set using manually prepared seepage annotations guided by inspection photographs, the final prediction achieved an F1-score of 0.913. Application without retraining to an unseen segment of the same highway tunnel retained high recall but reduced precision, while the external shield-tunnel dataset was used only as a cross-domain stress test and showed a larger precision decrease.
Structural instability in water diversion canals is frequently precipitated by concealed leakage voids beneath concrete linings. Traditional electromagnetic inspection methods, such as ground-penetrating radar (GPR), are often severely constrained in these environments by the rapid attenuation of signals within conductive, water-saturated media. To address this limitation, this study presents a novel non-destructive testing approach utilizing a low-temperature superconducting quantum interference device (SQUID) second-order axial gradiometer. However, detecting non-magnetic voids is challenging because their responses arise from the weak magnetic-susceptibility contrast between the soil and air. Under the reference soil and geomagnetic-field conditions, hemispheroidal voids with opening diameters of 0.10–0.20m and depths of 0.02–0.12m produced predicted peak amplitudes of the second-order finite-difference anomalies ranging from 3.94×10−11 to 5.86×10−10T, all below 10−9T. Consequently, we establish a physics-based forward model for the second-order finite-difference response of the axial magnetic flux density to characterize the gradiometer signals from a hemispheroidal void beneath a lining. Subsequently, we propose a quantitative inversion algorithm combining modified Akima cubic Hermite (Makima) interpolation and an exhaustive global grid search to estimate void geometry using the peak anomaly amplitude and full width at half maximum (FWHM) of the second-order finite-difference response. Results from field experiment inversions demonstrate that, for each of the five controlled void configurations, the estimated opening diameter and depth reproduced the corresponding mean experimental feature pair obtained from 18 independent scans with a combined response misfit not exceeding 5%. This approach demonstrates the feasibility of SQUID-based quantitative detection and characterization of concealed voids beneath water-saturated canal linings in high-conductivity environments.
Guided-wave damage imaging in realistic composite aircraft structures is difficult because anisotropic propagation, multimodal dispersion, internal interfaces, and boundary reflections distort wave packets and generate coherent clutter. Delay correction alone may still accumulate non-damage components and produce high-amplitude pseudo-foci. This study proposes an anisotropy-aware phase–frequency coherence imaging (PFCI) method for damage localization in a locally planar region selected from an Airbus A380 composite flap track fairing. Direction-dependent dispersion information is extracted from experimental guided-wave measurements to reconstruct a continuous anisotropy-aware wavenumber model for propagation compensation. After compensation, PFCI evaluates each trial point through a joint packet-level coherence test combining envelope-phase alignment and local instantaneous-frequency consistency near the focusing instant. This criterion enhances damage-consistent responses and suppresses pseudo-indications caused by reflections, direct-wave residue, and imperfect focusing. Experiments using a piezoelectric transducer network and full-matrix guided-wave measurements show that, for the representative single-defect case at 60 kHz with a 5-cycle tone burst, PFCI achieved a localization error of 5.831 mm, a peak-to-background ratio of 13.507 dB, and a contrast-to-noise ratio of 6.956. In the dual-defect case, two defect-related indications were identified, with a mean localization error of 7.697 mm. The results show that PFCI provides an effective feature-level imaging criterion for guided-wave damage localization in the tested composite aircraft structure under a healthy-baseline condition.
Step heating thermography (SHT) has emerged as a promising technique for non-destructive evaluation for its simplicity and operational efficiency, however, its application for coating thickness measurement has yet to be fully explored. This paper introduces a time-offset thermal response method for coating thickness measurement using SHT. By applying a selected time offset in the normalized surface temperature response formular, a distinct peak shows up on the originally monotonically rising curve. The peak time exhibits a strong linear correlation with the coating thickness. A comprehensive analysis is derived for the relationship of the time offset and the peak. Validating experiments on anti-corrosion and anti-icing coatings confirm the highly linear correlations (R2 > 0.998). Comparison with conventional thermographic signal reconstruction (TSR) method demonstrates that the proposed method is computationally more efficient, better immune from the noise, and inherently reference-free. Moreover, its ability to determine the thickness at earlier stage enables thinner coating detections. Therefore, the new method is more promising for real-world deployment.
Phased array ultrasonic imaging of coarse-grained 800HT alloy welds remains challenging, as pronounced acoustic anisotropy induces beam skewing, phase distortion, and strong structural noise, degrading imaging quality and defect localization accuracy. In this study, an anisotropy-corrected imaging framework is developed for ultrasonic inspection of coarse-grained welds by combining microstructural modeling, inspection parameter analysis, coherence-based imaging, and ray-tracing-based time-of-flight (TOF) correction. A two-dimensional finite element model was established on the basis of grain characteristics derived from Electron Backscatter Diffraction (EBSD), Voronoi-based geometric reconstruction, and stiffness matrix assignment using Bond transformation, and its validity was examined by comparison with experimental back-wall echo signals in both the time and frequency domains. Within the investigated conditions, frequencies of 4-5 MHz and 16 active elements provided the most favorable imaging performance. Among the coherence-based imaging methods considered, the delay multiply and sum with coherence factor (DMAS-CF) method produced the highest signal-to-noise ratio (SNR), although noticeable defect localization errors remained. To improve localization performance, a Fermat-principle-based spatial ray tracing method was introduced, and a slowness-conservation fast ray tracing (SCFRT) algorithm was further developed to improve computational efficiency. Unlike conventional imaging methods that rely on a priori sound velocity assumptions, the proposed framework directly incorporates microstructure-informed anisotropic corrections into the imaging workflow. The combined SCFRT-DMAS method improved defect localization while maintaining favorable imaging quality, and achieved a total end-to-end imaging time of about one second for experimental full matrix capture (FMC) data, indicating its potential for efficient near-real-time inspection of coarse-grained welds.
Pipeline networks, critical for energy transportation, are highly susceptible to damage from external impacts. To ensure efficient and cost-effective maintenance, it is essential to develop a rapid and accurate method for impact localization. Generally, sensors used in monitoring systems should be lightweight so as to preserve the structural integrity of the pipeline and maintain high reliability. However, patch-type sensors may exhibit limited conformity with curved pipelines, while direct-written sensor routes require process adaptation for in-situ deposition on cylindrical surfaces. Herein, an airbrush-based direct-writing method is proposed for the in-situ fabrication of a piezoelectric polymer sensor array on pipeline surfaces, enabling detection of impact-induced guided waves. A 3 × 2 sparse array comprising poly(vinylidene fluoride-co-trifluoroethylene) (P(VDF/TrFE))-based sensors is deposited via airbrush spraying, followed by in-situ annealing and corona poling to enhance piezoelectric performance. For impact source localization, the effective frequency band is first extracted using continuous wavelet transform (CWT). Then, a frequency-domain Multiple Signal Classification (MUSIC) method combined with an adaptive scan over trial propagation speeds is employed to iteratively focus the spatial spectrum. Guided by group-velocity dispersion analysis, the scanning interval is first initialized and then refined through focusing analysis, and the impact location is identified by the grid point with the highest focusing response. Experimental results show that the proposed system enables effective impact localization on the pipe surface, with an average localization error of 15.75 mm. Comparative and robustness analyses further support the feasibility of the proposed sensing-localization scheme for impact monitoring on dispersive pipe structures.
Non-destructive evaluation (NDE) is used to detect and characterise defects in safety-critical components. This paper focuses on manual pulse-echo ultrasonic testing applied to the sizing of surface-breaking crack-like defects. In the field, practitioners often use A-scans from single element, pulse-echo ultrasonic testing. Machine learning, and in particular neural networks, have huge potential for applications to NDE-type problems due to their ability to recognise patterns from signals. Accurate predictions require large, labelled databases yet there is a paucity of such ultrasonic data for surface-breaking thermally fatigued cracks. To address this deficit, this work demonstrates that a database composed entirely of simulated, synthetic A-scans provides promising predictions on measured data. Crack height predictions with a mean absolute error of around 0.124 mm are achieved for measured data (on cracks between 0.5 mm and 4 mm). When tested on simulated A-scans, both height and tilt angle predictions are extremely accurate (to within 0.035 mm and 0.41°, respectively). To achieve such accuracy the in-situ inspection techniques were mimicked, with particular attention paid to the characteristics of the thermally fatigued crack species, and the single element transducer and its input signal. The details of the finite element simulations that generated the database are presented here. The work outlined in this paper shows that A-scans can be used to size defects using neural networks trained entirely on synthetic data, informed by experimental measurements, with potential applications for improving the efficiency of sizing cracks in-situ.
Sonic-IR (vibro-thermography) technique is an effective non-destructive evaluation method that detects cracks by capturing ultrasonic-induced interior heat generation and surface temperature rise using infrared thermography. This study proposes a quantitative approach for identifying internal crack depth in mortar by integrating experimental Sonic-IR imaging with inverse thermal conduction analysis. By characterizing the transient heat transfer behavior, the internal heat source distribution corresponding to the crack geometry is reconstructed using time-dependent heat conduction model. The results demonstrate that this hybrid approach can accurately estimate the interior location and spatial profile of cracks. In particular, the visualized distribution of crack tip depth is consistent with X-ray CT images. Furthermore, integrating temperature data obtained from multiple observation angles significantly enhances the accuracy and stability of the reconstructed crack shape. The combination of the Sonic-IR method and inverse thermal analysis offers a promising framework for the rapid, in-situ quantitative assessment of internal structural defects in cementitious materials.
Interfacial debonding in bi-layer composites commonly manifests as millimeter-scale fractures that alter microwave scattering and attenuation, yet quantitative links between crack state and effective dielectric response remain insufficiently constrained. Here, we develop and validate a waveguide S-parameter–based characterization framework to retrieve the effective complex permittivity of bilayer structures containing interfacial fractures over 5.8 GHz to 8.2 GHz. The inversion workflow is verified against both full-wave numerical simulations and laboratory waveguide measurements, and is further employed to systematically elucidate the coupled effects of crack geometry, spatial distribution, and infilling media on reflection, transmission attenuation, and the partitioning of energy dissipation. Results show that dry interfacial cracks primarily affect the energy-storage response. The real part of the effective permittivity decreases monotonically as the crack volume fraction increases from 1 % to 10 %, producing deviations of 0.93 % to 8.72 % relative to the intact interface. When the crack filling transitions from air to distilled water, the response shifts from a permittivity-real-part offset to a loss-dominated regime, characterized by pronounced transmission suppression accompanied by enhanced absorption. Moreover, under identical constituent properties, changing the crack inclination leads to up to a 5.37 % variation in the real part of the effective permittivity, indicating that orientation-driven geometric anisotropy significantly modulates the dielectric response. This study presents a validated inversion route for retrieving effective complex dielectric properties, providing a non-destructive means to evaluate the interfacial integrity of bi-layer structures in subsurface energy engineering.
Shallow rail surface defects are challenging to detect due to their minute dimensions, concealed distributions, and interference from complex near-surface wavefields. Conventional ultrasonic phased array (UPA) imaging often suffers from low computational efficiency, near-surface blind zones, and insufficient spatial resolution. To overcome these limitations, this study proposes a frequency–wavenumber domain imaging method (PVF-ωk) that integrates diffuse acoustic field excitation with a phase vector factor (PVF). Utilizing diffuse acoustic fields, multichannel signals are first processed via cross-correlation to reconstruct inter-element Green's functions, thereby enhancing scattering responses from shallow defects while suppressing background noise. A hybrid matrix is then constructed to facilitate the joint recovery of scattering information from defects at varying depths. Subsequently, a PVF derived from the phase consistency of multichannel signals is introduced to weight and modulate the frequency–wavenumber domain results, completing image reconstruction. Experimental validation using micro-circular holes and inclined cracks demonstrates that, compared with the conventional Total Focusing Method, the proposed PVF-ωk method reduces the relative imaging time by 50% and increases the average signal-to-noise ratio (SNR) by 36.17%. Furthermore, the roundness of circular holes is improved by 91.67%, with an average localization deviation of 0.16 mm. For inclined cracks with different orientations, the crack length and angle measurement errors are reduced to 2.67% and 11.67%, respectively. In addition, the proposed method successfully resolves multiple closely spaced micro-defects located at a depth of 2 mm with a diameter of 0.7 mm and a center-to-center spacing of 3 mm. These findings confirm that the PVF-ωk method achieves an optimal balance of computational efficiency, SNR, spatial resolution, and near-surface defect detection capability for diverse shallow rail surface defects.
This study develops a simulation-assisted nondestructive characterization approach using laser-excited surface acoustic waves to quantify porosity and extract pore size distributions in additively manufactured metallic components. A finite-element model, covering a porosity range of 1–6% together with representative pore-size statistics, is established to systematically analyze how pore- size statistics influence surface wave velocity and frequency-domain attenuation. A quantitative porosity evaluation model based on surface wave velocity is constructed, while an inversion approach combining Rayleigh-scattering theory with the frequency-domain attenuation of surface waves is proposed for estimating pore-size-distribution parameters. Simulation results show that surface wave velocity decreases with increasing porosity and exhibits relatively low sensitivity to variations in pore-size distribution. In contrast, frequency-domain attenuation is strongly affected by porosity and also varies with pore-size statistics within the investigated parameter range. Laser-ultrasonic experiments were conducted on two additively manufactured 316L stainless steel samples with different porosities. After correction for the systematic discrepancy between the simplified numerical model and the experimental response, the measured attenuation exhibits the same overall trend as the simulation-calibrated inversion model. This work provides a feasible non-contact approach for porosity evaluation and attenuation-based estimation of pore-size-distribution parameters in metal additive manufacturing.
Additive manufacturing (AM) parameters directly affect its energy density and hatch overlap ratio, which are susceptible to induce various micro-defects (ranging from tens to hundreds of microns) including pores, cracks, and lack-of-fusion defects. A reliable ultrasonic testing technique is urgently desired for evaluating the integrity of AM components. Aiming at the limitations of conventional ultrasonic testing including limited detection depth, low signal-to-noise ratio (SNR), and inaccuracies in micro-defect quantification. This work develops a novel synthetic aperture focusing multi-resolution C-scan (SAF-MRC) technique for quantitatively detecting micro-defects in AM components utilizing a traditional ultrasound C-scan system. Specifically, the synthetic aperture focusing (SAF) algorithm is employed to improve the SNR of scanning signals and the detection depth of the geometric focusing probe. An adaptive variable bandwidth split spectrum process (SSP) is developed to decompose the synthesized signal into multiple frequency sub-signals for constructing a series of multi-resolution C-scan images and overcoming the constant resolution limitation of the focusing probe. Finally, wave field directivity function of probe is utilized to elucidate the relationship between the detected multi-resolution defect sizes and split frequencies, and the relationship is used to quantitative micro-defect size. These will enable the quantitative detection of micro-defects at substantial depths in AM components with high SNR and precision. The feasibility of the proposed method is verified on selective laser-melted (SLM) GH4169 specimen, which contains different depths and micrometers flat-bottom-hole (FBH) defects, using an ultrasound C-scan system with a 20 MHz water immersion focusing probe. Results show that compared with the traditional ultrasound C-scan, the proposed SAF-MRC method can significantly reduce the relative error of detected size from 108.2% to 7.1% for the Φ=198 μm FBH in the focal zone, while reducing the relative error from 308.1% to 27.8% for the Φ=198 μm FBH at a defocused distance of 6 mm. Furthermore, for the Φ=389 μm FBH at a defocused distance of 10 mm, the relative error of detected size was improved from 67.1% to 11.8%. Results indicate the proposed SAF-MRC method has comprehensive advantages of large detection depth, high SNR, and high precision for planar-like defects.
The fundamental shear horizontal (SH0) wave electromagnetic acoustic transducer (EMAT) is widely used for defect detection in metal plates. However, in bounded plates, the inherent bidirectional radiation characteristic of conventional EMATs induces boundary reflections and multipath interference, causing true defect features to be obscured by complex artifacts. To address this issue, this paper proposes a multi-dimensional artifact suppression framework to systematically improve the defect characterization capability of ultrasonic imaging. First, at the physical sensing level, a novel unidirectional interleaved grating coil (UIGC) EMAT featuring reduced backward sidelobes is proposed for the directional transmission and reception of SH0 waves, achieving multipath signal separation, thereby suppressing the generation of mirror artifacts. Second, at the signal preprocessing level, a joint denoising algorithm based on successive variational mode decomposition and wavelet threshold denoising (SVMD-WTD) is utilized to extract the effective signals, thereby suppressing the noise artifacts. Finally, at the imaging level, an improved frequency-domain synthetic aperture focusing technique (IF-SAFT) is introduced to reconstruct the acquired signals, mitigating ghosting artifacts caused by beam divergence. Experimental results demonstrate that the proposed framework effectively reduces these three typical artifacts and improves the quality of defect images. This approach enables more accurate characterization of defect distribution and reliable size quantification, providing an effective technical solution for guided-wave imaging in bounded industrial structures.
Nonlinear ultrasonic techniques have emerged as vital tools for high-sensitivity material stress evaluation. Unlike conventional methods that rely on the cumulative nonlinearity along the wave propagation path, this study proposes a novel stress evaluation method based on nonlinear shear horizontal (SH) waves generated at the source-region of a magnetostrictive electromagnetic acoustic transducer (EMAT). The effects of static and dynamic magnetic fields on the stress response of the acoustic nonlinearity parameter (ANP) were investigated, clarifying the modulatory role of electromagnetic excitation parameters in ANP-based stress evaluation. Furthermore, we proposed a method to screen and optimize magnetic field and current parameters to enhance evaluation performance, thereby establishing an ANP-stress evaluation model under optimal conditions. Experimental results show that, under the optimal condition, namely a magnetic field strength of 10.31 kA/m and an excitation current of 80.57 A, the ANP exhibits an excellent monotonic linear relationship with stress, with a coefficient of determination (R2) of 0.99. The source-region ANP can effectively characterize the stress state of ferromagnetic materials, thereby providing a new methodological basis and optimization strategy for non-contact, highly sensitive, and nondestructive stress evaluation of ferromagnetic materials.
While guided wave detection has been addressed successfully in composite panels, its application to metallic plates remains considerably more challenging because wave propagation in metallic structures is fundamentally different and the presence of a crack further complicates the behavior. As a result, a novel multi-stage framework for crack identification is presented, combining tip-induced diffraction with detailed Lamb wave analysis. The method proceeds through detection, localization, and characterization, ensuring accurate evaluation of crack presence, orientation, and severity. A virtual baseline is reconstructed using shapelets extracted from Lamb wave signals via physics-guided K-SVD dictionary learning, effectively eliminating the need for prior baseline data and improving the robustness of GWSHM under varying conditions. Notably, the framework enables detailed crack characterization using only sparse sensors, which is considerably more challenging than simple damage detection or localization in guided wave based SHM.The framework begins with a crack detection stage, in which a reliability-based threshold is established to confirm the presence of crack and suppress false alarms. Once detection is confirmed, crack localization is achieved through a fusion imaging algorithm that estimates the crack position with high precision. Subsequently, a time-of-flight–based tip estimation technique is employed to distinguish tip diffractions from specular reflections, allowing accurate identification of both crack tips and enabling continuous monitoring of crack growth.The proposed framework has been experimentally validated on both artificial cracks and naturally developed notches generated using an Instron 250 kN hydraulic fatigue machine, demonstrating adaptability under diverse defect scenarios and loading conditions.Overall, by integrating virtual baseline reconstruction, reliability threshold analysis, crack imaging technique, weak-amplitude diffraction isolation, this study establishes a systematic methodology that substantially enhances the reliability and accuracy of long-term crack detection, localization, and characterization in GWSHM.
Quantitative tomographic assessment of repair quality in timber structures--discriminating filling material, residual cavities, and the sound wood matrix within a single reconstruction--remains insufficiently addressed. This study develops a simultaneous iterative reconstruction technique (SIRT) ultrasonic tomography framework that integrates adaptive ray updating with bivariate spline post-processing for defect detection and repair assessment in timber cross-sections. The ray paths are regenerated in an effective slowness field to reduce the forward-model mismatch associated with fixed straight-line assumptions, whereas the spline operation is used only after reconstruction to improve the continuity of anomaly boundaries without introducing additional measurement information. The method was validated on seven New Zealand pine (Pinus radiata) specimens consisting of an original-defect group and a repair group covering complete, incomplete, and secondary-complete epoxy repair. A direct same-data comparison with fixed straight-ray SIRT shows that adaptive path updating consistently improves localization and area quantification. For single-hole cavities the minimum centroid deviation was 1.24 mm and the minimum area error was 0.92%; for the secondary fully repaired specimen the area error decreased to 0.48%. The contrast between incomplete- and secondary-repair cases confirms the ability of the framework to distinguish residual cavities from epoxy-filled regions. For crack-type specimens, the defects were localized but area errors remained relatively high because a crack width of about 20 mm was represented by only about three cells in the 31 x 31 reconstruction grid. The proposed framework therefore extends ultrasonic tomography in wood from original-defect detection to quantitative, spatially resolved assessment of repair completeness.
Phased array ultrasonic testing (PAUT) is widely utilized non-destructive testing (NDT) method in welding inspection; however, it requires considerable expertise and substantial analysis time for the defect evaluation. To address these challenges, extensive research has been conducted on automating PAUT inspection using artificial intelligence (AI). However, existing AI-based approaches are frequently overfitted to limited training data, resulting in performance degradation on unseen data. This limitation is particularly critical in PAUT, where inspection data are highly sensitive to variations in specimen characteristics, potentially producing numerous spurious echo features being generated from the geometry around the weld boundary. In response to this challenge, this study proposes a robust deep learning-based PAUT inspection framework that incorporates specimen-level preprocessing to effectively suppress specimen-induced interfering echoes and improve defect identification accuracy. The proposed framework comprises two key components: welding boundary-based normalization and specimen-wise median filtering, which serve to standardize data geometry and reduce the impact of geometric echoes and noise across specimens. The model's generalization capability is validated through specimen-level train-test separation, demonstrating the superior performance of the proposed method. These findings are expected to significantly advance the practical industrial application of AI-based automated PAUT inspection systems.
