Track fastener installation state detector

Abstract

The invention provides a track fastener installation state detector. The track fastener installation state detector comprises a detection module and an auxiliary walking support arm, the detection module comprises a module quick coupler, a handle, a module walking wheel, a module limiting wheel, a supporting leg, a supporting foot, a core detection unit, an additional sensor, a power supply system and a lighting system; the auxiliary walking support arm assists the detection module in pushing detection on the steel rail. The detector provided by the invention adopts a miniaturized and portable design, so that the working efficiency is improved, and the skylight occupation time is reduced; the modular design, the single-module design and the double-module design are adopted, so that the on-site actual requirements are met; an optimized binocular structure is combined with a self-research algorithm, so that the recognition precision is effectively improved, and the false alarm rate is reduced; and through cooperative application of multiple sensors, the equipment applicability is improved, and the equipment expandability is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of railway infrastructure detection, in particular to a track fastener installation state detector. BACKGROUND

[0002] The existing steel rail fastener detection methods mainly include:

[0003] 1. Track fastener detection method based on track comprehensive inspection system Since the 1990s, foreign countries have successively carried out research on track comprehensive detection systems. The West Japan Railway Company developed the 141 series comprehensive detection vehicle, which can detect the state of rail joints, fastener bolt tightening and track bed. The TCIS system developed by the ENSCO company of the United States and the RailCheck track automatic detection system developed by the Atlas Elektorik company of Germany are both equipped with high-precision image acquisition modules and image processing units, which can automatically detect track fastener missing, but cannot effectively detect fastener misalignment, broken spring bars and other apparent diseases. In recent years, with the rapid development of China's railway transportation industry, China's railway automatic inspection system has made outstanding progress from "import" to "leadership". Before 2008, China mainly had GJ-3, GJ-4 and GJ-5 track automatic detection systems, which could realize automatic detection of railway geometric state, but could not realize detection of fastener missing. In 2008, China developed No. 0 high-speed comprehensive detection vehicle, which can detect at a speed of 250km / h, but the vehicle also does not have the ability to detect fastener missing. Subsequently, China successively developed CRH380B-002 high-speed comprehensive detection train, CRH380A-001 high-speed comprehensive detection vehicle and GJ-6 system.

[0004] 2. Track fastener anomaly detection method based on artificial design features The traditional track fastener anomaly detection method based on artificial design features mainly obtains track images through high-speed cameras installed at the bottom of the train; then, adopts the way of artificial design features combined with statistical pattern recognition classifier to realize automatic detection of track fastener apparent diseases. For example, Haar features combined with LDA, Adaboost classifier, HOG features combined with K- nearest neighbor classifier, SVM classifier, etc.

[0005] In 2007, the Polytechnic University of Bari in Italy developed an image-based real-time fastener detection system. The system, installed on the underside of trains, processed the acquired fastener images through two discrete wavelet transforms and input them into a multi-layer neural network classifier. The output was categorized into two classes based on whether fasteners were missing. The system achieved a detection accuracy of 95% and could operate in real-time at speeds up to 200 km / h. However, the system was only suitable for hexagonal fasteners. In 2009, Ruvo et al. in Italy proposed a fastener detection method based on image Haar features and discrete wavelet transform, capable of detecting the presence of hexagonal bolts. In 2010, the University of Illinois at Urbana-Champaign developed a video-based track detection system. This system employed a whole-to-part target recognition method, incorporating edge and texture information of the target to narrow the search range, ultimately enabling the detection of track fasteners, anchor bolts, and turnout components. In 2012, the University of Maryland used the grayscale histogram method to extract fastener images from background images. The HOG features of the fastener images were then directly input into an SVM classifier to classify fastener types. This method achieved a 98% detection rate for defective fasteners. Molefe et al. used a visual bag-of-words algorithm combined with SVM to classify and detect rail welding defects, and experiments demonstrated that this algorithm has good classification performance.

[0006] Concurrently, many domestic scholars have also proposed various methods for detecting abnormal track fasteners. In 2010, Beijing University of Aeronautics and Astronautics proposed a fastener localization algorithm based on Haar features and a fastener detection method based on the AdaBoost algorithm for ballasted track fasteners, which can effectively detect damaged fasteners on ballasted tracks, with fastener localization and detection accuracies of 97.8% and 92%, respectively. In 2011, Shanghai Jiao Tong University proposed a track fastener detection method based on image orientation field template matching, which can detect whether fasteners are missing at speeds up to 400 km / h, with a detection accuracy of 99.99%. In 2014, Beijing University of Aeronautics and Astronautics proposed a track fastener detection method based on the LDA topic model, which can detect damaged and missing fasteners under complex lighting conditions, is applicable to various types of fasteners, and has a comprehensive accuracy of 99%. In 2015, Southwest Jiaotong University proposed a track fastener detection algorithm that integrates PHOG and MSLBP features. The two features are fused through hierarchical weighting and input into an SVM classifier. This method achieves a defective fastener recognition rate of 98.1%. In the same year, Southwest Jiaotong University proposed a track fastener detection method based on image symmetry and optimized sparse representation, which can simultaneously identify fastener types and fastener defects. Wei et al. used a grayscale projection algorithm and geometric prior knowledge of track fasteners for fastener localization, and then used dense SIFT as image features to achieve fastener defect detection by using SVM.

[0007] 3. Deep Learning-Based Anomaly Detection Method for Track Fasteners With the development of artificial intelligence technologies, represented by deep learning, pattern recognition systems that automatically learn features from datasets using deep networks have shown advantages far exceeding those of traditional "manually designed features." Therefore, the automated detection of surface defects in rail fasteners is also increasingly using deep learning for feature extraction.

[0008] Gibert et al. developed a segmented network framework based on deep learning for fastener defect detection, which can detect different types of abnormal fasteners, while improving detection speed and accuracy. However, due to its insufficient stability, it cannot be directly applied to domestic fastener detection tasks. Zhao Xinxin used AlexNet convolutional neural network and Siamese twin network to extract deep features from fastener images in railway fastener detection, and used the conventional deep learning classifier softmax for fastener status identification. SONG et al. constructed datasets for various fastener types and used the TITAN graphics processor to improve the accuracy of CNN models for fastener identification. Long Yan et al. developed an intelligent track fastener detection system based on the Faster R-CNN algorithm, but the complex network architecture increased the computational load of the model, resulting in a need to improve detection speed. Chandran et al. detected missing fasteners based on image processing and deep learning networks. They designed a 7-layer convolutional neural network and used ResNet-50 for fastener classification, but the detection speed was slow. Wei et al. used Faster R-CNN to detect defects in track fasteners, but its accuracy was limited by the imbalanced dataset, and the two-stage network framework was still slow. Therefore, Bai Tang et al. improved the Faster R-CNN model by optimizing the region candidate network bounding box information in the model using labeled data to achieve fast and accurate localization of fasteners. Wei proposed a fastener defect detection method based on TLMDDNet by improving YOLOv3, which significantly improved the model's speed and accuracy, but the problem of data imbalance remained unresolved. Wang Bingshui et al. improved the YOLO algorithm based on Darknet-53, using multi-scale feature detection to solve the problem of small target feature loss, achieving a recall rate of 95.2%. J. Chen et al. proposed a cascaded three-stage defect fastener detection network based on YOLO, achieving a high detection rate. Y. Li et al. combined depthwise separable convolution and feature pyramids into the YOLO V3 network to improve the accuracy of defect detection.

[0009] 4. Track Fastener Detection Method Based on 3D Point Cloud The track fastener inspection method based on 3D point cloud mainly uses a structured light sensor to obtain the spatial 3D information of the track fastener, and extracts the point cloud data of the track fastener based on this information to realize the measurement of key geometric parameters and automatic identification of defects.

[0010] Domestic scholars began research on fastener detection based on structured light point clouds quite early. In 2011, Shanghai Jiao Tong University used a line laser source and a camera to construct a structured light sensor to acquire the 3D contours of track fasteners and proposed a fastener detection method combining track fastener contours and neural networks, capable of detecting missing fasteners. In 2014, the University of Alcalá in Spain used a line structured light sensor to acquire 3D point clouds of track fasteners and rails, and proposed a fastener detection method based on 3D ICP matching. First, the fastener point cloud is extracted based on prior knowledge, and then the fastener point cloud is matched with a preset fastener point cloud to finally detect damaged fasteners. In 2015, Middle East Technical University in Turkey established a real-time fastener detection system using a high-speed line structured light sensor. This system is installed on the bottom of the train and detects fasteners through the histogram similarity of the point cloud. The maximum real-time detection speed is 100 km / h, and the overall detection accuracy is 95%. In 2016, Nanchang University developed a fastener tension detection system based on a line structured light sensor. This system first established a measurement model for the fastener height difference based on manual inspection principles. Then, it performed noise filtering, valid data determination, and feature point extraction on the collected fastener point cloud to measure the required height difference with an accuracy of 0.1 mm. In 2018, Mao et al. used a line structured light sensor to construct a three-dimensional point cloud of the fastener surface. They then used a decision tree classifier to classify faults and defects and assessed the fastener tightness. This method achieved high detection accuracy but was computationally complex and susceptible to noise. In 2020, Han et al. used fastener images based on two-dimensional grayscale and three-dimensional depth information to analyze the position of the nut and bolt relative to the rail surface to calculate the fastener tightness, achieving a detection accuracy of 0.5 mm. However, this method was difficult to adapt to changes in rail surface height. In 2022, Chen Wenting et al. used the overall depth parameter of the rail spike and nut as the detection threshold for fastener tightness, achieving the detection of fastener looseness, but the detection accuracy was relatively low. In the same year, Li Mingsen et al. used an improved gray-scale weighting model to accurately extract the center line of the structured light and used K-means clustering to detect defects in the three-dimensional reconstructed fasteners. This method has a good effect on detecting fastener defects, but it is not suitable for dark and humid environments.

[0011] 5. Vibration Signal-Based Inspection Method for Rail Fasteners Inspired by vibration-based structural damage identification methods, some scholars have conducted research on vibration-based track fastener detection methods. In 2016, Valikhani of the Iranian University of Science and Technology proposed a vibration-based track fastener tightness detection method. This method first installs an accelerometer on the rail, strikes the rail with a hammer, and records the accelerometer signal. The vibration signal is analyzed using conditional entropy and optimized Morlet wavelet transform to detect the tightness of the track fastener. In 2017, Wei of Dalian University of Technology constructed a vibration-based track fastener system. The signal acquisition unit of this system mainly consists of four single-axis accelerometers installed on the rail. The vibration signal of the accelerometers is excited by striking the rail with a hammer, and a vibration signal processing method based on discrete wavelet transform is proposed. This method can calculate the location and degree of damage to the damaged fastener. In 2023, North China Electric Power University, in collaboration with China Academy of Railway Sciences Group Co., Ltd., proposed a dynamic response diagnostic method for high-speed railway fastener defects. This method utilizes acceleration sensors installed on high-speed integrated inspection trains to collect vehicle vibration response signals in normal and failed sections of fasteners. It also uses generalized demodulation time-frequency analysis to extract characteristic indicators for fastener condition diagnosis and constructs a classification model in conjunction with SSA-SVM to achieve automatic diagnosis of high-speed railway fastener conditions.

[0012] The main drawbacks of existing technologies: 1. Track Fastener Inspection Method Based on Integrated Track Inspection System The track inspection vehicle does not primarily target fasteners for inspection, and its inspection of abnormal fasteners is limited to a single type, making it unable to effectively detect the diverse surface defects of fasteners in reality.

[0013] 2. Anomaly Detection Method for Track Fasteners Based on Manually Designed Features Traditional track fastener anomaly detection methods based on manually designed features have weak generalization ability when processing non-homogeneous data, resulting in a high false alarm rate and a high rate of missed detection of defective fasteners, which causes problems in engineering practice.

[0014] 3. Deep Learning-Based Anomaly Detection Method for Track Fasteners The deep learning-based method for detecting anomalies in track fasteners lacks depth information and cannot automatically detect the fastening status of the fastener system.

[0015] 4. Track Fastener Detection Method Based on 3D Point Cloud The track fastener anomaly detection method based on 3D point cloud does not acquire key point cloud data of track fasteners in a target-driven manner. It introduces a large amount of redundant point cloud data, which greatly increases the computational complexity of segmenting elastic clips, track spikes, anchor bolts, etc. It is costly, structurally complex, and lacks portable detection equipment, making it inconvenient for inspection personnel to carry and lift them onto the track.

[0016] 5. Vibration Signal-Based Inspection Method for Rail Fasteners The track fastener inspection method based on vibration signals requires the pre-installation of sensors such as accelerometers on the rails or high-speed integrated inspection trains, which is inconvenient to deploy. It may even require the use of hammers or other means to excite rail vibration, resulting in low inspection efficiency and limited inspection types, which cannot meet the daily static maintenance needs of railway sites. Summary of the Invention

[0017] To address the aforementioned problems and overcome the shortcomings of existing technologies, this invention provides a miniaturized, portable fastener installation status detector. It adopts a modular design, with a single detection module capable of inspecting fasteners on both the inner and outer sides of a single rail; it supports simultaneous use of two modules, including fastener tightening status, spring clip installation status, component damage, and fastener model inspection. It features a visual display function and report output capabilities.

[0018] The specific technical solution is as follows: A track fastener installation status detector includes: a detection module and an auxiliary traveling arm; the detection module includes a module quick-connect connector, a handle, module traveling wheels, module limit wheels, support legs, support feet, a core detection unit, additional sensors, a power system, and a lighting system; the auxiliary traveling arm assists the detection module in pushing and detecting on the rail; The core detection unit is designed based on the binocular detection principle, and the binocular camera layout is based on the size of the fastener. The additional sensors include a displacement sensor, a mileage sensor, and a radar sensor. The displacement sensor performs two main functions: threshold alarm and sleeper identification. It identifies sleeper features, fastener features, and bolt features based on changes in measured values. The threshold alarm function serves as a feature point for software identification, triggering an algorithm to comprehensively analyze the fastener installation status. The mileage sensor works in conjunction with the displacement sensor to identify and locate sleepers and fasteners. It can also locate problematic fastener mileage markers based on actual conditions, providing reliable reference for track maintenance personnel. The radar sensor monitors the working environment, providing early warnings of obstacles in the work area and identifying other workers, ensuring the safety of equipment and personnel.

[0019] Furthermore, the comprehensive analysis of the fastener installation status includes: qualitative analysis of the spring clip installation status, component defects, and fastener model inspection, and quantitative analysis of the fastener tightening status. The installation status of the spring bar refers to whether the spring bar is installed crookedly or backwards. The aforementioned component defects refer to: missing spring clips, gauge baffles, insulating blocks, spiral rail spikes, T-bolts, and anchor bolts; and broken spring clips. The fastener fastening status is defined as the difference between the gap between the lower jaw of the front end of the spring clip and the insulating block, the insulating gauge block, the gauge baffle, and the rail, or the difference between the distance between the top surface of the anchor bolt and the upper surface of the iron pad and the standard value. The track fastener installation status detector according to claim 2 is characterized in that the thickness of the pad is: The total thickness of the fastening system pad is the sum of the thicknesses of all pads between the rail base and the top surface of the rail support platform; for the WI-7 and WJ-8 type fastening systems, the thickness of the pad on the iron pad is the sum of the thicknesses of the pads between the rail base and the top surface of the iron pad, and the thickness of the pad under the iron pad is the sum of the thicknesses of the pads between the bottom surface of the iron pad and the top surface of the rail support platform.

[0020] Furthermore, the binocular camera includes a side-mounted camera and a vertically mounted camera, with the side-mounted camera being tilted.

[0021] Furthermore, the power supply system: The device has a built-in 24V, 20AH ternary lithium battery, and is equipped with a charging port and an external power input port. The single working time can reach more than 5 hours. The device has a multi-stage power conversion module inside, which can support sensor power supply in the range of DC 3.3-30V.

[0022] According to claim 2, the track fastener installation status detector is characterized in that the lighting system can take pictures with the core detection unit camera and can also provide good lighting at night, which facilitates manual re-inspection of the fastener status.

[0023] Furthermore, the main body of the auxiliary traveling arm adopts a three-fold design, which can be folded and stored, with a folded length of less than 60cm; it consists of a quick-connect connector for the arm, a support rod, a buckle, a top rail device, a traveling wheel for the arm, a limit wheel for the arm, a wrench, and a push rod; the push rod adopts a telescopic design.

[0024] According to claim 2, the track fastener installation status detector is characterized in that the track fastener installation status detector adopts a single module structure or a dual module structure, the dual module structure mainly consists of two sets of detection modules and auxiliary traveling arms, so as to realize one-time inspection of the upper and lower rails and the inner and outer sides.

[0025] Furthermore, the software control scheme of the detector: After the system boots up, it performs program initialization, creating mileage detection thread, displacement scanning thread, obstacle scanning thread, and detection algorithm thread; The mileage detection thread is responsible for mileage detection-related processing, recording the mileage data of the equipment running on the line; the mileage data can be used as an assessment standard for the workload of fastener installation status detection; it also provides the location data of abnormal fasteners; it works with the switch quantity of displacement detection to determine the sleeper position; and it is used as a trigger condition for taking pictures. The displacement scanning thread is responsible for collecting displacement data, cooperating with mileage detection to determine the sleeper position, and assisting in determining the fastener type based on analog data; when the switch value is 0, it means that the sleeper has not been reached, and no photo processing is performed at this time, but mileage recording can be performed in stages; when the switch value is 1, it means that the sleeper may be reached, and the mileage data can be compared with the mileage table to confirm that the sleeper has been reached. At this time, the mileage data is recorded in real time, and a photo is taken every time the mileage data changes by 1cm. The detection algorithm thread is responsible for recognizing and calculating image data, uploading the calculation results, and generating reports in conjunction with mileage data; The obstacle scanning thread is responsible for acquiring data from the ultrasonic sensors and is specifically used to determine whether there are obstacles hindering the fastener installation status detection operation.

[0026] A detection method for a track fastener installation status detector: After the equipment is turned on, it will operate along the rail at a speed of up to 7 km / h. When the displacement sensor output is 1 and the mileage sensor data matches the sleeper mileage meter data, the equipment will record the mileage data and collect the fastener status every 1 cm change based on the mileage data.

[0027] Technical effects of the present invention: Compared with the integrated track inspection system, the present invention features a portable design, high skylight utilization rate, and low requirement for skylight duration.

[0028] Compared with manually designed track fastener anomaly detection, the algorithm developed in this invention has a high recognition rate, high detection accuracy, and low false alarm rate.

[0029] Compared with deep learning-based track fastener anomaly detection, this invention incorporates depth information, enabling automatic detection of fastener tightness.

[0030] Compared with the detection of track fasteners using 3D point clouds, this invention does not require a large amount of redundant point cloud data, and provides intuitive detection of components such as spring bars and track spikes, resulting in low cost and high efficiency.

[0031] Compared with the detection of track fasteners based on vibration signals, this invention does not require other excitation conditions, is not affected by other factors on the rail, has high detection efficiency, and can detect various types of track fasteners. Attached Figure Description

[0032] Figure 1 This is a diagram of an existing foreign integrated track inspection system.

[0033] Figure 2 This is a diagram of a high-speed comprehensive inspection train in my country based on existing technology.

[0034] Figure 3 This is a schematic diagram of the structure of the present invention.

[0035] Figure 4 This is a schematic diagram of the detection module structure of the present invention.

[0036] Figure 5 This is a schematic diagram of the core detection unit structure of the present invention.

[0037] Figure 6 This is a schematic diagram of the additional sensor structure of the present invention.

[0038] Figure 7 This is a structural diagram of the power supply system of the present invention.

[0039] Figure 8 This is a structural diagram of the auxiliary traveling arm of the present invention.

[0040] Figure 9 This is a schematic diagram of the push rod structure of the present invention.

[0041] Figure 10 This is a schematic diagram of the dual-module structure used in this invention.

[0042] Figure 11 This is a system flowchart of the present invention.

[0043] Reference numerals: 1. Detection module; 2. Auxiliary traveling arm; 11. Module quick-connect connector; 12. Handle; 13. Module traveling wheel; 14. Module limit wheel; 15. Support leg; 16. Support foot; 17. Core detection unit; 18. Additional sensor; 19. Power system; 110. Lighting system; 171. Side-mounted camera; 172. Vertical-mounted camera; 181. Displacement sensor; 182. Mileage sensor; 183. Radar sensor; 21. Support arm quick-connect connector; 22. Support rod; 23. Buckle; 24. Top rail device; 25. Support arm traveling wheel; 26. Support arm limit wheel; 27. Wrench; 28. Push rod. Detailed Implementation

[0044] This invention relates to a track fastener installation status detector and detection method. To enable those skilled in the art to better understand the invention, the technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments.

[0045] The specific implementation of the present invention will be described in detail below with reference to specific embodiments. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this patent, and should not be construed as limiting this patent.

[0046] As shown in the figure, this invention mainly consists of a detection module 1 and an auxiliary walking arm 2. The detection module 1 has functions such as fastener installation status detection, mileage detection, obstacle scanning, and lighting. The auxiliary walking arm 2 has a connecting device function and an auxiliary walking function.

[0047] Detection Module 1 The detection module 1 consists of a module quick-connect connector 11, a handle 12, module running wheels 13, module limit wheels 14, support legs 15, support feet 16, a core detection unit 17, additional sensors 18, a power supply system 19, and a lighting system 110. The structure is as follows: Figure 4 As shown.

[0048] 1. Core Detection Unit 17 Designed based on the principle of binocular detection, the layout of the traditional binocular camera is optimized according to the size of the fastener, making it highly targeted. Combined with a self-developed algorithm, it effectively improves the accuracy of fastener status detection. The binocular camera includes a side-mounted camera 171 and a vertically mounted camera 172, with the side-mounted camera 171 set at an angle.

[0049] 2. Additional sensor 18: Displacement sensor 181 It achieves two main functions: threshold alarm and sleeper identification. The displacement sensor 181 can identify sleeper features, fastener features, bolt features, and other information based on changes in measured values. The threshold alarm function serves as a feature point for software identification, triggering an algorithm to comprehensively analyze the fastener installation status, including qualitative checks such as elastic clip installation status, component damage, and fastener model; and quantitative checks such as fastener tightness.

[0050] Mileage sensor 182 In conjunction with displacement sensor 181, it can realize sleeper identification and positioning, fastener positioning, and can locate the mileage marker of problematic fasteners according to the actual situation, providing a reliable reference for track maintenance personnel.

[0051] Radar sensor 183 By enabling monitoring of the work environment, early warnings can be given about obstacles in the work area, and other workers can be identified, thus ensuring the safety of equipment and personnel.

[0052] 3. Power System 19 The device has a built-in 24V, 20AH ternary lithium battery and is equipped with a charging port and an external power input port, allowing for a single operation time of over 5 hours. The device also features a multi-stage power conversion module, supporting sensor power supply within the DC 3.3-30V range.

[0053] 4. Lighting system 110 Considering that high-speed line maintenance work is concentrated at night, a lighting system 110 was specially designed. It can be used in conjunction with the camera of the core detection unit 17 to take pictures, and can also provide good lighting at night, which facilitates manual re-inspection of the fastener status.

[0054] Assisted traveling arm 2 When used in combination with the core detection unit 17, the auxiliary detection unit pushes the inspection unit on the rail for inspection.

[0055] The main body adopts a three-fold design, which can be folded for storage, with a folded length of less than 60cm. It consists of a quick-connect joint 21 for the support arm, a support rod 22, a buckle 23, a top rail device 24, a support arm traveling wheel 25, a support arm limiting wheel 26, a wrench 27, and a push rod 28.

[0056] The push rod 28 adopts a telescopic design with a maximum length of 1.5m and a retracted length of less than 65cm.

[0057] Dual-module structure description It mainly consists of two sets of detection modules 1 and auxiliary traveling arms 2. It enables one-time inspection of the upper and lower rails, as well as the inner and outer sides.

[0058] Software control scheme description After the system boots up, it initializes the program and creates threads for mileage detection, displacement scanning, obstacle scanning, and detection algorithms.

[0059] The mileage detection thread is responsible for mileage detection-related processing and records the mileage data of the equipment running on the line. The mileage data can be used as an assessment standard for the workload of fastener installation status detection; it also provides location data for abnormal fasteners; it works with the switch quantity of displacement detection to determine the sleeper position; and it is used as a trigger condition for taking pictures.

[0060] The displacement scanning thread is responsible for acquiring displacement data, coordinating with mileage detection to determine the sleeper position, and using analog data to assist in determining the fastener type. A switch value of 0 indicates the sleeper has not been reached; in this case, no photo is taken, but intermittent mileage recording is performed. A switch value of 1 indicates the sleeper may have been reached; by comparing the mileage data with the odometer, the arrival of the sleeper can be confirmed, and mileage data is recorded in real time. A photo is taken every 1cm change in mileage data.

[0061] The detection algorithm thread is responsible for recognizing and calculating image data, uploading the calculation results, and generating reports in conjunction with mileage data.

[0062] The obstacle scanning thread is responsible for acquiring data from the ultrasonic sensors and is specifically used to determine whether there are obstacles hindering the fastener installation status detection operation.

[0063] System workflow description: After the equipment is turned on, it will operate along the rail at a speed of up to 7 km / h. When the output of the displacement sensor 181 is 1 and the data of the mileage sensor 182 matches the data of the sleeper mileage meter, the equipment will record the mileage data and collect the fastener status every 1 cm change based on the mileage data.

[0064] Core components and functions of this invention: Detection Module: This is the core of the instrument, integrating: Core detection unit: Designed based on the "binocular detection principle" (i.e., using two cameras), it includes a side-mounted camera with an inclined orientation and a vertical camera to acquire the depth and image information of the fastener.

[0065] Additional sensors: Displacement sensor: Used to identify features such as sleepers, fasteners, and bolts, and triggers the algorithm through a threshold alarm function.

[0066] Mileage sensor: In conjunction with displacement sensor, it enables precise positioning of sleepers and fasteners, and records the mileage location of problematic fasteners.

[0067] Radar sensors: Monitor the working environment, warn of obstacles, or identify other workers to ensure safety.

[0068] This instrument can perform comprehensive qualitative and quantitative analysis of the installation status of fasteners.

[0069] Qualitative testing: Spring clip installation status: Check if the spring clip is installed backwards or crooked.

[0070] Missing components: Check for missing spring clips, bolts, insulating blocks, etc., or whether the spring clips are broken.

[0071] Fastener model check.

[0072] Quantitative detection: Fastener tightening status: This is achieved by measuring the gap between the front end of the spring clip and components such as the rail, or the distance between the top surface of the anchor bolt and the iron pad, and comparing it with the standard value.

[0073] Mechanism of action: 1. Resolved the contradiction between "2D inspection" and "3D inspection". Existing deep learning-based detection methods (mainly 2D images) can identify defects, but they lack depth information and cannot detect the fastener tightness (i.e., whether it is loose). While 3D point cloud-based methods can acquire depth, the equipment is expensive, structurally complex, and inconvenient, and generates a large amount of redundant point cloud data, making computation complex.

[0074] This invention does not use expensive and complex 3D laser point clouds, but instead employs an optimized "binocular structure" (side-mounted + vertical-mounted camera). This is a clever compromise: it can acquire enough depth information to achieve "quantitative" detection of "fastener fastening status," while avoiding the complexity and high cost of 3D point clouds, and its compact structure enables a portable design.

[0075] 2. It achieves a combination of "portability" and "functionality". Existing high-precision inspection equipment, such as the "railway integrated inspection vehicle," is bulky and mainly used for high-speed inspections, with limited coverage for a single type of fastener inspection. In contrast, the "vibration signal-based" method requires pre-installed sensors or hammering, which is inefficient and inconvenient.

[0076] This invention features a miniaturized and portable design, allowing for manual operation. Specifically designed for "maintenance windows" (railway maintenance windows), it offers short operation time and high efficiency. Furthermore, its modular design (can be used individually or in pairs), foldable and retractable support arms, and long-lasting battery significantly enhance the flexibility and practicality of on-site operations—features unmatched by larger equipment.

[0077] 3. Intelligent "sensor fusion" workflow Traditional "manually designed features" methods have poor generalization ability and high false alarm and false negative rates.

[0078] This invention integrates multiple sensors to design a highly efficient automated workflow. Its innovation lies in using a displacement sensor as a "physical trigger": 1) When the equipment is pushed forward, the displacement sensor first detects the "sleeper" feature.

[0079] 2) Once the arrival at the sleeper is confirmed (switch value is 1), immediately coordinate with the mileage sensor to record the precise location.

[0080] 3) Simultaneously trigger the core detection unit (binocular camera) to start high-frequency photography (e.g., take a picture once every 1cm change).

[0081] This automated "perception-location-photographing" process, combined with self-developed algorithms, ensures the accuracy and efficiency of data collection, effectively improves recognition accuracy, and reduces the false alarm rate.

[0082] The core innovation of this invention lies in the fact that it is not a mere accumulation of single technologies, but rather an organic combination of binocular vision, multi-sensor fusion, portable modular structure, and automated software process, tailored to the actual needs of daily railway maintenance. While ensuring detection accuracy (especially by adding quantitative detection of "tightness" that traditional 2D methods cannot achieve), it also achieves miniaturization, portability, and high efficiency.

[0083] Compared to large-scale integrated track inspection vehicles, this invention is compact, portable, requires less track maintenance time, and can detect a variety of abnormal fastener types. Compared to traditional detection methods based on manually designed features, this invention has a low false alarm rate, a low missed detection rate, and can automatically learn abnormal track fastener characteristics. Compared to fastener detection based on deep learning, this invention can automatically identify the fastener system's tightening status. Compared to fastener detection based on 3D point clouds, this invention effectively reduces the computational load of point cloud data, simplifies model segmentation calculations, and significantly reduces instrument size, making it easier for inspection personnel to carry on track. Compared to fastener detection based on vibration signals, this invention does not require other operations to excite rail vibration, resulting in high detection efficiency and a wide range of detection types, meeting the needs of daily static maintenance operations.

[0084] Terminology Explanation: Fastening system tightness: The gap between the lower jaw of the middle front end of the spring strip and the insulating block, insulating gauge block, gauge baffle, or rail, or the difference between the distance between the top surface of the anchor bolt and the upper surface of the iron pad and the standard value. Spring bar installation status: Check if the spring clip is installed crookedly (the angle between the axis connecting the two sides of the spring clip and the rail direction is greater than 15°) or if it is installed backwards.

[0085] Missing parts: The spring clip, gauge baffle, insulating block (insulated gauge block), spiral rail spike, T-bolt, and anchor bolt are missing, and the spring clip is broken.

[0086] Pad thickness: The total thickness of the fastening system pad is the sum of the thicknesses of all pads between the rail base and the top surface of the rail support platform. For WI-7 and WJ-8 type fastening systems, the thickness of the pad on the iron pad is the sum of the thicknesses of the pads between the rail base and the top surface of the iron pad, and the thickness of the pad under the iron pad is the sum of the thicknesses of the pads between the bottom surface of the iron pad and the top surface of the rail support platform.

[0087] The present invention has been described above by way of example. Obviously, the specific implementation of the present invention is not limited to the above-described manner. Any improvements made by adopting the inventive concept and technical solution of the present invention, or direct application to other occasions without modification, are all within the protection scope of the present invention.

Claims

1. A track fastener installation status detection instrument, characterized in that, include: The detection module (1) and the auxiliary traveling arm (2) are included; the detection module (1) includes a module quick-connect connector (11), a handle (12), a module traveling wheel (13), a module limit wheel (14), a support leg (15), a support foot (16), a core detection unit (17), an additional sensor (18), a power system (19), and a lighting system (110); the auxiliary traveling arm (2) assists the detection module (1) in pushing and detecting on the rail. The core detection unit (17) is designed based on the binocular detection principle, and the binocular camera layout is based on the size of the fastener. The additional sensors (18) include: a displacement sensor (181), a mileage sensor (182), and a radar sensor (183). The displacement sensor (181) realizes two major functions: threshold alarm and sleeper identification. It identifies sleeper features, fastener features, bolt features, etc., based on changes in the measured values. The threshold alarm function is used as a feature point for software identification, triggering an algorithm to comprehensively analyze the fastener installation status. The mileage sensor (182) works in conjunction with the displacement sensor (181) to realize sleeper identification and positioning, fastener positioning, and can locate the mileage marker of problematic fasteners according to the actual situation, providing a reliable reference for track maintenance personnel. The radar sensor (183) realizes the monitoring of the working environment, can provide early warning of obstacles in the working area, identify other workers, and ensure the safety of equipment and workers.

2. The track fastener installation status detector according to claim 1, characterized in that, The comprehensive analysis of the fastener installation status includes: qualitative analysis of the spring clip installation status, component damage, and fastener model inspection; and quantitative analysis of the fastener tightening status. The installation status of the spring bar refers to whether the spring bar is installed crookedly or backwards. The aforementioned component defects refer to: missing spring clips, gauge baffles, insulating blocks, spiral rail spikes, T-bolts, and anchor bolts; and broken spring clips. The fastener fastening status is defined as the difference between the gap between the lower jaw of the front end of the spring clip and the insulating block, the insulating gauge block, the gauge baffle, and the rail, or the difference between the distance between the top surface of the anchor bolt and the upper surface of the iron pad and the standard value.

3. The track fastener installation status detector according to claim 2, characterized in that, Pad thickness: The total thickness of the fastening system pad is the sum of the thicknesses of all pads between the rail base and the top surface of the rail support platform; for the WI-7 and WJ-8 type fastening systems, the thickness of the pad on the iron pad is the sum of the thicknesses of the pads between the rail base and the top surface of the iron pad, and the thickness of the pad under the iron pad is the sum of the thicknesses of the pads between the bottom surface of the iron pad and the top surface of the rail support platform.

4. The track fastener installation status detector according to claim 2, characterized in that, The binocular camera includes a side-mounted camera (171) and a vertically mounted camera (172), wherein the side-mounted camera (171) is tilted.

5. The track fastener installation status detector according to claim 2, characterized in that, The power supply system (19): The device has a built-in 24V, 20AH ternary lithium battery, and is equipped with a charging port and an external power input port. The single working time can reach more than 5 hours. The device has a multi-stage power conversion module inside, which can support sensor power supply in the range of DC 3.3-30V.

6. The track fastener installation status detector according to claim 2, characterized in that, The lighting system (110) can be used in conjunction with the camera of the core detection unit (17) to take pictures, and can also provide good lighting at night, making it convenient to manually re-inspect the status of fasteners.

7. The track fastener installation status detector according to claim 2, characterized in that, The main body of the auxiliary traveling arm (2) adopts a three-fold design, which can be folded and stored, with a folded length of less than 60cm; it consists of a quick-connect connector (21), a support rod (22), a buckle (23), a top rail device (24), a traveling wheel (25), a limiting wheel (26), a wrench (27), and a push rod (28); the push rod (28) adopts a telescopic design.

8. The track fastener installation status detector according to claim 2, characterized in that, The track fastener installation status detector adopts a single-module structure or a dual-module structure. The dual-module structure mainly consists of two sets of detection modules (1) and an auxiliary traveling arm (2) to realize one-time inspection of the upper and lower rails and the inner and outer sides.

9. The track fastener installation status detector according to claim 2, characterized in that, The software control scheme of the detector: After the system boots up, it performs program initialization, creating mileage detection thread, displacement scanning thread, obstacle scanning thread, and detection algorithm thread; The mileage detection thread is responsible for mileage detection-related processing, recording the mileage data of the equipment running on the line; the mileage data can be used as an assessment standard for the workload of fastener installation status detection; it also provides the location data of abnormal fasteners; it works with the switch quantity of displacement detection to determine the sleeper position; and it is used as a trigger condition for taking pictures. The displacement scanning thread is responsible for collecting displacement data, cooperating with mileage detection to determine the sleeper position, and assisting in determining the fastener type based on analog data; when the switch value is 0, it means that the sleeper has not been reached, and no photo processing is performed at this time, but mileage recording can be performed in stages; when the switch value is 1, it means that the sleeper may be reached, and the mileage data can be compared with the mileage table to confirm that the sleeper has been reached. At this time, the mileage data is recorded in real time, and a photo is taken every time the mileage data changes by 1cm. The detection algorithm thread is responsible for recognizing and calculating image data, uploading the calculation results, and generating reports in conjunction with mileage data; The obstacle scanning thread is responsible for acquiring data from the ultrasonic sensors and is specifically used to determine whether there are obstacles hindering the fastener installation status detection operation.

10. A detection method for a track fastener installation status detector as described in any one of claims 1-9, characterized in that: After the equipment is turned on, it will be put into operation and will advance along the rail at a speed of less than 7 km / h. When the output of the displacement sensor (181) is 1 and the data of the mileage sensor (182) matches the data of the sleeper mileage table, the equipment will record the mileage data and collect the fastener status every 1 cm change based on the mileage data.

Images