Pub Date : 2026-01-16DOI: 10.1016/j.ndteint.2026.103650
Tong Tong , Wan Qu , Jiadong Hua , Daogui Chen , Jinghan Tan , Jing Lin
Composite materials are widely employed in many industrial fields, and transmitted Lamb wave-based methods, represented by tomography, have been widely utilized for delamination detection in composite laminates. Nevertheless, conventional Lamb wave tomography may suffer from large artifacts and other problems. To break these limitations, a Lamb wave tomographic method based on sparse and probabilistic reconstruction for delamination detection in composite laminates is proposed in this study. Firstly, Lamb wave propagation in delaminated laminates is analyzed, from which it can be derived that delamination can cause the time-of-flight (ToF) delay of A0 mode. Then, differences in ToF between intact and delaminated laminates are calculated and constitute the time difference vector, which can be represented by the product of the length matrix and the slowness difference vector. Since the delamination distribution is sparse, the slowness difference vector satisfies the sparse assumption, which indicates that it can be solved with sparse reconstruction techniques. Furthermore, to improve the quality of sparse reconstruction, the probability distribution is introduced as a prior weight during the solving procedure. Finally, numerical and experimental investigations are implemented. The imaging results can provide a more precise estimation of delamination size and location, which demonstrates the performance improvement of the presented approach.
{"title":"Delamination detection in composite laminates using Lamb wave tomographic method based on sparse and probabilistic reconstruction","authors":"Tong Tong , Wan Qu , Jiadong Hua , Daogui Chen , Jinghan Tan , Jing Lin","doi":"10.1016/j.ndteint.2026.103650","DOIUrl":"10.1016/j.ndteint.2026.103650","url":null,"abstract":"<div><div>Composite materials are widely employed in many industrial fields, and transmitted Lamb wave-based methods, represented by tomography, have been widely utilized for delamination detection in composite laminates. Nevertheless, conventional Lamb wave tomography may suffer from large artifacts and other problems. To break these limitations, a Lamb wave tomographic method based on sparse and probabilistic reconstruction for delamination detection in composite laminates is proposed in this study. Firstly, Lamb wave propagation in delaminated laminates is analyzed, from which it can be derived that delamination can cause the time-of-flight (ToF) delay of A0 mode. Then, differences in ToF between intact and delaminated laminates are calculated and constitute the time difference vector, which can be represented by the product of the length matrix and the slowness difference vector. Since the delamination distribution is sparse, the slowness difference vector satisfies the sparse assumption, which indicates that it can be solved with sparse reconstruction techniques. Furthermore, to improve the quality of sparse reconstruction, the probability distribution is introduced as a prior weight during the solving procedure. Finally, numerical and experimental investigations are implemented. The imaging results can provide a more precise estimation of delamination size and location, which demonstrates the performance improvement of the presented approach.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103650"},"PeriodicalIF":4.5,"publicationDate":"2026-01-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146034887","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-14DOI: 10.1016/j.ndteint.2026.103639
Vladimir Vavilov, Arsenii Chulkov, Olesia Ganina, Marina Kuimova, Oleg Makushev
This study presents a comprehensive methodology for characterizing air-filled finite-size defects in materials with varying thermal properties using pulsed thermal nondestructive testing (TNDT). We numerically solve the three-dimensional heat transfer problem for 729 test cases encompassing defects with different lateral dimensions, depths, and thicknesses in both metallic and non-metallic materials. The analysis yields maximum temperature contrasts and their corresponding observation times, while investigating the influence of defect geometry on thermal signatures. An analytical expression for predicting observation times is derived to complement the numerical results.
The computational results are fitted with polynomial functions to enable rapid estimation of optimal TNDT parameters. This approach provides a practical framework for evaluating detection limits across a wide range of material properties and defect geometries. System-wide analysis reveals mean errors of 60 % for temperature contrast evaluation and 36 % for determination of observation times. Experimental validation using reference samples demonstrates measurement accuracies of 14–35 % for temperature contrasts and 2–8 % for observation times. The proposed inverse solution achieves particularly accurate depth characterization (<14 % error), though thickness estimation shows greater variability (up to 61 % error).
{"title":"The methodology of defect thermal characterization in pulsed thermal NDT based on 3D numerical solutions and polynomial approximation","authors":"Vladimir Vavilov, Arsenii Chulkov, Olesia Ganina, Marina Kuimova, Oleg Makushev","doi":"10.1016/j.ndteint.2026.103639","DOIUrl":"10.1016/j.ndteint.2026.103639","url":null,"abstract":"<div><div>This study presents a comprehensive methodology for characterizing air-filled finite-size defects in materials with varying thermal properties using pulsed thermal nondestructive testing (TNDT). We numerically solve the three-dimensional heat transfer problem for 729 test cases encompassing defects with different lateral dimensions, depths, and thicknesses in both metallic and non-metallic materials. The analysis yields maximum temperature contrasts and their corresponding observation times, while investigating the influence of defect geometry on thermal signatures. An analytical expression for predicting observation times is derived to complement the numerical results.</div><div>The computational results are fitted with polynomial functions to enable rapid estimation of optimal TNDT parameters. This approach provides a practical framework for evaluating detection limits across a wide range of material properties and defect geometries. System-wide analysis reveals mean errors of 60 % for temperature contrast evaluation and 36 % for determination of observation times. Experimental validation using reference samples demonstrates measurement accuracies of 14–35 % for temperature contrasts and 2–8 % for observation times. The proposed inverse solution achieves particularly accurate depth characterization (<14 % error), though thickness estimation shows greater variability (up to 61 % error).</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103639"},"PeriodicalIF":4.5,"publicationDate":"2026-01-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979242","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-12DOI: 10.1016/j.ndteint.2026.103648
Wenyi Xu, Jing Yu, Yuxin Chang, Ruiyuan Niu, Guanglei Zhu, Ning Tao, Jiangang Sun
In this study, a data conversion method between modulated thermal imaging and flash thermal imaging is derived theoretically and demonstrated experimentally. The method allows for modulated data acquired at one frequency to be forwardly converted to a full flash data which can then be backwardly converted to a modulated data at a different frequency. The experimental demonstrations were carried out using a glass fiber reinforced plastic (GFRP) plate sample that contains flat bottom holes located at various depths. From a forward conversion of measured modulated data, the converted flash data was processed for defect detection by using the thermal effusivity tomography method and the results were compared with the corresponding ones obtained from a flash experiment on the same sample. In addition, backward conversions from the converted flash data to new sets of modulated data at various other frequencies were demonstrated and verified. The results show that this data-conversion method can address the detection of subsurface defects within different depths, which will eradicate the blind-frequency problem and eliminate the need for performing multiple tests with different modulation frequencies.
{"title":"Method and application of data conversion between modulated and flash thermal imaging","authors":"Wenyi Xu, Jing Yu, Yuxin Chang, Ruiyuan Niu, Guanglei Zhu, Ning Tao, Jiangang Sun","doi":"10.1016/j.ndteint.2026.103648","DOIUrl":"10.1016/j.ndteint.2026.103648","url":null,"abstract":"<div><div>In this study, a data conversion method between modulated thermal imaging and flash thermal imaging is derived theoretically and demonstrated experimentally. The method allows for modulated data acquired at one frequency to be forwardly converted to a full flash data which can then be backwardly converted to a modulated data at a different frequency. The experimental demonstrations were carried out using a glass fiber reinforced plastic (GFRP) plate sample that contains flat bottom holes located at various depths. From a forward conversion of measured modulated data, the converted flash data was processed for defect detection by using the thermal effusivity tomography method and the results were compared with the corresponding ones obtained from a flash experiment on the same sample. In addition, backward conversions from the converted flash data to new sets of modulated data at various other frequencies were demonstrated and verified. The results show that this data-conversion method can address the detection of subsurface defects within different depths, which will eradicate the blind-frequency problem and eliminate the need for performing multiple tests with different modulation frequencies.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103648"},"PeriodicalIF":4.5,"publicationDate":"2026-01-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146034846","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-12DOI: 10.1016/j.ndteint.2026.103646
Mingtao Liu , Xue Bai , Fei Shao , Jian Ma
This paper addresses the challenge of accurately detecting surface and subsurface defects in laser additively manufactured components characterized by high surface roughness. A novel laser ultrasonic imaging method is proposed based on regularized Expectation-Maximization (EM) clustering. The theoretical foundation exploits the observation that ultrasonic feature signal intensities, derived from both transmitted Rayleigh waves (for surface defects) and time-delayed superposed scattered echo signals (for subsurface defects), conform to a Gaussian Mixture Model (GMM). By constructing a GMM and implementing the EM algorithm, the proposed method enables the adaptive separation of defect signals from background noise arising from surface roughness. To improve algorithmic stability and robustness, an adaptive regularization technique based on differential evolution was incorporated, addressing covariance singularity and accelerating convergence. The performance of the proposed method was validated on AlSi10Mg and Ti6Al4V samples. Even under challenging conditions of high surface roughness (Ra = 37.5 μm), the method successfully detects submillimeter surface defects with diameters as small as 0.4 mm. Additionally, the regularized EM clustering approach demonstrates excellent resolution for subsurface defects from 0.5 mm down to sub-wavelength depths (1.1 mm, ∼0.9λ) with a diameter of 0.5 mm. The method also shows strong adaptability in limited sample and high-noise scenarios, outperforming a convolutional neural network-based benchmark in detection accuracy and false detection rate. The core innovation of this approach lies in clustering feature signal data to distinguish defect-related signals from noise, enabling adaptive noise reduction on rough surfaces and minimizing the false detection rate. The proposed method offers a promising application pathway for both online defect detection during the laser additive manufacturing process and comprehensive defect evaluation in components with high surface roughness.
针对激光增材制造零件表面粗糙度高的特点，提出了精确检测表面和亚表面缺陷的难题。提出了一种基于正则化期望最大化聚类的激光超声成像方法。理论基础是基于对透射瑞利波（用于表面缺陷）和延时叠加散射回波信号（用于亚表面缺陷）的超声特征信号强度符合高斯混合模型（GMM）的观察。该方法通过构造GMM和实现EM算法，实现了缺陷信号与表面粗糙度引起的背景噪声的自适应分离。为了提高算法的稳定性和鲁棒性，引入了基于差分进化的自适应正则化技术，解决了协方差奇异性，加快了收敛速度。在AlSi10Mg和Ti6Al4V样品上验证了该方法的性能。即使在具有挑战性的高表面粗糙度条件下（Ra = 37.5 μm），该方法也能成功检测到直径小至0.4 mm的亚毫米表面缺陷。此外，正则化EM聚类方法对直径为0.5 mm的从0.5 mm到亚波长深度（1.1 mm, ~ 0.9λ）的亚表面缺陷具有出色的分辨率。该方法在有限样本和高噪声场景下也表现出较强的适应性，在检测精度和误检率方面优于基于卷积神经网络的基准。该方法的核心创新点在于对特征信号数据进行聚类，将缺陷相关信号与噪声区分开来，实现粗糙表面的自适应降噪，最大限度地降低误检率。该方法为激光增材制造过程中的在线缺陷检测和高表面粗糙度部件的综合缺陷评估提供了一条有前景的应用途径。
{"title":"Regularized expectation-maximization clustering enhanced laser ultrasonic imaging for defects in laser additively manufactured components with high surface roughness","authors":"Mingtao Liu , Xue Bai , Fei Shao , Jian Ma","doi":"10.1016/j.ndteint.2026.103646","DOIUrl":"10.1016/j.ndteint.2026.103646","url":null,"abstract":"<div><div>This paper addresses the challenge of accurately detecting surface and subsurface defects in laser additively manufactured components characterized by high surface roughness. A novel laser ultrasonic imaging method is proposed based on regularized Expectation-Maximization (EM) clustering. The theoretical foundation exploits the observation that ultrasonic feature signal intensities, derived from both transmitted Rayleigh waves (for surface defects) and time-delayed superposed scattered echo signals (for subsurface defects), conform to a Gaussian Mixture Model (GMM). By constructing a GMM and implementing the EM algorithm, the proposed method enables the adaptive separation of defect signals from background noise arising from surface roughness. To improve algorithmic stability and robustness, an adaptive regularization technique based on differential evolution was incorporated, addressing covariance singularity and accelerating convergence. The performance of the proposed method was validated on AlSi10Mg and Ti6Al4V samples. Even under challenging conditions of high surface roughness (Ra = 37.5 μm), the method successfully detects submillimeter surface defects with diameters as small as 0.4 mm. Additionally, the regularized EM clustering approach demonstrates excellent resolution for subsurface defects from 0.5 mm down to sub-wavelength depths (1.1 mm, ∼0.9λ) with a diameter of 0.5 mm. The method also shows strong adaptability in limited sample and high-noise scenarios, outperforming a convolutional neural network-based benchmark in detection accuracy and false detection rate. The core innovation of this approach lies in clustering feature signal data to distinguish defect-related signals from noise, enabling adaptive noise reduction on rough surfaces and minimizing the false detection rate. The proposed method offers a promising application pathway for both online defect detection during the laser additive manufacturing process and comprehensive defect evaluation in components with high surface roughness.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103646"},"PeriodicalIF":4.5,"publicationDate":"2026-01-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979264","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Oil and gas pipelines serve as critical global energy infrastructure, where structural integrity is paramount for ensuring energy security and preventing catastrophic accidents. However, in-service pipelines operating in harsh environments present significant challenges for non-destructive testing due to severely constrained inspection spaces. Limited-angle computed tomography (CT) has emerged as a useful method for detecting pipeline defects, but incomplete projection data leads to severe reconstruction artifacts when using conventional algorithms, substantially compromising defect detection accuracy. While unsupervised deep learning methods show promise without requiring paired training data, existing approaches primarily rely on implicit network priors, making it difficult to guarantee geometric fidelity of reconstructed structures. To address this challenge, this study proposes a novel Contour Guided-Deep Radon Prior (CG-DRP) unsupervised reconstruction framework. The key innovation incorporates known geometric contours of pipeline structures as explicit physical constraints deeply integrated into the Deep Radon Prior (DRP) optimization process, achieving optimal fusion of physical prior accuracy and unsupervised learning flexibility. The framework additionally incorporates Convolutional Block Attention Module (CBAM) to enhance feature extraction capabilities. Experimental validation using simulated and real pipeline data under 90°and 120°limited-angle conditions demonstrates that CG-DRP comprehensively outperforms traditional algorithms (FBP, SART, ADMM-TV) and advanced unsupervised methods (DIP, RBP-DIP, DRP). Reconstructed images achieve optimal PSNR and SSIM performance, effectively suppressing artifacts while preserving structural details and minor defects. The research confirms CG-DRP’s robustness and superiority, providing an efficient solution for industrial CT applications in pipeline integrity assessment.
{"title":"Contour guided-deep radon prior: A robust unsupervised framework for limited-angle CT inspection of oil and gas pipelines","authors":"Jintao Fu, Tianchen Zeng, Jiahao Chang, Peng Cong, Ximing Liu, Yuewen Sun","doi":"10.1016/j.ndteint.2026.103647","DOIUrl":"10.1016/j.ndteint.2026.103647","url":null,"abstract":"<div><div>Oil and gas pipelines serve as critical global energy infrastructure, where structural integrity is paramount for ensuring energy security and preventing catastrophic accidents. However, in-service pipelines operating in harsh environments present significant challenges for non-destructive testing due to severely constrained inspection spaces. Limited-angle computed tomography (CT) has emerged as a useful method for detecting pipeline defects, but incomplete projection data leads to severe reconstruction artifacts when using conventional algorithms, substantially compromising defect detection accuracy. While unsupervised deep learning methods show promise without requiring paired training data, existing approaches primarily rely on implicit network priors, making it difficult to guarantee geometric fidelity of reconstructed structures. To address this challenge, this study proposes a novel Contour Guided-Deep Radon Prior (CG-DRP) unsupervised reconstruction framework. The key innovation incorporates known geometric contours of pipeline structures as explicit physical constraints deeply integrated into the Deep Radon Prior (DRP) optimization process, achieving optimal fusion of physical prior accuracy and unsupervised learning flexibility. The framework additionally incorporates Convolutional Block Attention Module (CBAM) to enhance feature extraction capabilities. Experimental validation using simulated and real pipeline data under 90°and 120°limited-angle conditions demonstrates that CG-DRP comprehensively outperforms traditional algorithms (FBP, SART, ADMM-TV) and advanced unsupervised methods (DIP, RBP-DIP, DRP). Reconstructed images achieve optimal PSNR and SSIM performance, effectively suppressing artifacts while preserving structural details and minor defects. The research confirms CG-DRP’s robustness and superiority, providing an efficient solution for industrial CT applications in pipeline integrity assessment.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103647"},"PeriodicalIF":4.5,"publicationDate":"2026-01-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Sensors are the primary component in the study of health monitoring of various structures. Numerous cutting-edge smart sensors have been utilized to improve monitoring technologies, however, the necessity to patch them to the structure in close contact still creates major complications in their actual deployment. Moreover, the contact sensors add a mass penalty to the structural element, causing a challenge for thin and flexible structures. In this paper, we introduce a contactless approach for damage localization in metallic plates using ultra-wide-band (UWB) antennas to overcome the limitations of contact-based approaches. The UWB antenna array is placed at a distance from the structure and is used to transmit and receive electromagnetic (EM) waves in the radio frequency (RF) range. Additionally, an imaging algorithm is developed to locate the damage in the structure. The simulation and experimental results demonstrate that the algorithm accurately estimates the damage locations. Furthermore, the estimated results of the proposed RF-based approach are comparatively validated with the existing ultrasonic sensor-based contact approach. Our simulation and experimental results show that both techniques (ultrasonic and RF) have a par accuracy of 99.93% for damage localization with respect to actual damage locations. The comparative study confirms that the UWB antennas are equally efficient in multi-damage localization in metallic plates, with the additional advantage of eliminating sensor patching onto the structure. This leads to the conviction that the UWB antennas are a novel addition to contactless SHM for various metallic structures. The technique is further extended to non-metallic airfoil structures for damage localization, and the computed accuracy of located damage is 95.1% with respect to actual damage location.
{"title":"RF antenna array for contactless structural health monitoring: Ultrasonic benchmarking and application to airfoil structure","authors":"Deepak Kumar , Yogesh Kumar Yadav , Sahil Kalra , Prabhat Munshi","doi":"10.1016/j.ndteint.2026.103643","DOIUrl":"10.1016/j.ndteint.2026.103643","url":null,"abstract":"<div><div>Sensors are the primary component in the study of health monitoring of various structures. Numerous cutting-edge smart sensors have been utilized to improve monitoring technologies, however, the necessity to patch them to the structure in close contact still creates major complications in their actual deployment. Moreover, the contact sensors add a mass penalty to the structural element, causing a challenge for thin and flexible structures. In this paper, we introduce a contactless approach for damage localization in metallic plates using ultra-wide-band (UWB) antennas to overcome the limitations of contact-based approaches. The UWB antenna array is placed at a distance from the structure and is used to transmit and receive electromagnetic (EM) waves in the radio frequency (RF) range. Additionally, an imaging algorithm is developed to locate the damage in the structure. The simulation and experimental results demonstrate that the algorithm accurately estimates the damage locations. Furthermore, the estimated results of the proposed RF-based approach are comparatively validated with the existing ultrasonic sensor-based contact approach. Our simulation and experimental results show that both techniques (ultrasonic and RF) have a par accuracy of 99.93% for damage localization with respect to actual damage locations. The comparative study confirms that the UWB antennas are equally efficient in multi-damage localization in metallic plates, with the additional advantage of eliminating sensor patching onto the structure. This leads to the conviction that the UWB antennas are a novel addition to contactless SHM for various metallic structures. The technique is further extended to non-metallic airfoil structures for damage localization, and the computed accuracy of located damage is 95.1% with respect to actual damage location.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103643"},"PeriodicalIF":4.5,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979241","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-09DOI: 10.1016/j.ndteint.2026.103640
Seyed Hamidreza Afzalimir, Parisa Shokouhi, Cliff J. Lissenden
Hydrogen embrittlement encompasses many material degradation mechanisms that lead to loss of ductility and brittle fracture. Ultrasound testing, as a structural integrity evaluation method, will be shown to detect diffused hydrogen. Cubic samples were extracted from a cold-drawn Al2024 bar and charged with hydrogen. Ultrasound testing was performed in the three principal directions of the cubic samples: L (longitudinal, parallel to the drawing direction that elongated the grains), T (long transverse), and S (short transverse), both before and after hydrogen charging. Linear ultrasound testing – specifically using the pulse-echo mode for wave speed and attenuation measurements – shows moderate sensitivity to hydrogen charging. Nonlinear ultrasound testing – specifically for second-harmonic generation (SHG) – exhibits high sensitivity to hydrogen charging with wave propagation in the L direction, moderate sensitivity in the T direction, and low sensitivity in the S direction. We interpret these SHG results with respect to recent predictions of the effect that solute H atoms near a grain boundary have on the acoustic nonlinearity parameter. Model results show that the acoustic nonlinearity parameter increases dramatically for waves parallel to the grain boundary. Moreover, the acoustic nonlinearity parameter is predicted to decrease modestly for ultrasonic waves normal to the grain boundary. The cold-drawn bar has many grain boundaries parallel to the L-direction, but relatively few parallel to the S-direction. Thus, the SHG results in the L- and S-directions correspond roughly to the waves parallel and normal, respectively, to the grain boundary in the model. This study improves our understanding of how nonlinear ultrasound testing can be applied effectively as a diagnostic tool to detect hydrogen embrittlement.
{"title":"Nonlinear ultrasound to detect hydrogen embrittlement in Al2024","authors":"Seyed Hamidreza Afzalimir, Parisa Shokouhi, Cliff J. Lissenden","doi":"10.1016/j.ndteint.2026.103640","DOIUrl":"10.1016/j.ndteint.2026.103640","url":null,"abstract":"<div><div>Hydrogen embrittlement encompasses many material degradation mechanisms that lead to loss of ductility and brittle fracture. Ultrasound testing, as a structural integrity evaluation method, will be shown to detect diffused hydrogen. Cubic samples were extracted from a cold-drawn Al2024 bar and charged with hydrogen. Ultrasound testing was performed in the three principal directions of the cubic samples: L (longitudinal, parallel to the drawing direction that elongated the grains), T (long transverse), and S (short transverse), both before and after hydrogen charging. Linear ultrasound testing – specifically using the pulse-echo mode for wave speed and attenuation measurements – shows moderate sensitivity to hydrogen charging. Nonlinear ultrasound testing – specifically for second-harmonic generation (SHG) – exhibits high sensitivity to hydrogen charging with wave propagation in the L direction, moderate sensitivity in the T direction, and low sensitivity in the S direction. We interpret these SHG results with respect to recent predictions of the effect that solute H atoms near a grain boundary have on the acoustic nonlinearity parameter. Model results show that the acoustic nonlinearity parameter increases dramatically for waves parallel to the grain boundary. Moreover, the acoustic nonlinearity parameter is predicted to decrease modestly for ultrasonic waves normal to the grain boundary. The cold-drawn bar has many grain boundaries parallel to the L-direction, but relatively few parallel to the S-direction. Thus, the SHG results in the L- and S-directions correspond roughly to the waves parallel and normal, respectively, to the grain boundary in the model. This study improves our understanding of how nonlinear ultrasound testing can be applied effectively as a diagnostic tool to detect hydrogen embrittlement.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103640"},"PeriodicalIF":4.5,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-08DOI: 10.1016/j.ndteint.2026.103642
Yang Yang , Bofan Liu , Zongfei Tong , Hong-En Chen , Jianguo Zhu , Cuixiang Pei , Shejuan Xie , Hao Su , Zhenmao Chen
Fatigue crack is a typical defect initiated in key engineering structures under dynamic loads. The propagation of fatigue cracks would significantly shorten the structural service life and even cause serious accidents. The vibrothermography (VT), as a promising non-destructive testing (NDT) technique, presents great potential for fatigue crack inspection due to its internal heating mode and applicable for both metallic and nonmetallic materials. However, the multi-parameters optimization and agent model building of VT system put forward higher requirements of an efficient numerical simulation technique for VT signals. In this paper, a fast numerical method for low-power VT under high frequency excitation is proposed and validated. For efficient simulation of dynamic displacement, the element birth and death method is utilized to adjust the coefficient matrix of finite element based on the contact or separation state of crack surface. This method can cope with the complex nonlinear phenomenon of crack closing properly while maintaining computational feasibility during vibration analysis. For the simulation of temperature field of VT, the energy equivalent method proposed by authors is employed to address the efficiency problem of the direct time domain integration for the high-frequency excitation. By linearizing the heat source, the present method can reduce computational burden while preserving numerical accuracy, enabling efficient simulation of the thermal field during VT process. Finally, the proposed method is validated via numerical simulations and experiments which show that the method is over six times faster than the commercial software but with a comparablenumerical precision.
{"title":"A fast numerical method for low-power vibrothermography nondestructive testing of fatigue cracks","authors":"Yang Yang , Bofan Liu , Zongfei Tong , Hong-En Chen , Jianguo Zhu , Cuixiang Pei , Shejuan Xie , Hao Su , Zhenmao Chen","doi":"10.1016/j.ndteint.2026.103642","DOIUrl":"10.1016/j.ndteint.2026.103642","url":null,"abstract":"<div><div>Fatigue crack is a typical defect initiated in key engineering structures under dynamic loads. The propagation of fatigue cracks would significantly shorten the structural service life and even cause serious accidents. The vibrothermography (VT), as a promising non-destructive testing (NDT) technique, presents great potential for fatigue crack inspection due to its internal heating mode and applicable for both metallic and nonmetallic materials. However, the multi-parameters optimization and agent model building of VT system put forward higher requirements of an efficient numerical simulation technique for VT signals. In this paper, a fast numerical method for low-power VT under high frequency excitation is proposed and validated. For efficient simulation of dynamic displacement, the element birth and death method is utilized to adjust the coefficient matrix of finite element based on the contact or separation state of crack surface. This method can cope with the complex nonlinear phenomenon of crack closing properly while maintaining computational feasibility during vibration analysis. For the simulation of temperature field of VT, the energy equivalent method proposed by authors is employed to address the efficiency problem of the direct time domain integration for the high-frequency excitation. By linearizing the heat source, the present method can reduce computational burden while preserving numerical accuracy, enabling efficient simulation of the thermal field during VT process. Finally, the proposed method is validated via numerical simulations and experiments which show that the method is over six times faster than the commercial software but with a comparablenumerical precision.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103642"},"PeriodicalIF":4.5,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979263","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-07DOI: 10.1016/j.ndteint.2026.103638
Alicia Ortiz-Chiliquinga , Fernando Moreno-Haya , Carlos López-Pingarrón , Carlos Jesús Vega-Vera , José S. Torrecilla
The reliable evaluation of lubricating oil condition is critical for ensuring the safety and operational efficiency of heavy-duty equipment in both civilian and defense sectors. Conventional laboratory-based physicochemical analyses, although effective, are inherently time-consuming and do not enable real-time diagnostics or on-site decision-making. In this work, we introduce an innovative approach that leverages infrared thermography coupled with deep learning to achieve rapid, non-destructive, and fully automated classification of lubricating oil samples as either “compliant” (fit for use) or “non-compliant” (unfit for use). The study focuses on two reference lubricants (O-1178 (5W30), gearbox oil and O-1236 (15W40), engine oil) widely deployed in military vehicles, with ground-truth class labels established via standardized laboratory protocols. A comprehensive dataset of over 10,000 thermographic images was generated through controlled cooling cycles, providing the foundation for model development. After comparative analysis of several state-of-the-art convolutional neural network architectures, ResNet-34 and ResNet-50 were selected for their superior performance. The models, trained and validated on stratified and balanced datasets, consistently achieved classification accuracies above 99 %, with the ResNet-34 model delivering 100 % sensitivity and specificity for the detection of non-compliant samples in both oil types. Complementary metrics, including ROC/AUC (≈1.0) and F1-scores near unity, together with stable training–validation loss convergence, confirmed that the classifiers operated in a saturated performance regime with robust generalization. Interpretation with Grad-CAM heatmaps revealed that the model's decisions are grounded in physically meaningful thermal micropatterns directly linked to lubricant degradation. This strategy not only minimizes unnecessary oil changes and associated environmental impact, but also elevates predictive maintenance capabilities by enabling immediate, reliable diagnostics in dual-use (civil and military) settings. The proposed methodology establishes a robust and versatile framework for advanced lubricant condition monitoring, readily adaptable to other industrial fluids and diverse operational scenarios requiring rapid, on-site assessment. Future work will extend this framework to additional lubricant types and broader real-world conditions to further consolidate these findings.
{"title":"Infrared thermography coupled with deep learning for fast and reliable predictive monitoring of lubricating oils in dual-use heavy-duty vehicles","authors":"Alicia Ortiz-Chiliquinga , Fernando Moreno-Haya , Carlos López-Pingarrón , Carlos Jesús Vega-Vera , José S. Torrecilla","doi":"10.1016/j.ndteint.2026.103638","DOIUrl":"10.1016/j.ndteint.2026.103638","url":null,"abstract":"<div><div>The reliable evaluation of lubricating oil condition is critical for ensuring the safety and operational efficiency of heavy-duty equipment in both civilian and defense sectors. Conventional laboratory-based physicochemical analyses, although effective, are inherently time-consuming and do not enable real-time diagnostics or on-site decision-making. In this work, we introduce an innovative approach that leverages infrared thermography coupled with deep learning to achieve rapid, non-destructive, and fully automated classification of lubricating oil samples as either “compliant” (fit for use) or “non-compliant” (unfit for use). The study focuses on two reference lubricants (O-1178 (5W30), gearbox oil and O-1236 (15W40), engine oil) widely deployed in military vehicles, with ground-truth class labels established via standardized laboratory protocols. A comprehensive dataset of over 10,000 thermographic images was generated through controlled cooling cycles, providing the foundation for model development. After comparative analysis of several state-of-the-art convolutional neural network architectures, ResNet-34 and ResNet-50 were selected for their superior performance. The models, trained and validated on stratified and balanced datasets, consistently achieved classification accuracies above 99 %, with the ResNet-34 model delivering 100 % sensitivity and specificity for the detection of non-compliant samples in both oil types. Complementary metrics, including ROC/AUC (≈1.0) and F1-scores near unity, together with stable training–validation loss convergence, confirmed that the classifiers operated in a saturated performance regime with robust generalization. Interpretation with Grad-CAM heatmaps revealed that the model's decisions are grounded in physically meaningful thermal micropatterns directly linked to lubricant degradation. This strategy not only minimizes unnecessary oil changes and associated environmental impact, but also elevates predictive maintenance capabilities by enabling immediate, reliable diagnostics in dual-use (civil and military) settings. The proposed methodology establishes a robust and versatile framework for advanced lubricant condition monitoring, readily adaptable to other industrial fluids and diverse operational scenarios requiring rapid, on-site assessment. Future work will extend this framework to additional lubricant types and broader real-world conditions to further consolidate these findings.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103638"},"PeriodicalIF":4.5,"publicationDate":"2026-01-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145940690","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-01-07DOI: 10.1016/j.ndteint.2026.103641
Weiying Cheng
In the inspection of partial-circumferential pipe wall thinning (PCPWT) using mode microwaves, higher-order mode microwaves are excited at various frequencies when the primary mode interacts with the thinning region. This alters the reflection of mode waves, and consequently affects the signals. The behavior of signals varies across different frequency ranges. Therefore, in this study, we analyzed the signals within specific corresponding frequency bands: (1) low frequencies, where higher-order modes have not yet been generated; (2) intermediate frequencies, where the mode is excited but the mode is not yet; and (3) higher frequencies, where both the and mode are excited. In the signal analysis, Singular Spectral Analysis (SSA) was employed to decompose the simulated signal into two components: a slowly oscillating component - exhibiting beating patterns particularly in the low-frequency range - and a residual component, characterized by irregular oscillation attributed to higher-order modes, especially at intermediate and higher frequencies. The results showed that both the thinning thickness and circumferential extent can be characterized using features derived from the two components. In the experimental study, a variety of signal processing techniques have been applied to measurement signals, which include reflections other than from the PCPWT. By using SSA and transforming the measurement signals across various domains – namely, frequency, spatial, and - domains – signals most strongly associated with PWT were successfully extracted. These signals exhibited features consistent with simulation results, validating their potential for characterizing higher order modes and, consequently, PCPWT.
{"title":"Features extraction for characterizing partial-circumferential pipe wall thinning using TM01 mode microwaves","authors":"Weiying Cheng","doi":"10.1016/j.ndteint.2026.103641","DOIUrl":"10.1016/j.ndteint.2026.103641","url":null,"abstract":"<div><div>In the inspection of partial-circumferential pipe wall thinning (PCPWT) using <span><math><mrow><msub><mrow><mi>T</mi><mi>M</mi></mrow><mn>01</mn></msub></mrow></math></span> mode microwaves, higher-order mode microwaves are excited at various frequencies when the primary <span><math><mrow><msub><mrow><mi>T</mi><mi>M</mi></mrow><mn>01</mn></msub></mrow></math></span> mode interacts with the thinning region. This alters the reflection of <span><math><mrow><msub><mrow><mi>T</mi><mi>M</mi></mrow><mn>01</mn></msub></mrow></math></span> mode waves, and consequently affects the <span><math><mrow><msub><mi>S</mi><mn>11</mn></msub></mrow></math></span> signals. The behavior of <span><math><mrow><msub><mi>S</mi><mn>11</mn></msub></mrow></math></span> signals varies across different frequency ranges. Therefore, in this study, we analyzed the signals within specific corresponding frequency bands: (1) low frequencies, where higher-order modes have not yet been generated; (2) intermediate frequencies, where the <span><math><mrow><msub><mrow><mi>T</mi><mi>E</mi></mrow><mn>21</mn></msub></mrow></math></span> mode is excited but the <span><math><mrow><msub><mrow><mi>T</mi><mi>M</mi></mrow><mn>11</mn></msub></mrow></math></span> mode is not yet; and (3) higher frequencies, where both the <span><math><mrow><msub><mrow><mi>T</mi><mi>E</mi></mrow><mn>21</mn></msub></mrow></math></span> and <span><math><mrow><msub><mrow><mi>T</mi><mi>M</mi></mrow><mn>11</mn></msub></mrow></math></span> mode are excited. In the signal analysis, Singular Spectral Analysis (SSA) was employed to decompose the simulated <span><math><mrow><msub><mi>S</mi><mn>11</mn></msub></mrow></math></span> signal into two components: a slowly oscillating component - exhibiting beating patterns particularly in the low-frequency range - and a residual component, characterized by irregular oscillation attributed to higher-order modes, especially at intermediate and higher frequencies. The results showed that both the thinning thickness and circumferential extent can be characterized using features derived from the two components. In the experimental study, a variety of signal processing techniques have been applied to measurement signals, which include reflections other than from the PCPWT. By using SSA and transforming the measurement signals across various domains – namely, frequency, spatial, and <span><math><mrow><mi>Ω</mi></mrow></math></span>- domains – signals most strongly associated with PWT were successfully extracted. These signals exhibited features consistent with simulation results, validating their potential for characterizing higher order modes and, consequently, PCPWT.</div></div>","PeriodicalId":18868,"journal":{"name":"Ndt & E International","volume":"160 ","pages":"Article 103641"},"PeriodicalIF":4.5,"publicationDate":"2026-01-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145979337","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号