Journal of Nondestructive Evaluation最新文献

Wind turbine blades (WTBs) have increased in size and complexity, resulting in higher operational demands and maintenance costs. Damage to these blades can significantly reduce turbine performance, lifespan, and power generation, while increasing safety risks. Effective structural health monitoring (SHM) is therefore essential for early damage detection and failure prevention. This paper presents a comprehensive review of various SHM techniques for WTBs, categorizing each technique into sensing methods (data acquisition) and analysis methods (data processing and interpretation). The review also addresses the causes and types of blade damage, severity ratings along with corresponding maintenance actions, and fatigue-induced damage progression. Advanced approaches, including machine learning, signal processing, hybrid methods, and emerging techniques such as piezo-based active sensing, electromechanical impedance, and Lamb wave tomography, are also explored for their potential to enhance SHM capabilities. Additionally, commercially available SHM systems and inspection platforms, such as unmanned aerial vehicles, are reviewed to highlight practical applicability. The review covers strain-based methods, acoustic emission, vibration analysis, thermography, ultrasonic testing, radiography, machine vision, and electromagnetic techniques, highlighting their advantages, limitations, and future research directions for improving SHM for WTBs.

A method is proposed for increasing the signal-to-noise ratio SNR of surface eddy current magnetic permeability meters with orthogonal rectangular coils for a moving strip of ferromagnetic rolled metal. As a result of robust parameter design of the probes, their optimal structures are determined using the design of experiments theory. Specific examples demonstrate the effectiveness of the method, which, in addition to increasing the SNR, also simultaneously reduces variations in the output signal caused by interfering factors. To create a Taguchi design of experiments, a magnetodynamic model of the probe over a moving test object was used along with orthogonal arrays. Numerical computer simulations confirm the reliability of the identified optimal structures of eddy current probes. Using variance ANOVA analysis, the ranks of influence of design and operating parameters on the SNR of probes were established. Based on this, technical requirements for the accuracy of manufacturing their structures and conditions for maintaining the stability of the probe’s operating modes have been formulated.

Pork is an important meat product for the European Union, which exported over 4.2 million tons in 2023, valued at €8.1 billion. Automating the labor-intensive deboning process is of significant interest, particularly through the development of advanced inline inspection systems capable of analyzing pork shoulder bone structures. While computed tomography (CT) systems provide high-contrast 3D reconstructions, their large size and high-cost present substantial barriers to adoption in industrial meat processing. This study addresses these challenges by introducing a novel approach that uses a single X-ray projection in combination with deep neural networks to predict the 3D segmentation map of pork shoulder bone structures using conventional reconstruction algorithms. To this end, U-Net neural network variants were trained on high-resolution CT scans of 90 pork shoulders. These scans were augmented with synthetic data to simulate different orientations on a conveyor belt, ensuring the model’s robustness. The minimum number of X-ray projections needed for accurate reconstruction was determined based on simulations, and 60 evenly spaced projections between 0° and 180° were found optimal. The Feldkamp-Davis-Kress (FDK) algorithm was chosen for its efficiency and cost-effectiveness in inline processing. The model achieved a Dice score of 0.94 and an SSIM of 0.96 on test data, demonstrating its ability to predict 59 missing projections and reconstruct the 3D bone structure accurately. The method that is proposed in this paper has the potential to advance meat processing by enhancing deboning precision, reducing waste, and streamlining operations. Our best model achieved an average dice score of 0.94 ± 0.03 and a maximum voxel-wise error of 16 mm on our test set, indicating high segmentation accuracy and spatial consistency in bone structure reconstruction. While the results are promising, the current evaluation is based on synthetic X-ray projections. Future work will focus on validating the method with real inline acquisitions and assessing its impact on cutting precision and waste reduction in robotic deboning.

Accurate identification of structural states during the hoisting of high-speed railway (HSR) precast box girders is essential for ensuring construction safety. However, efficient identification of girder status during hoisting remains a major challenge due to the complexity of mechanical responses at lifting points. To explore the feasibility and effectiveness of acoustic emission (AE) monitoring in identifying hoisting states, this study proposes a status recognition method that integrates AE sensing at lifting points with an optimized classification model. Specifically, (1) AE signals are collected during various hoisting scenarios of standard 32.6 m HSR box girders to capture characteristic changes in signal features such as amplitude, energy, ring counts, etc. (2) Based on a comprehensive feature set extracted from the AE signals, a light gradient boosting machine (LGBM) classification model optimized via Bayesian algorithm is developed for hoisting state recognition. Experimental tests conducted in a prefabrication yard demonstrate that the proposed method effectively distinguishes different hoisting conditions, particularly capturing potential anomalies caused by lifting asynchrony or stress concentration. The results validate the applicability of AE technology for non-invasive, efficient status identification during girder hoisting, providing a technical foundation for the intelligent monitoring of construction safety.

Open-loop permanent magnet magnetizers are widely employed in magnetic flux leakage (MFL) testing of steel wire ropes due to their structural simplicity and operational convenience. However, the underlying magnetization mechanism remains inadequately understood, and systematic investigations into their optimal structural configurations are still lacking. In this study, a combined approach of finite element simulation and experimental validation is utilized to systematically examine the influence of axial length on the magnetization performance of ring-type permanent magnet magnetizers for steel wire ropes modeled as equivalent steel tubes. The results reveal a nonlinear relationship between axial length and the peak axial magnetic flux density in air: the intensity initially increases and then decreases with length. In contrast, both the magnetic field uniformity and the effective magnetization region improve monotonically with increasing length. These two mechanisms jointly modulate the internal magnetization state of the tube, giving rise to an optimal axial length. Further analysis confirms that for a steel wire rope with a diameter of 50 mm (or its equivalent tube), the optimal axial length range of the open-ring magnetizer lies between 95 mm and 110 mm, within which an optimal balance between field strength.

Real-time monitoring of bridge deterioration remains a major challenge in structural health monitoring (SHM), as traditional inspection methods lack sensitivity to early-stage damage and cannot provide real-time evaluation. This study aims to develop a practical approach for damage stage assessment of prestressed hollow-slab bridges using acoustic emission (AE) parameters and a lightweight machine learning model. First, scaled model tests were conducted to collect AE signals under different loading stages, and parameters such as amplitude, energy, duration, ring-down count, and impact velocity were extracted. Second, correlation analyses (Pearson, Spearman, Kendall) were performed to identify the most representative parameters for structural degradation. Third, a K-nearest neighbors (KNN) model was then constructed to classify bridge damage stages in real-time. Finally, comparative analysis with decision tree, support vector machine (SVM), one-dimensional convolutional neural network (1DCNN), and long short-term memory (LSTM) models was conducted. Impact Velocity was determined to have the strongest correlation with damage stages and the KNN model achieved competitive accuracy (0.9103) with superior computational efficiency (1.30 × 10⁻⁶ s per sample), offered the best computational efficiency, making it highly suitable for real-time applications. This research establishes an efficient, data-driven framework for real-time bridge health monitoring, demonstrating the practical viability of combining AE technology with lightweight machine learning for structural damage assessment. The methodology shows particular promise for implementation in real-time monitoring systems for prestressed concrete structures.

Corrosion is a critical factor in the degradation of metallic structures, especially in sectors such as maritime, aerospace, and civil infrastructure. Traditional corrosion assessment techniques, while widely used, are limited by their point-based measurements, labor intensity, and susceptibility to human error. This study proposes an advanced non-destructive evaluation (NDE) method combining guided wave propagation with probabilistic imaging to assess global corrosion damage and surface roughness in metallic plates. The approach utilizes Lamb waves, which are sensitive to thickness variations and capable of propagating over large areas with minimal sensor deployment. Key indicators such as wavefront asymmetry, residual signal energy, and correlation-based metrics (RAPID) were analyzed both numerically and experimentally. Corroded plates were modeled using Gaussian random fields and validated through controlled electrochemical degradation experiments. Results demonstrate that guided wave-based indices can effectively detect and monitor corrosion progression, with sensitivity to both the degree of material loss and surface irregularity. Additionally, a reference-free method based on wavefront asymmetry showed potential for practical, in-situ applications. The findings confirm the viability of guided waves as a powerful tool for structural health monitoring, offering enhanced spatial coverage, automation potential, and early-stage damage detection capabilities.

This study investigates the impact of subsurface defects on surface temperature distributions through Pulsed Infrared Thermography (PIRT), employing a connection between MATLABl software and COMSOL Multiphysics for the simulation of 300 distinct defect depths, culminating in the analysis of 450,000 thermal images. The Hough Circle Transform (HCT) was applied to a specific time frame from the infrared video sequence, where defect signatures were pronounced, to precisely localize defect regions. For defect quantification, a normalization method was implemented, calculating the ratio of the thermal norm within defect regions to that of non-defective areas. A Long Short-Term Memory (LSTM) network was then trained to map the temporal evolution of this norm ratio to a specific defect detection time. A Photopolymer Resin sample, fabricated with varying depths of defects via Stereolithography Apparatus (SLA) 3D printing, underwent Non-destructive testing (NDT) using Pulsed Infrared Thermography (PIRT). An integrated algorithm was employed for the identification of defect regions and the subsequent calculation of the times at which defects were identified. The depths of unknown defects were predicted employing a least square fitted curve, derived from four samples with known defect depths. This methodology achieved a notable degree of precision. This research delineates a relationship between the depth of a defect and its detection time, introducing an innovative approach for the accurate prediction of defect depths within the realm of Non-destructive testing (NDT).

This study presents a comparative evaluation of statistical and machine learning (ML) models for predicting moisture content (MC) in lightweight foamed concrete (LFC) using nondestructive microwave reflection parameters. Five models were examined: Multiple Linear Regression (MLR), Support Vector Machines (SVM), Random Forest (RF), Levenberg-Marquardt Neural Network (LMNN), and Radial Basis Function Network (RBF). Two dataset formats were used: a frequency-structured dataset (FSD) capturing full-spectrum information, and an instance-based dataset (IBD) designed for single-frequency applications. Model performance was assessed using R², RMSE, MAE, training time, and prediction speed, with five-fold cross-validation applied to evaluate generalization. Results showed that ML models outperformed the statistical model in capturing non-linear relationships. RF and LMNN achieved the highest accuracy and stability across both datasets, while RBF and MLR showed signs of overfitting, especially on FSD. Sensitivity analysis using permutation feature importance revealed that S11 magnitude was most influential in structured data, whereas input importance was more evenly distributed in IBD. The findings emphasize the importance of aligning model choice with dataset structure to improve accuracy and robustness. This study supports the development of real-time, nondestructive moisture monitoring systems in LFC for sustainable construction applications.

In order to mitigate the loss caused by hole damage during the service of carbon fiber-reinforced polymer (CFRP) axle tubes, this paper proposes a modal parameter-based approach to identify hole damage. The method focuses on single-hole, double-hole, and triple-hole damage as the objects of study, with fiber Bragg grating sensors for data collecting and strain mode shapes serving as the indicator for damage determination. The damage area of the axle tubes is localized based on the difference in strain mode shapes, and the degree of damage is identified using deep neural networks (DNN). The results indicate that the method of identifying the hole damage of CFRP axle tubes based on modal parameters is highly accurate, with all damage locations reliably identified, and the maximum relative error in damage degree identification is -12.95%. This study is highly significant for enhancing maintenance efficiency and prolonging the service life of CFRP axle tubes.