Journal of Nondestructive Evaluation最新文献

Terahertz (THz) imaging has emerged as a powerful modality for nondestructive testing (NDT), especially for inspecting non-conductive materials where traditional X-rays and ultrasound techniques fall short. Here, a compressive sensing-based THz single-pixel imaging system optimised for real-time nondestructive testing and evaluation of composite materials has been employed. Different structured and random sensing masks have been employed, namely, the discrete cosine transform (DCT), Hadamard, Gaussian, Bernoulli, and random. The impact of various mask reordering strategies, including Cake-Cutting, Total Gradient, and Total Variation, on the image quality has been systematically examined. Image quality has been quantitatively assessed using Mean Square Error, Peak Signal-to-Noise Ratio, and Structural Similarity Index Measure metrics across different sampling ratios and noise levels. A novel Deconvolved Energy (DE) reordering has been proposed and implemented, where a descending reordering has been carried out based on the energy of the mask pattern deconvolved with the Tikhonov regularised blur kernel. From the results, it is evident that DCT-based masks consistently outperform others in terms of THz image reconstruction fidelity and computational efficiency, especially when paired with DE reordering. The generalizability of the proposed methodology has been validated by different THz images acquired with a variety of defects across different composite materials. From the results, it is evident that the proposed methodology achieves robust THz image reconstruction even in under-sampled scenarios and in the presence of noise, with significantly reduced CPU time, establishing a high-performance and scalable framework ideally suited for THz-based nondestructive testing and real-time imaging applications.

Vibrothermography using vibration excitation at specific frequency to activate a resonance in a defective area (local defect resonance, LDR) is promising for magnifying vibration induced heating and facilitating the detection of defects. However, the technique is limited by the requirement for a knowledge of vibrational responses to determine the resonance frequency. In this paper, a method for the automated determination of an optimal probing frequency only from surface temperature is proposed. A 3D finite element model and an experimental setup are established. The rate of temperature rise is proposed to better indicate the occurrence of LDR of an artificial delamination in carbon fiber reinforced polymer (CFRP). The defect-to-background contrast (DBC) is defined to quantify the enhancement of thermal imaging. Results from long-pulse vibrothermographic experiments show that the highest signal-to-noise ratio of delamination detection is achieved when the probing frequency is selected at the peak of DBC calculated from the rate of temperature rise. The identified probing frequency is stable for various bandwidths of sweep. The proposed method can improve the signal-to-noise ratio of thermal imaging of delamination in CFRP.

Sparse-view computed tomography (CT) can reduce acquisition times, supporting inline industrial inspection in suitable settings. In practice, scan time may also be shortened by lowering exposure per view, using faster detectors or motion systems, or leveraging partial/parallel acquisition; here we focus on reducing the number of projections. Fewer projections, however, can introduce streak artifacts that cause measurement deviations during metrological evaluations. This paper presents two self-supervised deep learning approaches using implicit neural representations (INR) to mitigate sparse-view artifacts and enhance measurement accuracy. Both methods represent the 3D object volume using a multi-layer perceptron (MLP) optimized individually for each scan through an incremental forward-backward strategy. The first approach, Neural Representation with Sparse-View Volume-based Loss (NR-SVOL), employs volume-domain training using an initial filtered back-projection (FBP) volume, enabling rapid artifact reduction with limited computational overhead. The second, Neural Representation with Sparse-View Projection-based Loss (NR-SPRO), directly optimizes the INR to match measured sparse projections, analogous to Neural Radiance Fields (NeRF), yielding superior artifact compensation at the expense of increased computation. Comprehensive evaluations were conducted on three industrial objects, a gear, a cylinder head, and a connector, at varying sparse-view configurations (32–256 projections). Both NR-SVOL and NR-SPRO demonstrated substantial artifact reduction, decreasing surface deviations by up to an order of magnitude in standard deviation. NR-SVOL achieved results within approximately five minutes, suggesting compatibility with some inline cycle times for our tested parts, while NR-SPRO delivered even higher accuracy when allowed more computation. This study highlights a practical trade-off between speed and precision, showcasing the potential of these methods for sparse-view inline industrial CT for improved metrological quality.

Manual evaluation and interpretation of Impact echo (IE) data is often labor-intensive and time-consuming, motivating the growing interest in applying machine learning (ML) techniques to this non-destructive testing (NDT) method. However, the scarcity of labeled datasets limits the generalizability of ML models to new, unseen data. This study investigates strategies to integrate a-priori knowledge into Convolutional Neural Networks (CNNs) for improved prediction of the S1 Lamb wave frequency from IE signals. To this end, time–frequency representations of IE signals, derived using the Short-Time Fourier Transform (STFT), are used as model inputs. A-priori knowledge is introduced in the form of initial frequency estimates obtained through manual evaluation. Additionally, transfer learning is employed to enrich the limited measurement dataset with data from 2D numerical simulations. The results demonstrate that, although training loss curves remain similar across models, incorporating additional information significantly enhances performance on unseen datasets. Furthermore, pre-training with simulation data accelerates convergence during early fine-tuning stages. The highest predictive accuracy was achieved when the initial guess was directly embedded into the loss function.

This article examines the effectiveness of non-destructive testing (NDT) in assessing wood under impact loadings. Our research was to evaluate the feasibility of using the frequency resonance technique (FRT), to predict the behaviour under impact of thermally modified timber (TMT) compared with a control sample of untreated wood. Wooden planks from five different species were subjected to a thermal modification process (TMP) under two different regimes. Both the TMT and control samples were evaluated using NDT to measure their dynamic modulus of elasticity (MOED), logarithmic decrement of damping (LDD) and acoustic conversion efficiency (ACE). Subsequently, wood samples from the same species were tested using drop-weight impact tests to measure their inflicted maximum force and impact bending strength (IBS), while high-speed cameras recorded the impacts to measure the maximum deflection of the specimens. The results revealed that the only relatively efficient prediction of FRT was the relationship between MOED and IBS. The ACE and LDD results did not show any acceptable correlations with impact tests, indicating that NDT is not reliable for assessing maximum force and deflection in the wood species under impact. Our study also found that the efficiency of the results and predictions were influenced by the wood species and the TMP conditions, necessitating a large number of samples for each species and heat modification temperature to achieve accurate NDT results. Our study found that the efficiency of NDT predictions was significantly influenced by both wood species and the TMP conditions. Specifically, oak showed a relatively higher coefficient of determination, while ash had the lowest. The thermal treatment also had a varied effect on NDT's ability to determine IBS, increasing its efficiency for larch specimens while decreasing it for ash and beech, with no significant effect on oak and spruce. These findings imply that future NDT methodologies must be developed with a species-specific approach and calibrated for each unique modification condition. Consequently, achieving accurate NDT results will require comprehensive data sets with a large number of samples for each species and heat modification temperature.

Diffuse ultrasound is a promising technique for estimating the depth of surface-breaking cracks in concrete. However, practical use in the field has been limited by the necessity of establishing the crack depth-lag time relationship, only derived by time-consuming finite element simulations. Running these time-consuming simulations on-site is impractical, especially when rapid assessment of damage over large areas is crucial. This research addresses this limitation by recently proposed theoretical diffuse energy velocity concepts, to directly correlate with depth of surface-breaking cracks. Existing datasets for artificial notches in concrete specimens and actual surface-breaking cracks in reinforced concrete beams subjected to four-point bending are utilized to evaluate the performance of the proposed method. The results indicate that the diffuse ultrasonic method based on the diffuse energy velocity provides more accurate crack depth predictions compared with conventional approaches. More importantly, the simplicity of using the theoretical diffuse energy velocity approach eliminates the need for time-consuming finite element simulations, enabling rapid, on-site crack depth measurements. This enhancement significantly improves the utility of the diffuse ultrasonic method for field applications. Therefore, this research highlights the potential of the diffuse ultrasonic method, enhanced by the theoretical diffuse energy velocity approach, to serve as a reliable and efficient commercial tool for field inspections of concrete structures.

The aging of transportation infrastructure has highlighted the need for reliable bridge deck assessment methods. Among various non-destructive technologies, ground penetrating radar (GPR) stands out for its ability to evaluate both concrete sections and reinforcement conditions without structural interference. While GPR offers significant advantages over traditional inspection methods such as chain dragging, data interpretation remains challenging due to signal complexity and environmental factors. Recent advances in signal processing, machine learning, and artificial intelligence (AI) have opened new possibilities for enhancing GPR data interpretation and automation. This review paper synthesizes and critically examines recent developments in GPR data analysis for bridge deck evaluation, from conventional signal processing to emerging computational approaches. Advances in five key areas are explored: basic GPR processing algorithms, traditional concrete evaluation methods, machine learning applications in concrete assessment, conventional reinforcement analysis techniques, and artificial intelligence-based reinforcement evaluation. Traditional methods and emerging AI approaches each offer distinct capabilities, with traditional techniques providing the foundation for targeted assessments, while machine learning and deep learning techniques introduce new potential for automated analysis. Studies across various test beds reveal that performance metrics are strongly influenced by testing conditions, data acquisition parameters, and structural characteristics. This diversity in reported outcomes highlights both the significant progress made in GPR data analysis and the continuing challenges in achieving reliable results across varied field conditions.

Accurate semantic segmentation of point cloud data is vital for safety monitoring and intelligent management in tunnel construction. However, challenges such as cluttered environments, occlusions, and the lack of annotated domain-specific datasets hinder effective application of deep learning techniques. To address these issues, this study presents TCS-Net, a novel point cloud segmentation network tailored for under-construction tunnels. A large-scale annotated dataset, named 3D Tunnel, was constructed using handheld laser scanning and contains over 60 million points across eight structural categories, filling a critical data gap in the field. TCS-Net introduces a multi-module fusion framework that combines spatial attention mechanisms, an inverted residual MLP (InvResMLP) for enriched feature representation, and a Kd-tree–based Gaussian upsampling with channel attention for enhanced feature propagation. An optimized training strategy incorporating AdamW, cosine decay, and label smoothing further improves learning robustness. Experimental results on the 3D Tunnel dataset demonstrate that TCS-Net achieves superior segmentation performance, with 94.38% mean IoU and 98.23% overall accuracy, validating its effectiveness and practical potential in tunnel construction scenarios.

This article presents a high-speed z-shaped laser point scanning thermography detection method and its corresponding experimental system. The system is designed for the rapid inspection of specimens. By analyzing the thermal response of the z-shaped scan thermography, a thermal response signal reconstruction method based on the restored pseudo heat flux (RPHF) theory, which is autocorrelated RPHF (RPHF-autocorr), is proposed, and the principle of the process is discussed. Additionally, the proposed method corrects the speckle and time offsets in the point laser scanning infrared thermal imaging. Experiments were conducted on carbon fiber reinforced polymer (CFRP) composites, where the raw thermal images are obtained and processed using the proposed algorithm. The signal-to-noise ratio (SNR) of defects is calculated and used to evaluate the defect detectability. The results are compared to the pseudo-static matrix reconstruction time truncation (PSMRT) and PSMRT pulse phase thermography (PSMRT-PPT). The experimental results show that the RPHF-autocorr algorithm provides a relatively high SNR. In summary, the RPHF-autocorr algorithm significantly improves the SNR of defects, increases defect detection sensitivity, and reduces the impact of lateral thermal diffusion caused by the moving laser spot.

The analysis of secondary quantitative data extracted from high-resolution synchrotron X-ray computed tomography scans represents a significant challenge for users. While a number of methods have been introduced for processing large three-dimensional images in order to generate secondary data, there are only a few techniques available for simple and intuitive visualization of such data in their entirety. This work employs the AccuStripes visualization technique for that purpose, which enables the visual analysis of secondary data represented by an ensemble of univariate distributions. It supports different schemes for adaptive histogram binnings in combination with several ways of rendering aggregated data and it allows the interactive selection of optimal visual representations depending on the data and the use case. We demonstrate the usability of AccuStripes on two high-resolution synchrotron scans of particle-reinforced metal matrix composite samples, each containing millions of particles. Through AccuStripes, detailed insights are facilitated into distributions of derived particle characteristics of the entire sample. Furthermore, research questions such as how the overall shape of the particles is or how homogeneously they are distributed across the sample can be answered.