High-speed Detection Research Articles

LHB-YOLOv8: An Optimized YOLOv8 Network for Complex Background Drop Stone Detection

Real-time detection of rockfall on slopes is an essential part of a smart worksite. As a result, target detection techniques for rockfall detection have been rapidly developed. However, the complex geologic environment of slopes, special climatic conditions, and human factors pose significant challenges to this research. In this paper, we propose an enhanced high-speed slope rockfall detection method based on YOLOv8n. First, the LSKAttention mechanism is added to the backbone part to improve the model’s ability to balance the processing of global and local information, which enhances the model’s accuracy and generalization ability. Second, in order to ensuredetection accuracy for smaller targets, an enhanced detection head is added, and other detection heads of different sizes are combined to form a multi-scale feature fusion to improve the overall detection performance. Finally, a bidirectional feature pyramid network (BiFPN) is introduced in the neck to effectively reduce the parameters and computational complexity and improve the overall performance of rockfall detection. In addition we compare the LSKAttention mechanism with other attention mechanisms to verify the effectiveness of the improvements. Compared with the baseline model, our method improves the average accuracy mAP@0.5 by 4.8%. Moreover, the amount of parameters is reduced by 20.2%. Among the different evaluation criteria, the LHB-YOLOv8 method shows obvious advantages, making it suitable for engineering applications and the practical deployment of slope rockfall detection systems.

Breathers in Mode-Locked Lasers Based on Saturable Absorbers

Breathing pulses, as a unique nonlinear pulse phenomenon, play a key role in laser performance optimization, nonlinear optical processes, and complex signal transmission. Unlike stable solitons, the energy of breathing pulses fluctuates periodically over time, exhibiting periodic changes in both pulse frequency and amplitude. Through appropriate nonlinear effects, lasers can generate stable breathing pulses, achieving a mode-locked state that demonstrates a periodic "breathing" pattern. Based on this, a fiber laser incorporating a saturable absorber as the mode-locking element was designed and built, and stable breathing states were successfully observed at lower pump power levels. High-speed detection techniques and time-stretched dispersive Fourier transform (TS-DFT) technology were used to time-amplify and spectrally analyze the rapid pulses, while monitoring the evolution of the breathing pulse in both time and frequency domains. Experimental results indicate that changes in pump power significantly affect the periodic modulation induced by additional oscillations, thereby controlling the breathing ratio and ultimately leading to the formation of a stable soliton. When the pump power is between 470 <i>mW</i> and 480 <i>mW</i>, the formation of the breathing pulse was first observed, with a breathing ratio of up to 4.5. As the pump power increases, the breathing effect gradually diminishes, and at 510 <i>mW</i>, it completely disappears, with the breathing ratio dropping to 1.<br>These results confirm the critical role of pump power in controlling the breathing pulse state and its transition, demonstrating the potential for controlling pump power in ultrafast laser technology and nonlinear optics. The breathing pulse phenomenon, as a periodic pulse behavior, reflects the complex dynamical characteristics between nonlinear optical effects and cavity parameters. When combined with the natural synchronization system formed between the breathing frequency and the cavity frequency (determined by the cavity length), the periodic change of the breathing pulse becomes a crucial factor for controlling laser output. By adjusting parameters such as the laser’s nonlinearity and dissipation, the characteristics of the breathing pulse and breathing ratio can be precisely controlled, thus achieving precise control of the laser output. The periodic oscillatory characteristics of the breathing pulse inside the laser cavity lead to the non-uniform distribution of pulses, a feature that demonstrates enormous potential in pulse shaping, ultrashort pulse generation, and precise frequency comb control. Additionally, the presence of the breathing pulse may impact the stability and energy conversion efficiency of the laser, offering new perspectives for laser design and optimization.

Design and Development of a Precision Defect Detection System Based on a Line Scan Camera Using Deep Learning

The manufacturing industry environment is rapidly evolving into smart manufacturing. It prioritizes digital innovations such as AI and digital transformation (DX) to increase productivity and create value through automation and intelligence. Vision systems for defect detection and quality control are being implemented across industries, including electronics, semiconductors, printing, metal, food, and packaging. Small and medium-sized manufacturing companies are increasingly demanding smart factory solutions for quality control to create added value and enhance competitiveness. In this paper, we design and develop a high-speed defect detection system based on a line-scan camera using deep learning. The camera is positioned for side-view imaging, allowing for detailed inspection of the component mounting and soldering quality on PCBs. To detect defects on PCBs, the system gathers extensive images of both flawless and defective products to train a deep learning model. An AI engine generated through this deep learning process is then applied to conduct defect inspections. The developed high-speed defect detection system was evaluated to have an accuracy of 99.5% in the experiment. This will be highly beneficial for precision quality management in small- and medium-sized enterprises

Lane Detection Based on ECBAM_ASPP Model.

With the growing prominence of autonomous driving, the demand for accurate and efficient lane detection has increased significantly. Beyond ensuring accuracy, achieving high detection speed is crucial to maintaining real-time performance, stability, and safety. To address this challenge, this study proposes the ECBAM_ASPP model, which integrates the Efficient Convolutional Block Attention Module (ECBAM) with the Atrous Spatial Pyramid Pooling (ASPP) module. Building on traditional attention mechanisms, the ECBAM module employs dynamic convolution kernels to eliminate dimensionality reduction, enhancing the efficiency of feature channel learning and local interactions while preserving computational efficiency. The ECBAM_ASPP model incorporates the ECBAM attention mechanism into the feature extraction network, effectively directing the network to focus on salient features while suppressing irrelevant ones. Additionally, through variable sampling of the input, the model achieves multi-scale feature extraction, enabling it to capture richer lane-related feature information. Experimental results on the TuSimple and CULane datasets demonstrate that the ECBAM_ASPP model significantly improves real-time performance while maintaining high detection accuracy. Compared with baseline methods, the proposed model delivers superior overall performance, showcasing greater robustness and practicality.

High-temperature ultraviolet photodetector and amplifying integrated circuits based on AlGaN/GaN heterostructure

Abstract In this work, an ultraviolet (UV) photodetection and amplifying integrated circuit (IC) based on an AlGaN/gallium nitride (GaN) heterostructure is demonstrated. The IC consists of a metal-heterostructure-metal photodetector (MHM-PD) and a high-electron-mobility transistor (HEMT)-based amplifier. The photoresponse of the MHM-PD increases at elevated temperatures due to the spatial separation of the photocarriers under the polarization electric field at the AlGaN/GaN heterointerface, as well as the photo-enhanced leakage current through the metal-heterostructure junction. At 250 °C, MHM-PD achieves a peak photoresponsivity of 14.5 A W−1 and a UV-to-visible rejection ratio of 104. As the thermal chuck temperature increases from 25 °C to 250 °C, the performance of the HEMT-based amplifier shows good thermal stability. Finally, the IC achieves a photoresponse of over 106 V W−1 and a switching frequency of 50 kHz at 250 °C with rise and decay time constants of 3.95 μs and 2.8 μs, respectively. These results show that the IC has a high-sensitivity and high-speed UV detection capability.

A Multi-Weapon Detection Using Deep Learning

The escalating usage of light weapons in criminal and terrorist activities has necessitated the development of advanced weapon detection systems. This study centers on the application of the high-speed deep learning detection model, You Only Look Once version 7 (YOLOv7), to address the need for efficient and swift threat identification. The proposed system, trained on a self-curated dataset rich in images of various dangerous light weapons, is designed to recognize and distinguish multiple weapons simultaneously. The unique value of YOLOv7 in object and specifically weapon detection lies in its outstanding speed and accuracy, as demonstrated by certain weapon categories achieving a mean average precision (mAP) of 97%. The system’s performance can potentially be further enhanced by augmenting the image dataset for each weapon category. This study, therefore, not only validates the critical importance of YOLOv7 in advancing detection methodologies but also presents a practical solution for bolstering public safety.

Single-pixel orbital angular momentum detection in the temporal domain

Abstract Orbital angular momentum (OAM) beam’s spatial information captured using a camera, employs millions of single-pixel detectors arranged in a two-dimensional matrix. Poses challenges in OAM detection due to significant storage requirements, reduced speed due to low frame rate, and increased post-processing time. To address this bottleneck, a novel OAM detection technique utilizing only a diffuser and a single-pixel fast photodiode (FPD) is reported. The sixteen OAM beams, l = [ − 8 + 8 ] interact with the rotating diffuser resulting in a temporally varying speckle field. The 1D temporal speckle information (TSI) has been segregated from the temporally varying 2D speckle pattern images for each beam. The segregated 1D TSI has been classified by employing a machine learning model and achieved a classification accuracy of 97.1% and 97.3% on simulated and experimental data respectively. To further validate the proposed single-pixel OAM detection method, the 1D TSI is captured directly using a FPD. The experimentally captured 1D TSI captured via photodiode has been classified by a lightweight custom-designed 1D convolutional neural network and achieved an increased classification accuracy of 96%. This approach redefines OAM detection, operating in the temporal domain with a single-pixel FPD, thereby significantly reducing storage and computational costs while promising high-speed OAM detection.

Implementation of Deep Learning Algorithm for Vehicle Count Monitoring System

Vehicle detection plays a crucial role in various applications such as traffic surveillance, license plate recognition, and the development of autonomous vehicles. The You Only Look Once (YOLO) object detection method is renowned for its high-speed real-time object detection capabilities. In this study, YOLO is employed to detect vehicles in images and videos. YOLO treats object detection as a direct regression problem for bounding boxes and class predictions. The aim of this research is to develop a vehicle counting system using the YOLO method. The Midpoint algorithm is utilized to calculate the midpoint between two points in a coordinate plane. Another objective is to analyze the strengths and weaknesses of the method and algorithm in the context of vehicle detection while identifying related research trends. The test results indicate that the system is capable of detecting vehicles with an average accuracy of 92.42% across four different time periods. In the morning, the system detected 156 vehicles (manual count: 147, accuracy: 94.23%); at midday, it detected 246 vehicles (manual count: 225, accuracy: 91.46%); in the evening, 377 vehicles were detected (manual count: 351, accuracy: 93.10%); and at night, the system identified 526 vehicles (manual count: 225, accuracy: 92.58%). This study contributes to the development of a more effective vehicle counting system for smart city applications while also paving the way for further research on vehicle detection under varying lighting and environmental conditions.

Improved Algorithm for Vehicle Bottom Safety Detection Based on YOLOv8n: PSP-YOLO

The safety inspection system for the underside of vehicles demands both high detection speed and accuracy, necessitating a model with small parameters and high network efficiency. To address these issues and improve the real-time performance of under-vehicle safety inspection, the YOLOv8n object-detection algorithm was enhanced, resulting in the PSP-YOLO algorithm. Firstly, the improved PSP-YOLO network introduces a P2 small-object detection head, which is particularly effective in detecting tiny objects, such as small dangerous objects under vehicles, thereby significantly enhancing the ability to detect such small targets. Secondly, the Space-to-Depth Conv (SPD-Conv) module was introduced into the backbone network, which improves the detection performance of low-quality samples and small objects while ensuring high precision and accurate localization of small targets in images. Finally, the Partial Convolution (PConv) lightweight module was also added to the feature fusion network, effectively reducing the model’s parameters and computational load, conserving computing resources, and improving detection speed and accuracy. Experimental results demonstrate that the modified model increased mAP@0.5 by 3.9% and mAP@0.5-0.95 by 3.6% on the dataset used, reduced the number of parameters to 2.53 M, and achieved a detection speed of 104.6 f/s. Therefore, the improved YOLOv8n algorithm significantly enhances the detection performance of dangerous objects, meeting the high-precision and real-time requirements for vehicle underside safety inspections.

A Deep Learning Approach to Plastic Bottle Waste Detection on the Water Surface using YOLOv6 and YOLOv7

Deep learning is a branch of machine learning with many layers, such as the You Only Look Once (YOLO) method. From various versions of YOLO, YOLOv6 and YOLOv7 are considered more prominent because they achieve high Mean Average Precision (mAP) values. Both versions of YOLO have been implemented into various problems, especially in the waste detection problem. Plastic bottle waste is one of the most common types of waste that pollutes Indonesian waters. This study aims to solve this problem by helping to sort waste in surface waters by applying YOLOv6 and YOLOv7. FloW-Img was used, obtained on request from the Orcaboat website. The dataset consists of 500,000 bottle objects in 2,000 images. The YOLOv6 and YOLOv7 models were evaluated using mAP and running time. The results show that YOLOv6 and YOLOv7 can handle bottle waste detection well, with mAP values of 0.873 and 0.512, respectively. In addition, YOLOv6 (4.21 m/s) has a higher detection speed than YOLOv7 (13.7 m/s). However, in tests with images that do not have bottle objects, YOLOv7 provides better detection accuracy and consistency results, making it more suitable for real-world applications that demand high accuracy in environments with much visual noise.

Multichannel Signal Integration Approach for High-Speed Target Detection in Coherent MIMO Radar System

Multichannel Signal Integration Approach for High-Speed Target Detection in Coherent MIMO Radar System

Порівняльний аналіз ефективності методів розпізнавання автомобілів на аерокосмічних знімках

Introduction. Object recognition on aerial images is an urgent task in modern conditions, especially in cases requiring accurate and fast car recognition. Traditional contour extraction methods, such as Canny, Sobel, Laplacian, Prewitt, and Scharr, are based on gradient analysis and are known for their ease of implementation. This is an essential step for further recognition, as the correct definition of contours contributes to more accurate object identification. However, the effectiveness of the above methods could be improved, especially in complex environments with high object density, uneven brightness, and noise. Neural network models, such as YOLO (You Only Look Once), offer new possibilities, providing more accurate and reliable recognition, even in difficult situations. Purpose. This study compares the effectiveness of classical contour extraction methods and the YOLOv6n neural network model for vehicle recognition in aerial images. The accuracy of vehicle detection is evaluated by the main metrics: Precision, Recall, and F1-measure, which allow the determination of each method’s efficiency level in specific conditions. Methods. The study includes testing the classical Canny, Sobel, Laplacian, Prewitt, and Scharr algorithms for car outline detection and analyzing the results of the YOLOv6n model for deep-learning object detection. Classical methods use image processing to identify contours based on pixel gradients, which allows for extracting structures in an image. The YOLOv6n model is based on a neural network approach, considering complex image features for more accurate and faster object detection. Results. The data analysis showed that classical methods, although they can detect contours, have limited accuracy in conditions of high object density and sharp changes in brightness. The accuracy (Precision) and F1 Score for traditional methods was low, indicating a significant number of false positives and false negatives. In particular, the Sobel and Scharr methods showed the highest Recall but significantly decreased accuracy. In contrast, the YOLOv6n neural network model demonstrated high results in all primary metrics: Precision – 97.9%, Recall – 94.8%, F1 Score – 96.32%, and maP – 97.6%, which confirms its advantages in providing accurate and reliable vehicle recognition in aerial images. Conclusions. The study has shown that traditional contour extraction methods can serve as auxiliary tools for image preprocessing. Still, they need to provide adequate accuracy for the final stages of vehicle recognition. Neural network approaches, such as YOLOv6n, significantly outperform classical methods by providing high detection speed and accuracy, making them recommended for use in high-precision object recognition tasks in aerial images.

Spin photo detector by using a CoFeB magnetic tunnel junction

Abstract Ultra-fast photoelectric conversion devices are a crucial element in photonics applications and are the subject of intense research and development. In conventional high-speed photo detectors, photoelectric charge generation in semiconductors is rooted in a mechanism that directly outputs current. Owing to the dilemma of the amount of generated charge and high-speed output, it is being faced with fundamental difficult to achieve a higher response. Spin order, rather than charge generation, also exists as a phenomenon that enables ultrafast photoresponses, and magnetic tunnel junction (MTJ) are well-known high-speed electrical detection elements for magnetic states. In this study, we experimentally confirmed the magnetic switching of photoresponsive MTJ using a practical single CoFeB free layer with a high MR ratio of 80% by irradiation with laser light. Furthermore, we experimentally confirmed the operation of the reversible photo detector, and its rise time reached an ultrafast speed of 20 ps. We named it spin photo detector and it has huge potential applications that require ultrafast photo detection.

Dynamic Mechanisms of High-Pressure Hydrogen Self-Ignition and Flame Evolution in Obstructed Tubes: Experimental Insights and Implications

Experimental investigations on high-pressure hydrogen release into obstructed tubes with varying obstacle positions are conducted to explore the dynamic mechanisms of shock propagation, self-ignition and flame evolution. High-speed photography, pressure sensor and light detector are used in the experiments. The results reveal that obstacle position significantly affects self-ignition by the multi-dimensional interactions and the turbulence-promoted mixing. The reflected shock results in an overpressure rise of twice that behind the leading shock. In addition, the intensity of reflected shock decreases when it propagates upstream. The critical burst pressure for self-ignition first decreases and then increases with a further obstacle position away from the burst disk. The lowest burst pressure for self-ignition is 2.08 MPa arising in the tube with obstacles located at 200 mm axially, half that of the smooth tube. In obstructed tubes, three possible self-ignition regions exist, including inclined obstacle walls, tube centerline and tube sidewalls. The patterns of self-ignition are determined by burst pressure and obstacle position. Besides, flame acceleration is observed in the obstructed, which is related to turbulence promoted by the shock-obstacle interaction. The farther obstacles from the burst disk results in more marked effects on flame acceleration.

The Development of an Electron Pulse Dilation Photomultiplier Tube Diagnostic Instrument.

A new pulse-dilated photomultiplier tube (PD-PMT) with sub-20 ps temporal resolution and associated drivers have been developed for use detection and signal amplification in the inertial confinement fusion (ICF) community. The PD-PMT is coupled to a transmission line output in order to provide a continuous time history of the input signal. Electron pulse dilation provides high-speed detection capabilities by converting incoming signals into a free-electron cloud and manipulating the electron signal with electric and magnetic fields. This velocity dispersion is translated into temporal separation after the electrons transit into a drift space. The free electrons are then detected by using conventional time-resolved methods and the effective temporal resolution is improved about 12 times. In order to accurately obtain the actual device input signal, we experimentally investigated the relationship between microchannel plate (MCP) gain and electron energy during the first collision. We report the measurements with the PD-PMT, and the error source of the amplitude of the compressed signal is analyzed, which provides a reference for subsequent accurate construction.

Traffic Accident Prediction based on CNN and Time of Commuting after Accident

To address the challenges associated with traffic accident detection and the subsequent impact on commute times, a method combining CNN and Linear Regression models is proposed. Initially, this approach involves classifying specific accident data using images and detailed accident records. The images are then categorized into binary groups—accident and non-accident. Post-accident detection does not necessitate high-speed detection or extensive data processing; CNN are sufficiently capable for this purpose. Traffic accidents are an inevitable part of daily driving and can lead to significant delays in commuting. In rare cases, accidents may result in participants falling into a coma, further delaying rescue efforts. The proposed method aims to enhance road safety and traffic efficiency by promptly identifying accidents and minimizing the disruption caused to commute times through advanced technological solutions. By leveraging CNN for image analysis and Linear Regression for data modeling, this approach seeks to streamline accident detection processes, thereby improving overall traffic management and safety measures. This methodological framework underscores the importance of integrating cutting-edge technology to mitigate the adverse effects of traffic accidents on daily commuting.

Cathode Reaction Studies in Lithium-Oxygen Battery through Operando XRD Measurements

Introduction The lithium-Oxygen (Li-O2; LOB) battery has obtained significant attention as a promising candidate for next-generation storage batteries due to its higher specific energy density compared to conventional lithium-ion batteries [1]. However, the mechanism of the cathode reaction remains unclear, and the type of Li2O2 produced during cathode discharge is still uncertain, mainly because it's challenging to directly observe the compounds generated at the cathode during LOB operations. By newly designing observation cells and improving laboratory XRD system, we clarified the discharge products and investigated the discharging process using operando XRD measurements and Rietveld refinements[2]. In this study, the cathode reaction mechanism was discussed based on the results of the operando XRD measurements. Experimental A newly designed electrochemical cell, covered with a Kapton film dome to enable constant oxygen flow to the cathode side of the LOB, was set to the diffractomator of the Rigaku Ultima IV XRD measurement system. In the XRD system, Cu Kα was utilized as the incident X-ray, and high-speed one-dimensional detector served as the detector. In the electrochemical cell, Ketjen black self-standing films were employed as cathode materials, and tetraglyme solution containing 0.5 M LiTFSI + 0.5 M LiNO3 + 0.2 M LiBr was used as the electrolyte solution. The LOB was discharged and charged at a constant current density of 0.4 mA/cm2 for 10 hours each (4.0 mAh/cm2), in which XRD profiles were recorded every 20 minutes. Rigaku's SmartLab Studio II software was utilized for Rietveld analysis of each XRD profile to determine the detailed cathode product and its crystal structure. Results and discussion Upon conducting operando XRD measurements and Rietveld analyses, the cathode reactions of the Li-O2 battery were investigated. The discharging and charging reactions at the cathode of the Li-O2 battery, as well as the x values of the major product Li2-xO2 during discharging, have been investigated and determined. In the XRD profile measured during discharging, several peaks attributed to the cathode product were observed. Detailed Rietveld analysis of these peaks revealed that LiO2 with the same P63/mmc space group structure as Li2O2 was immediately formed after discharge initiation, followed by the deposition of Li2-xO2 accompanied by an increase in lithium occupancy. As discharging progressed, the lattice constant gradually extended towards the c-axis direction, leading to the formation of Li1.5O2 at a discharge capacity of 4.0 mAh/cm2. In this presentation, we will discuss the discharge and charge reactions of the Li-O2 battery cathode in detail, and we will also discuss the charge-discharge cycle characteristics.

Cathode Reaction Analysis of Lithium-Oxygen Battery Based on Operando XRD Measurements

Introduction The lithium-air battery (LAB), which has ten times the energy density of conventional lithium-ion batteries (LIB), is one of the promising candidates for the next-generation storage batteries, and there are great expectations for its development [1]. However, the reaction mechanism of the cathode reaction is still unclear, and the type of Li2O2 that is the product of the cathode in discharging is still unknown. The main reason for this is considered to be that it was difficult to directly observe the compounds generated at the cathode during LAB operation. By improving the laboratory XRD and the observation cell, we succeeded in operando XRD measurements of the cathode reaction in the LAB. In addition, under general conditions, the cathode product was Li1.5O2 with some Li missing from the crystal lattice of Li2O2 [2]. Here, we describe the details and discuss the reaction mechanism. Experimental An electrochemical cell covered with a Kapton® dome was constructed to allow constant oxygen flow to the cathode side of the LAB, and was attached to the diffractomator of a Rigaku Ultima IV XRD measurement system. A Cu Ka was used as the incident x-ray, a high-speed one-dimensional detector was used as the detector, a Ketjen black free-standing film was used as the cathode material, and tetraglyme solution with 0.5 M LiTFSI + 0.5 M LiNO3 + 0.2 M LiBr was used as the electrolyte solution. The LAB was discharged and charged at a constant current density of 0.4 mA/cm2 for 10 hours (4.0 mAh/cm2) each, and the XRD profile was recorded every 20 minutes. Each XRD profile was subjected to Rietveld analysis using Rigaku’s SmartLab Studio II software to determine the cathode product and its crystal structure. Results and discussion In the XRD profile measured during discharging, a lot of peaks due to the cell parts were observed. But, several peaks due to the cathode product were also observed. As a result of detailed Rietveld analysis of these peaks, the cathode product at the discharging capacity of 4.0 mAh/cm2 was not Li2O2 but Li1.5O2. Instead of the normal LiO2 with a rectangular crystal structure, LiO2 with a hexagonal crystal structure, which is the same as that of Li2O2, was observed immediately after the discharge. As the discharge progressed, Li was added to it, and the lattice constant gradually extended in the c-axis direction, forming Li1.5O2 at a discharge of 4.0 mAh/cm2. In the presentation, the charging process and the charge-discharge cycle characteristics would also be discussed.

Laser ultrasonic inspection of surface and subsurface defects in materials based on deep learning

ABSTRACT Laser ultrasonic testing is an advanced non-destructive testing with high frequency and fast detection speed. However, most laser ultrasonic testing characterises defects by manually extracting the features of signals. To address this, deep learning is introduced to identify defects intelligently. Firstly, theoretical models based on finite elements are built to study the the surface and subsurface defects’ effects on surface acoustic wave (SAW) propagation. Experimental studies are also conducted on 5 samples with surface defects and 15 with subsurface defects. Then, Gaussian white noise is added to both simulation and experimental signals to expand the data set and imitate the complex conditions. The wavelet transform images of the reflected and transmitted signal are combined into a data set for training the convolutional neural network (CNN). The subsurface defects’ data set is similarly augmented. Finally, a total of 40,000 surface defect and 21,000 subsurface defect data are collected and split into training, test, and validation sets (6:2:2 ratio). The CNN accurately classifies surface defect penetration depth (minimum error 5.87%), while subsurface defect widths are determined with transfer learning (minimum error 5.6%). It proves that it is feasible to classify and identify defects by combining the laser ultrasonic method with deep learning.

Eddy current quantitative evaluation of high-speed railway contact wire cracks based on neural network

PurposeThe purpose of this study is to study the quantitative evaluation method of contact wire cracks by analyzing the changing law of eddy current signal characteristics under different cracks of contact wire of high-speed railway so as to provide a new way of thinking and method for the detection of contact wire injuries of high-speed railway.Design/methodology/approachBased on the principle of eddy current detection and the specification parameters of high-speed railway contact wires in China, a finite element model for eddy current testing of contact wires was established to explore the variation patterns of crack signal characteristics in numerical simulation. A crack detection system based on eddy current detection was built, and eddy current detection voltage data was obtained for cracks of different depths and widths. By analyzing the variation law of eddy current signals, characteristic parameters were obtained and a quantitative evaluation model for crack width and depth was established based on the back propagation (BP) neural network.FindingsNumerical simulation and experimental detection of eddy current signal change rule is basically consistent, based on the law of the selected characteristics of the parameters in the BP neural network crack quantitative evaluation model also has a certain degree of effectiveness and reliability. BP neural network training results show that the classification accuracy for different widths and depths of the classification is 100 and 85.71%, respectively, and can be effectively realized on the high-speed railway contact line cracks of the quantitative evaluation classification.Originality/valueThis study establishes a new type of high-speed railway contact wire crack detection and identification method, which provides a new technical means for high-speed railway contact wire injury detection. The study of eddy current characteristic law and quantitative evaluation model for different cracks in contact line has important academic value and practical significance, and it has certain guiding significance for the detection technology of contact line in high-speed railway.