Journal of Nondestructive Evaluation最新文献

The belt conveyor is an important continuous transport device in modern industrial production. The conveyor belt, a crucial part of the belt conveyor, is vulnerable to damage since it works for lengthy periods of time at high speeds and large loads. If these damages are not detected and addressed in a timely manner, they may hasten the conveyor belt’s wear and even lead to safety accidents. This paper suggests a conveyor belt damage detection and segmentation network, BDSE-YOLO, based on an enhanced YOLOv11, to address the problems of low detection accuracy, poor real-time performance, and insufficient adaptability to complex backgrounds in the current conveyor belt damage detection methods. First, the YOLOv11 architecture is optimized by introducing the ACmix module in the feature extraction module. A new C2PSA_ACmix module is designed to leverage the self-attention characteristics of the ACmix module, enhancing the network’s capacity to extract both local and global characteristics, thereby improving the performance of damage segmentation and detection, particularly for small or complex damages. Additionally, the iRMB module is added to the backbone network to enhance information flow. This module captures long-range dependencies while maintaining the lightweight nature of the network, enhancing the efficiency and accuracy of segmentation tasks. On this basis, a damage evaluation method based on geometric features and size quantification is proposed. The rupture direction is determined using an ellipse fitting algorithm, while size quantification techniques are employed to accurately analyze the damage morphology and eight quantification indicators are established. Experimental results on a self-made dataset and two public datasets demonstrate that the suggested model attains 96.2%, 81.0% and 92.7% accuracy rates, respectively, outperforming the comparison models and demonstrating high detection accuracy and robustness. The model exhibits strong adaptability in complex industrial environments, and the eight proposed evaluation indicators provide reliable criteria for evaluating rupture propagation trends and the severity of damage. The proposed network and method offer an effective solution for the intelligent detection and evaluation of damage to conveyor belts.

Spectral X-ray computed tomography enables material decomposition by leveraging energy-dependent X-ray attenuation properties. However, material decomposition with spectral CT requires a longer acquisition time to obtain sufficient numbers of photons in each energy bin. Sparse-view offers a practical solution to reduce acquisition time, but it introduces ill-posedness, degrading decomposition accuracy. This study introduces a material decomposition framework based on Neural Radiance Fields where material maps are represented using a multilayer perceptron (MLP). The material maps are then optimized through a spectral forward projection process based on the Lambert–Beer’s law, while a partition of unity (PoU) loss ensures the physical constraint on material maps. Our method was evaluated using simulated and real spectral CT datasets and compared with a traditional statistical approach. The results demonstrated that our method performs well in material decomposition under sparse-view conditions. The results suggest that our “neural material composition fields” framework offers accurate material decomposition robust to sparse-view conditions without requiring labeled training data.

Carburization is a critical degradation mechanism in high-performance (HP) steel furnace tubes, impairing structural integrity during prolonged high-temperature service. This study proposes a machine learning-assisted ultrasonic testing framework to classify four levels of carburization damage in Cr‒Ni‒Nb HP steel alloys. A total of 80 A-scan signals were acquired per frequency (2.25 and 5 MHz) across four distinct damage classes, with spectral features extracted via discrete cosine transform (DTC). Microstructural analysis confirmed a linear increase in the volumetric fraction of chromium carbides from 9.5% (SP01, low carburization) to 40.5% (SP04, severe carburization). Among the classifiers evaluated, the K-Nearest Neighbors (KNN) and Quadratic Support Vector Machine (QSVM) achieved 100% accuracy (AUC = 1.00) at 2.25 MHz for advanced damage levels. However, early-stage detection remained challenging, with GNB attaining only 83.1% accuracy and AUC = 0.91 for SP01. Classification performance deteriorated significantly at 5 MHz due to increased signal attenuation and noise, with accuracy falling to 47.3–53.5%. These findings underscore the influence of ultrasonic frequency on damage detectability and model reliability. The integration of frequency-optimized ultrasonic inspection with machine learning delivers a scalable approach for real-time, non-destructive monitoring of carburization in industrial HP steel components, offering critical insights for predictive maintenance and structural health assessment.

This paper presents a method to determine thermal parameters, which can neither be calculated analytically nor measured directly. Specifically, the convective heat transfer coefficient is discussed, which is usually determined using empirical models, namely the Nusselt correlations. To overcome the lack of information about the relation between temperature and parameters, the methodology of data assimilation is applied. Therefore, a dynamic model is combined with measurement data from a thermography experiment avoiding interference with the actual process enabling inline parameter identification. The method is first applied on artificially created simulation data and second on real measurement data. This paper shows that the developed method estimates the heat transfer coefficient in agreement with the well-known Nusselt correlations. Moreover, the present work compares different estimation strategies and gives a recommendation regarding state-parameter, pure parameter or combined estimation, including a detailed analysis of the variation of a smoothing parameter.

Si3N4 bearing roller surface defects are characterized by complex gray-scale texture features and diverse morphology. An integrated support vector machine (SVM) classification method based on adaptive inertia weight particle swarm optimization (AIW-PSO) is proposed in this paper to achieve comprehensive classification of Si3N4 bearing roller surface defect images. By analyzing the complex feature information of these defect images, an ensemble SVM-based classification model is designed. To realize high-precision classification, the model’s prediction accuracy is improved through integrated learning. The AIW-PSO method effectively avoids underfitting during model training by optimizing SVM hyperparameters. The method significantly realizes the accurate classification of Si3N4 bearing roller surface defect images and greatly improves the performance of the classification model.

This research proposes an approach for integrating realistic defects into computed tomography (XCT) scans by using X-ray simulations. It allows full control over different scenarios and measuring the detection algorithm efficiency in real-world situations. Using real XCT data of a pin-fin cooler made of aluminum alloy with complex internal structures, synthetic spherical and irregular defects ranging from 56 (upmu )m to 300 (upmu )m in diameter are superimposed to create a comprehensive dataset that mimics a wide range of realistic scenarios. This XCT dataset with superimposed defects is then utilized to apply a probability of detection analysis to detect defects of varying sizes and shapes. This analysis shows that for spherical pores, the detectability limit is up to 2.5 times higher in the superimposed case with a minimum voxel similarity of 95%, while for irregular pores, this limit is 3.3 times higher when a minimum voxel similarity of 80%. The integration of synthetic defects into real XCT images allows for a more rigorous and controlled assessment of detection algorithms, providing valuable insights into their performance under realistic conditions. Our findings demonstrate that this method can significantly improve the accuracy and reliability of measurements of defect detectability, offering a powerful tool for quality assurance in critical manufacturing processes.

In this study, the coating thickness of a double-sided coating processed by cathodic electrodeposition was measured and visualized using the acoustic resonant imaging technique. This method analyzes the resonant frequency and amplitude spectrum of echoes, enabling the visualization of thin coating thickness. Because changes in the coating thickness owing to the characteristics of the electrodeposition process are expected to influence the properties, we experimentally investigated the relationship between the coating thickness and the resonant frequency. An experimentally established relationship between coating thickness and resonant frequency indicated that a 33% reduction in thickness corresponded to an approximate 10% increase in sound velocity. Based on this relationship, the coating thicknesses on both surfaces were simultaneously visualized. Validation through cross-sectional microscopy confirmed the accuracy of the thickness measurements, with an average error of less than 2%, demonstrating the accuracy of this method for precise and simultaneous coating thickness evaluation.

X-Ray Computed Tomography (CT) is a widely adopted tool in the non-destructive quality assurance of additive manufacturing (AM). Porosity in AM can be assessed via CT without compromising the integrity of the part and without reliance on witness specimen. Reliable pore criticality analysis, essential for AM fatigue assessments, hinges on precise determination of pore dimensions. This work investigates CT data by comparing the pore sizes and shapes from two different data sources (CT and metallography), originating from the same samples. The comparison indicates a pore size underestimation in the CT data by an average of 20%. A subsequent rescaling and smoothing workflow on the CT pore data compensates this underestimation. This workflow reduces the mean pore size deviations between both data sources by up to 50% compared to the original data, allowing a more accurate pore assessment. Additionally the smoothing process reduces errors introduced by the CT reconstruction, lowering the average and scatter in mean curvature between pores. The rescaled and smoothed pores serve as an improved starting point for investigations regarding the effect of porosity on fatigue in AM.

Electromagnetic ultrasonic measurement of preload in highly attenuating superalloy bolts faces two critical challenges: the low ultrasonic signal amplitude and impracticality of measurement without prior knowledge of the bolt’s initial condition. To address these issues, this study proposes a dual-mode electromagnetic acoustic transducer (EMAT) featuring a new permanent magnet configuration. For superalloy bolt specimens, the proposed configuration not only resolves the challenge of insufficient Longitudinal Wave (LW) excitation inherent to traditional EMAT but also achieves a 280.78% enhancement in Shear Wave (SW) signal amplitude. The experimental results on bolt preload measurement demonstrate that the regression model established from measurement data obtained by the new permanent magnet-configured EMAT exhibits a 0.0672 higher R² coefficient compared to that of the traditional EMAT. In the comparative analysis of bolt preload measurement accuracy between mono-wave and bi-wave methods, the relative errors for SW mono-wave, LW mono-wave, and bi-wave methods are 0.45%, 0.18%, and 0.51%, respectively. The new dual-mode EMAT proposed in this study provides a robust methodology and critical data references for aerospace engine bolt preload monitoring.

Drilled shafts are the foundation of choice for heavily loaded structures, particularly in urban areas. However, their in-situ concrete casting process is vulnerable to the formation of foundation defects, requiring full-volume imaging of as-built drilled shafts for quality assurance. This study presents a novel two-dimensional (2D) acoustic full-waveform inversion (AFWI) method for high-resolution ultrasonic imaging of drilled shafts, capturing details both inside and outside the rebar cage at centimeter-scale resolution. The method is formulated using 2D acoustic wave equations and adjoint-state optimization, integrating Tikhonov and Total Variation (TV) regularizations to enhance solution stability and preserve sharp structural boundaries. Additionally, an approximate Hessian matrix is incorporated in the regularization gradient, significantly improving inversion accuracy, particularly in regions beyond the rebar cage. Validated through synthetic experiments, the method successfully reconstructs shaft boundaries and detects defects without requiring prior knowledge of design diameter. The mean radial boundary errors of 2.4 m diameter shafts without and with defect are 1.2 cm and 4.4 cm, respectively. To further evaluate its real-world performance, the method is applied to a full-scale drilled shaft measuring 2.4 m in diameter and 21.3 m in length. Experimental ultrasonic data are collected by the standard cross-hole sonic logging (CSL) at depths along the shaft length and inverted to obtain a 2D image of P-wave velocity (V_p) at each depth. Individual 2D V_p images are then combined into a 3D image of the whole drilled shaft. Results confirm that the AFWI approach effectively characterizes the entire shaft, providing high-fidelity imaging and precise boundary delineation with the mean radial error of about 3 cm. To our knowledge, this is the first reported application of full-waveform inversion on an actual drilled shaft, marking a significant advancement in quality assurance of cast-in-place foundations.