Journal of Nondestructive Evaluation最新文献

This work presents an investigation of acoustic emission (AE) behavior during compression-after-impact (CAI) tests on thermoplastic composites subjected to different impact energy levels. AE sensing is employed to detect and evaluate damage that may not be immediately visible. To the best of the authors’ knowledge, existing literature provides limited insight into CAI performance of thermoplastic composites, especially under relatively high impact conditions, an important gap given the rising use of thermoplastics in advanced air mobility applications. The primary objective is to analyze AE signals recorded during CAI tests and characterize their features across various impact energies, with the longer-term goal of enabling applications in advanced methods such as machine learning and artificial intelligence. For each test, results include time-dependent signal density, peak frequencies, and amplitudes, along with cumulative signal strength (CSS) to track the progression of damage at each impact level. The discussion further explores correlations between AE features and includes metrics like peak frequency density to highlight the relative influence of different features and link specific frequency ranges to distinct failure modes. Additionally, optical microscopy revealed four main failure mechanisms: matrix cracking, delamination, debonding between fiber and matrix, and fiber breakage.

In the nondestructive testing industry, probability of detection (POD) studies can be prohibitively expensive because of specimen and testing costs. In model-assisted probability of detection (MAPOD) analysis, physics-based model simulations are used to reduce the number of specimens needed to compute POD curves, but current MAPOD practices require precision simulations. These simulations need to have high enough accuracy to be treated as equivalent to real experimental data, perhaps after calibrating a “transfer function” that corrects their output. Not all simulators will have that level of accuracy, but when there are multiple inspection scenarios, the simulator accuracy can be assessed as part of the MAPOD process. This paper proposes Model-Quality-Adaptive POD (MoQuAPOD) which adapts to the demonstrated quality of a simulator across multiple inspection scenarios. The approach extends the traditional MAPOD concept of a simulation-to-experiment transfer function by acknowledging the uncertainty in its estimation. Stochastic transfer functions are evaluated and calibrated as a population over the range of inspection scenarios as part of the POD process. A hierarchical Bayesian model uses simulator predictions to draw strength across that population, reducing the number of specimens required to achieve a particular POD and confidence. We obtain POD estimates and population statistics from a Markov Chain Monte Carlo (MCMC) sampler. Population uniformity implies trustworthiness of the simulator, so that trustworthy simulators need less experimental data. An example illustrates model-quality-adaptive multi-scenario POD methods reducing the required number of specimens by 25%, compared to single-scenario POD methods, using synthetically generated data.

In the design of lightweight and extended-span structures, cable-based systems have been widely used and may be the only solution, particularly for superlong-span bridges and spatial structures. Prestressed anchor cables are the most crucial element in cable-based structures and directly influence their performance. However, prestressed anchor cables are prone to corrosion and fatigue due to environmental factors, complex loads, and chemical influences. Therefore, to ensure safety, developing a real-time monitoring method to evaluate the status of the prestressed cable is urgently needed. This paper introduces Acoustic Emission (AE) integrated with machine learning to enhance the diagnosis and prognosis of prestressed anchor cables. In the laboratory, twelve cables with varying defects were tested to failure to optimize the machine learning framework. Meanwhile, AE signatures due to damage progression were characterized and identified. Moreover, a machine learning framework for prestressed anchor cables was developed in terms of k-means and k-Nearest Neighbor (K-NN) clustering to distinguish AE signatures precisely due to different AE sources from large-scale data. A precursor signal for cable fracture was also extracted and recommended for early-stage warning. Furthermore, the in-situ recorded AE signal has validated the proposed framework. This study provides guidelines for using AE as a valuable tool for the Structural Health Monitoring (SHM) of prestressed anchor cables.

Detection of debonds in honeycomb sandwich-type structures has been a subject of interest for many researchers. However, detection and quantification of debonds in sandwich structures by methods adaptable for structural health monitoring is still being pursued. Low frequency guided waves are being used for the detection, localization and quantification of the debonds, with some limitations, like quantification of the damage being dependent on the distance of the sensor from the debond, need for a reference signal from the pristine structure etc. The present work investigates the potential of the low frequency guided waves in the detection, localization and quantification of debonds in Honeycomb Sandwich Structures (HSS) and addresses some of these limitations. A unique debond quantification curve for the given HSS is generated using a 3D finite element model and validated experimentally. Unlike in earlier works, these curves are independent of the distance of the sensor from the debond. The methodology is demonstrated in pulse-echo and pitch-catch configurations, and it does not need a reference signal from the pristine structure. The developed method is effective in detecting and quantifying the debond located on both the face skins of the sandwich, with the sensor mounted only on one face skin. An efficient methodology to assess the size of the debond is proposed, based on the results obtained from this study. The guided waves are actuated and sensed by Lead Zirconium Titrate (PZT) transducers which facilitate the implementation of structural health monitoring.

In intelligent nondestructive evaluation (NDE), overfitting on small datasets poses a significant limitation to the generalization of ultrasound classification models across different materials. Consequently, the development of effective regularization techniques is crucial for designing robust multi-material NDE systems. In this paper, we propose a versatile and lightweight dual-regularization module comprising two sub-modules: Acoustic Velocity-Guided Dropout (AVGD) and Squeeze-and-Excitation (SE) attention. The AVGD integrates physical domain knowledge with conventional regularization methods by dynamically adjusting the dropout rate of feature channels based on acoustic velocity information. Meanwhile, the SE attention mechanism enhances critical features in conjunction with dropout, thereby improving the model’s learning capacity. Both sub-modules are encapsulated into a dropout layer and an SE block, respectively, and seamlessly integrated into a classical neural network architecture. The proposed method is evaluated on a collected ultrasound signal dataset and compared against standard regularization mechanisms. Experimental results demonstrate that the dual-regularization mechanism significantly enhances the generalization capability of the baseline model.

This study systematically addresses critical challenges associated with blurred edges of wafer defects, including multi-dimensional feature aggregation, abrupt gradient decline, and hierarchical information loss. To tackle these issues, an innovative, precise segmentation method is proposed based on a dual cross-attention mechanism and a feature gradient histogram. Through in-depth analysis of edge-blurring characteristics in wafer defects, a multi-scale embedding matrix equation was formulated to optimize the contour extraction process. Additionally, a multi-level encoder architecture was implemented to enhance the efficiency of edge contour information extraction. To address boundary information loss during segmentation, a boundary gradient optimization model was constructed using multi-scale differential equations, enabling accurate fitting of boundary gradients through feature recombination vectors. Experimental results demonstrate the effectiveness of the proposed wafer defect segmentation approach. The method achieves an average accuracy of 97.51%, with mean intersection-over-union (mIoU) scores exceeding 89% across three distinct types of wafer defect detection tasks. By effectively mitigating the adverse effects of edge blur on segmentation precision, this approach offers a comprehensive solution for wafer defect detection. The contributions of this research not only enhance the accuracy and reliability of defect identification but also provide robust technical support for improving product quality and manufacturing efficiency in high-end semiconductor industries. These advancements hold significant practical value in promoting the high-quality development of the semiconductor sector.

X-ray computed tomography (XCT) is essential for nondestructive evaluation and quality control of large-scale metal components. XCT imaging, however, faces significant challenges from metal artifacts, particularly those caused by Compton scattering, which degrade image quality and obscure critical details. Hardware-based solutions (e.g. scatterControl) offer advancements by intercepting scattered photons and reducing artifacts, but they can be time-consuming and require additional processing. Here, we propose modifying and leveraging a novel deep learning (DL) framework, Simurgh, to enhance and accelerate scatter correction in XCT. By combining scatterControl with DL-based artifact removal, we demonstrate significant reduction in scan time while producing high-quality reconstructions. Through extensive evaluation on industrial XCT data, we show that our methods reduce scan time by up to more than 10(times ) while preserving flaw detectability. Quantitative analysis across multiple segmentation techniques confirms that Simurgh-based reconstructions consistently outperform traditional Feldkamp-Davis-Kress, model-based iterative reconstruction, and commercial DL models in both pixel-level and task-specific evaluations, enabling scalable, high-throughput XCT workflows for characterization of large scale components in applications such as casting and metal additive manufacturing.

Cone-beam X-ray computed tomography (XCT) is an essential imaging technique for generating 3D reconstructions of internal structures, with applications ranging from medical to industrial imaging. Producing high-quality reconstructions typically requires many X-ray measurements; this process can be slow and expensive, especially for dense materials. Recent work incorporating artifact reduction priors within a plug-and-play (PnP) reconstruction framework has shown promising results in improving image quality from sparse-view XCT scans while enhancing the generalizability of deep learning-based solutions. However, this method uses a 2D convolutional neural network (CNN) for artifact reduction, which captures only slice-independent information from the 3D reconstruction, limiting performance. In this paper, we propose a PnP reconstruction method that uses a 2.5D artifact reduction CNN as the prior. This approach leverages inter-slice information from adjacent slices, capturing richer spatial context while remaining computationally efficient. We show that this 2.5D prior not only improves the quality of reconstructions but also enables the model to directly suppress commonly occurring XCT artifacts (such as beam hardening), eliminating the need for artifact correction pre-processing. Experiments on both experimental and synthetic cone-beam XCT data demonstrate that the proposed method better preserves fine structural details, such as pore size and shape, leading to more accurate defect detection compared to 2D priors. In particular, we demonstrate strong performance on experimental XCT data using a 2.5D artifact reduction prior trained entirely on simulated scans, highlighting the proposed method’s ability to generalize across domains.

This study investigates the trade-off between minimizing wall thickness and through-hole formation in AlSi10Mg thin hollow lattice structures produced via laser powder bed fusion. X-ray computed tomography (XCT) is employed as a metrological tool to evaluate the effects of laser linear energy density (LED) across conditions ranging from under-melting to over-melting using a single laser track strategy. An XCT-based algorithm is developed for automated through-hole detection, providing quantitative data on through-hole count and size. The algorithm's capability is evaluated through leakage tests. The substitution method, adapted from ISO 15530–3 for tactile coordinate measuring machines (CMM), is employed to assess XCT measurement uncertainty for hollow lattice dimensions. As a new addition to the conventional substitution method, the effects of high-density data generated by XCT are assessed against the calibrated diameters obtained from low-density CMM data and used for the calculation of wall thickness. Experimental results show that under-melting conditions can produce wall thicknesses of 0.135 mm to 0.212 mm, with an exponential increase in through-hole formation as LED decreases. A linear relationship between LED and wall thickness is observed, enabling identification of optimal parameters for producing defect-free thin-walled structures.

In response to the low defect detection accuracy caused by small defect areas and large differences in defect scales in EL images of photovoltaic cell components, a defect detection algorithm for photovoltaic cell modules based on traditional image processing and deep learning is proposed. Firstly, a traditional image processing algorithm is designed to segment the images of the cell modules into individual solar cells for detection. Secondly, the accuracy of defect detection is improved by enhancing the YOLOv8 network. The specific details are as follows: First of all, a dynamic receptive field selection structure called C2DLSK (C2f and Dynamic Large Selective Kernel Module) is designed to replace the C2f module in the backbone. It dynamically selects the appropriate receptive field size for the current target during the feature extraction process to more accurately extract the features of defects. Then the CARAFE (Content-Aware ReAssembly of Features) is used to replace the first nearest-neighbor upsampling module in the neck. At the same time, a bidirectional weighted fusion method called BiConcat is used for feature fusion, which fully utilizes semantic information while enhancing the weight of important features in feature fusion. Finally, the MPDIoU loss function is used to replace the CIoU loss function, further improving the accuracy of defect detection. The experiment shows that under the condition of ensuring real-time detection, the average precision mean average precision (mAP) of this algorithm for defect detection in photovoltaic cell components reaches 85.8%, which is an improvement of 1.9% compared to the original network. Compared with the current mainstream YOLOv3-tiny, YOLOv5s, YOLOv7-tiny, and YOLOv8s, it improves the detection accuracy of photovoltaic cell components by 5.3%, 2.9%, 1.6%, and 0.9% respectively.