This study addresses the issues of high signal interpretation complexity and low efficiency in manual defect identification associated with alternating current field measurement (ACFM) technology in practical industrial inspections. A defect identification method that integrates a bidirectional long short-term memory (BiLSTM) network with an attention mechanism is proposed. By constructing a hybrid model with the capability for temporal feature selection, the method enhances the extraction of subtle defect features from ACFM signals. The experimental results show that on a test set containing 500 multi-condition signals (covering defects in 20# steel plates with depths ranging from 0.5 mm to 3 mm and lengths from 10 mm to 50 mm), the model achieves an accuracy rate of 92.10% ± 1.2%. Compared to traditional manual inspection, the average single determination time is reduced to 17.14 ms (an efficiency improvement of approximately 68%), with the smallest reliably identifiable defect size being 0.5 mm (depth) × 10 mm (length). Compared to the baseline long short-term memory (LSTM) model, the F1 score improves by 10.07 percentage points to 94.65% and the memory usage of the model is 2.28 MB, meeting the deployment requirements for industrial embedded devices. This method provides a new technical pathway for the automation of ACFM inspections.

In pipelines, weld regions are the primary areas for defect inspection and a critical guarantee for the validity of defect detection results. However, due to the characteristics of X-ray images of pipeline girth welds, such as their large length and complex background information, existing weld region segmentation techniques are prone to under-segmentation, error segmentation and discontinuous segmentation issues of the weld region. To address these problems, this paper proposes an intelligent segmentation technology for the pipeline girth weld region (PGWR) that fuses data-inherent knowledge. Initially, a deep attention adjust module fused with the U-Net network (DAAM-U-Net) is constructed to perform preliminary weld region segmentation, achieving accurate localisation and overall shape segmentation of the weld regions. Furthermore, constraints based on the data-inherent knowledge from X-ray images of weld regions are established to refine the segmentation, overcoming the challenges of under-segmentation, error segmentation and discontinuous segmentation. The proposed method is illustrated and validated using X-ray images of the PGWR. The results show that the segmentation performance of weld regions in complex backgrounds reaches a mean intersection over union (mIoU) of 94.52% and a Dice coefficient of 86.29%, not only demonstrating the effectiveness of the proposed method but also indicating that the approach of combining artificial intelligence (AI) technology with data-inherent knowledge effectively bridges the gap between AI technologies and their isolated applications in the non-destructive testing field.

Rails are a key component of railway infrastructure and their condition is crucial to the safe operation of trains. Ultrasonic guided wave technology is a method for long-distance detection and monitoring of rails. However, the complex structure of rails and the dispersion phenomenon of ultrasonic guided waves can lead to mode mixing, making it difficult to extract defect scattering signals and thus affecting the accuracy of defect detection. Therefore, this paper proposes an ultrasonic guided wave imaging method based on adaptive beamforming to improve the imaging accuracy of rail defects. Firstly, the guided wave signal is focused based on time reversal (TR). Then, according to the deviation between the reversed focused signal and the original excitation signal, the damage index (DI) for each propagation path is calculated. Finally, these indices are used as weighting vectors to perform adaptive weighting on the full matrix capture (FMC) signals for imaging. This method can narrow the main lobe width (MLW) and reduce the side lobe peak, thereby simultaneously enhancing the image contrast and resolution. Simulation and experimental results show that the proposed method can effectively suppress guided wave dispersion, increase the defect echo amplitude and significantly improve the imaging quality of rail defects. Compared with traditional total focusing method (TFM) imaging, the proposed method has significant improvements in the defect location, defect size and signal-to-noise ratio (SNR).

Welds are critical pipe connections, where weld defects can create stress concentrations and pose significant safety risks. The metal magnetic memory (MMM) method is widely employed for detecting pipeline weld defects. In this paper, the relationship between weld defects of different sizes and MMM signals is simulated by numerical simulation. The results show that the MMM signal of defects changes in a regular fashion with the defect size. Subsequently, the correctness of this numerical simulation conclusion is verified by experiments. A method based on the combination of a two-stream convolutional neural network (TSCNN) and a grid search (GS)‐optimised multi‐output least-squares support vector regression (MLSSVR) for welding defect size prediction is proposed to address challenges such as quantifying weld defects, the limited number of detection samples and the complexity of manual feature extraction. This method improves the quantification accuracy and reduces the computation time compared to traditional machine learning and provides a new method for welding defect quantification.

Traditional dam deformation prediction methods tend to ignore the influence of deformation in neighbouring areas on monitoring points and face the problem of long‐ and short‐term dependence, resulting in less accurate prediction results. For this reason, this paper proposes a dam deformation prediction method based on long short‐term memory and Transformer (LSTM-Transformer) that considers the spatiotemporal correlation of multiple measurement points. Firstly, the spatiotemporal correlation points related to the deformation of the monitoring points are mined using association rules and a dam deformation prediction model considering the spatiotemporal correlation of multiple measurement points is constructed based on the consideration of the influence of environmental factors. Secondly, the fusion of LSTM with Transformer enables the model to better capture global information in the sequence, which solves the short‐ and long‐term dependence problems in the dam deformation prediction model. Finally, the universality of the model and the superiority of the method is verified through two experiments.

Advanced ultrasonic techniques are now widely adopted in industry, though they present a risk of operators misunderstanding key features and misinterpreting results. The new phased array post-processing techniques, such as the total focusing method (TFM) and phase coherent imaging (PCI), have been implemented in the absence of agreed standards and consistent training. As human factors play a crucial role in the reliability of non-destructive testing (NDT), the effectiveness of these advanced techniques depends heavily on the proficiency of the inspector. The development of new equipment and techniques must be accompanied by appropriate training to mitigate the risk of human error. An assessment of inspectors’ perceptions regarding the reliability and implementation of phased array ultrasonic testing (PAUT) in weld inspection, with a specific focus on TFM and PCI, was conducted. The primary objective was to evaluate the industry’s proficiency, challenges and constraints in applying these advanced methods. Surveying Level 2 inspectors predominantly in the oil and gas sector, 76% of respondents have utilised TFM/PCI within the industry. However, 74% expressed little to no confidence in their ability to use these methods effectively. A significant concern highlighted was the inadequate training available, with over 75% of respondents indicating they had received little to no formal instruction on these techniques. This lack of comprehensive training underscores the importance of addressing human factors to enhance NDT reliability. Respondents also expressed a strong need for more detailed guidelines and standardisation across original equipment manufacturers (OEMs), specifically regarding pre-system set-up, including defining the region of interest (ROI) and selecting the correct wave set. Limited understanding of relevant codes and standards results in a shortage of qualified written instructions and structured procedures, while the high cost of training remains a major concern.

This paper investigates the acoustic frequency response of a Magnox package with internal pressure ranging from below atmospheric pressure to 3 bar. Vibrational resonant frequencies are measured using electromagnetic acoustic transducers (EMATs) and a finite element (FE) model is used to aid in interpreting the experimental results. This work builds on two previous studies that explored the vibrational resonant frequency response of Magnox packages at internal pressures above atmospheric pressure and addresses a gap in the literature by extending the analysis over a wider pressure range than these packages. The results demonstrate a strong correlation between the measured and modelled frequencies, validating the accuracy of the FE model in predicting the vibrational behaviour of the Magnox package under different pressure conditions. This work provides deeper insights into the acoustic characteristics of Magnox packages, which are essential for their ongoing safe and effective management.

Metal surfaces on industrial equipment are susceptible to fracture when subjected to extreme pressure and speed conditions. The safety of the parts is at risk and their service life is decreased by these fractures. One of the primary difficulties in developing eddy current non-destructive testing (NDT) sensors is locating fractures in any direction. The limitation is that fractures that are not perpendicular to the current flows produced cannot be detected by eddy current-based approaches. To locate arbitrary orientation fractures in conductive non-ferrous materials, this work explores the use of a rotating electromagnetic field technique and provides a novel rotating uniform eddy current (RUEC) system. This novel approach has greatly improved the detection of fractures in any orientation. The impedance signature of the fracture is investigated using a three-dimensional finite element (FE) simulation model of the new RUEC system. The rotating alternating current field measurement (RACFM) approach, which was designed to detect artificial cracks at any angle in underwater structures, is used in[1] to validate the model proposed in this study. The efficacy of the RUEC system and its potential in the detection of different fracture orientations in the aluminium specimen are demonstrated by the simulation results.

Defect classification of wind turbine blades is a popular topic in non-destructive health testing of complex components. This paper studies the wind turbine blade non-destructive testing (NDT) defect classification method based on a wavelet packet transform-convolutional neural network (WPT-CNN). Samples of wind turbine blades with different defect types were made and the datasets were collected using ultrasonic non-destructive inspection. The performance of two types of convolutional neural network (CNN) structure in predicting defects in wind turbine blades was compared. The results indicate that deep convolutional neural networks (DCNNs) are effective for defect detection in wind turbine blades, demonstrating greater feasibility and effectiveness in automatic classification based on deep learning models.

In the fault diagnosis of rolling bearings, compared to using vibration signals as the diagnostic target, utilising motor stator current signals can effectively reduce diagnostic costs and improve diagnostic efficiency. However, current signals often contain abundant noise that tends to obscure the characteristic signals related to bearing faults. Furthermore, the manifestation in signals of fault features for complex faults in rolling bearings is more intricate than that for single-point faults. Therefore, this paper proposes a signal processing method that can effectively suppress noise in current signals and further enhance the weak bearing composite fault features. The main content of this study leverages the inherent characteristics of motor stator current signals to eliminate a substantial amount of noise interference unrelated to fault characteristic signals, enabling the expression of weak fault characteristic signals to be further emphasised, thus achieving an effective representation of bearing composite fault features in current signals. Finally, the processed fault signals are used for fault classification through convolutional neural networks (CNNs). Experimental results effectively demonstrate that this method also has certain advantages and reliability in detecting and extracting weak compound fault features of bearings.