Trackside device bolt fastening operation intelligent management and control system and method with visual AI

Abstract

The application provides a trackside equipment bolt fastening operation intelligent management and control system and method fused with visual AI, and relates to the technical field of electric service intelligent operation and maintenance.The method comprises the following steps: acquiring first information and a first image;collecting the torque value of the current operation bolt based on a preset torque measurement sensor to obtain second information;matching the first information based on a preset all-trackside-equipment-bolt image recognition model, and recognizing and processing the first image based on the matching result to obtain third information;comparing the values in the second information with the values in the third information to obtain an operation result.The application provides an intelligent solution for the needs of automatic identification of bolt position, accurate matching of torque standard, and effective recording of operation data, etc.by fusing visual AI technology, edge computing, intelligent hardware, and a management platform, and realizes comprehensive management and control of trackside equipment bolt fastening operation.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for electrical systems, and more specifically, to an intelligent control system and method for bolt tightening operations of trackside equipment that integrates visual AI. Background Technology

[0002] Trackside electrical equipment is the core execution unit of the railway signaling system, mainly encompassing three key types of equipment: switch machines, signal boxes, and track circuit boxes. Its operational stability directly determines railway traffic safety. Fixing bolts, as core connecting components of this type of equipment, bear multiple functions, including structural fastening and signal transmission assurance, and are considered "small structures with big hidden dangers."

[0003] In terms of operating tools, mechanical and digital torque wrenches are still the mainstream equipment in China. Some companies have tried to develop smart wrenches, but their functions are limited. Although they can accurately detect torque values, they do not have the ability to identify bolt positions and still require manual assistance. Furthermore, equipment such as switch machines and signal boxes involve 8 to 20 types of bolts, with corresponding torque values ​​that can vary by more than 3 times. Manual setting is not only cumbersome but also prone to problems such as insufficient torque, overload, or underreporting of operations.

[0004] In terms of technological application, existing research mainly focuses on bolt defect detection, and has not yet achieved deep integration of visual recognition and torque control. It cannot automatically match tightening standards for different bolts and lacks real-time control over the tightening process. Furthermore, existing tools cannot record torque curves, making it difficult to assess bolt health status, and the lack of effective monitoring methods for the operation process poses potential safety hazards for long-term equipment operation. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent control system and method for bolt tightening operations on trackside equipment that integrates visual AI, in order to improve the aforementioned problems. To achieve this purpose, the technical solution adopted by this invention is as follows:

[0006] Firstly, this application provides an intelligent control method for trackside equipment bolt tightening operations that integrates visual AI, including:

[0007] Acquire first information and first image, wherein the first information is the equipment information to be operated and the first image is the tightening operation image of the bolt currently being operated;

[0008] The torque value of the currently working bolt is collected based on a preset torque measurement sensor to obtain the second information, which is the real-time torque value of the currently working bolt.

[0009] The first information is matched based on the preset image recognition model of all trackside equipment bolts, and the first image is recognized based on the matching result to obtain the third information, which is the torque limit of the currently working bolt.

[0010] The result of the task is obtained by comparing the values ​​in the second information with the values ​​in the third information.

[0011] Secondly, this application also provides an intelligent control system for trackside equipment bolt tightening operations that integrates visual AI, including:

[0012] An electronic force limiting wrench, wherein the electronic force limiting wrench integrates a high-precision torque measurement sensor and is equipped with a wired communication interface;

[0013] An edge vision processing module is embedded in the end of the electronic force limiting wrench. The edge vision processing module is equipped with a binocular camera or a wide-angle fisheye camera. The edge vision processing module is connected to the electronic force limiting wrench through the wired communication interface.

[0014] A handheld mobile terminal, wherein the handheld mobile terminal and the edge vision processing module communicate wirelessly;

[0015] An intelligent management platform, which can store the job results obtained by the handheld mobile terminal through a server.

[0016] The beneficial effects of this invention are as follows:

[0017] This invention proposes an intelligent control system and method for trackside equipment bolt tightening operations that integrates visual AI. It provides an intelligent solution to address the subjective errors inherent in manual operation, achieve automatic bolt position identification, accurate torque standard matching, and effective recording of operation data. By integrating visual AI technology, edge computing, intelligent hardware, and a management platform, it realizes comprehensive control over trackside equipment bolt tightening operations and constructs a closed-loop management system for trackside equipment bolt torque, encompassing "operation-monitoring-evaluation-early warning."

[0018] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the intelligent control method for trackside equipment bolt tightening operations that integrates visual AI, as described in this embodiment of the invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0022] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] Example 1:

[0024] like Figure 1 As shown, this embodiment provides an intelligent control method for trackside equipment bolt tightening operations that integrates visual AI, including steps S1 to S4:

[0025] S1: Obtain first information and first image, where the first information is the equipment information to be operated and the first image is the image of the current bolt tightening operation.

[0026] To clarify the specific methods for acquiring the first information and the first image, step S1 includes S11 to S16, specifically:

[0027] S11: Obtain the QR code information of the equipment to be operated.

[0028] Specifically, the equipment on site has a QR code label affixed to a designated location.

[0029] S12: Scan and identify the QR code information to obtain the ID of the device to be operated.

[0030] S13: Based on the ID of the equipment to be operated, match the information with the preset trackside equipment information to obtain the first information.

[0031] Specifically, the preset trackside equipment information includes equipment type, equipment number, equipment ID, track bed type, number of bolts, bolt number, bolt type, and corresponding torque limit for each bolt. That is, the first information includes the equipment type, equipment number, equipment ID, track bed type, number of bolts on the equipment, bolt number of each bolt on the equipment, bolt type of each bolt on the equipment, and corresponding torque limit for each bolt on the equipment.

[0032] S14: If the QR code information of the equipment to be operated cannot be obtained, the latitude and longitude coordinates of the equipment to be operated are obtained based on the positioning system, and the number plate of the equipment to be operated is recognized based on the OCR algorithm to obtain the number of the equipment to be operated.

[0033] Specifically, the equipment on site is equipped with numbered tags at designated locations.

[0034] S15: Based on the latitude and longitude coordinates of the equipment to be operated and the number of the equipment to be operated, match them with the preset trackside equipment information to obtain the first information.

[0035] S16: Based on the preset camera module, the working position of the current working bolt is captured to obtain the first image.

[0036] Specifically, the edge vision processing module is equipped with the camera module, and after the operation starts, the camera module is controlled to capture a first image.

[0037] S2: The torque value of the currently working bolt is collected based on the preset torque measurement sensor to obtain the second information, which is the real-time torque value of the currently working bolt.

[0038] Specifically, the electronic force limiter is equipped with the torque measurement sensor. When the electronic force limiter performs bolt tightening operations, the torque measurement sensor can collect the real-time torque value of the bolt being tightened.

[0039] S3: Match the first information based on the preset image recognition model of all trackside equipment bolts, and perform recognition processing on the first image based on the matching result to obtain the third information, which is the torque limit of the currently working bolt.

[0040] To clarify the specific method for obtaining the third information, step S3 includes S31 to S33, specifically as follows:

[0041] S31: Match the first information with all preset trackside equipment bolt image recognition models to determine the trackside equipment bolt image recognition model corresponding to the first information.

[0042] In this step, there are multiple preset image recognition models for all trackside equipment bolts. Each trackside equipment bolt image recognition model corresponds to one type of trackside equipment for recognition. The trackside equipment bolt image recognition model contains fastening operation image information of all bolts on the corresponding type of trackside equipment under different shooting angles, application location information, and torque limit value of each bolt.

[0043] Specifically, based on the first information and the preset image recognition models of all trackside equipment bolts, the image recognition model of the trackside equipment bolts corresponding to the equipment to be operated is determined.

[0044] S32: Based on the trackside equipment bolt image recognition model corresponding to the first information, the first image is processed to obtain the data information corresponding to the currently operating bolt.

[0045] Specifically, based on the image recognition model of the trackside equipment bolt corresponding to the first image and the first information, the data information corresponding to the current working bolt is obtained. The data information corresponding to the current working bolt includes the application position information and torque limit of the current working bolt.

[0046] S33: Determine the third information based on the data information corresponding to the currently working bolt.

[0047] S4: Based on the comparison between the values ​​in the second information and the values ​​in the third information, the operation result is obtained.

[0048] In this step, it is determined whether the value in the second information is greater than or equal to the value in the third information. If the value in the second information is less than the value in the third information, the second information is retrieved again until the value in the second information is greater than or equal to the value in the third information. Then, the operation is stopped and the operation result is output.

[0049] If the value in the second information is greater than or equal to the value in the third information, then stop the operation and output the operation result.

[0050] Specifically, the electronic torque limiter can automatically disengage and release force when the torque value reaches the predetermined torque limit, thus stopping the operation.

[0051] Example 2:

[0052] This embodiment further optimizes Embodiment 1. Specifically, it matches the first information based on a preset image recognition model of all trackside equipment bolts, and performs recognition processing on the first image based on the matching result to obtain the third information. Before the third information is the torque limit of the currently working bolt, it also includes steps S301 to S303:

[0053] S301: Based on a preset camera module, images of the working positions of all bolts on the trackside equipment are acquired from multiple preset shooting angles to obtain a second image. The second image is a fastening operation image of all bolts on the trackside equipment under different shooting angles.

[0054] Specifically, before starting the operation, a preset camera module is controlled to capture the second image, and the multiple preset shooting angles include different starting positions for the bolt tightening operation.

[0055] S302: Based on the second image, perform preprocessing, and based on the preprocessed second image, perform feature extraction and semantic segmentation to obtain fourth information, which is the application position information of all bolts;

[0056] Specifically, for feature extraction, this invention adopts a dual-stream fusion feature extraction method of "convolutional network to extract local features + attention mechanism to capture global structural relationships", which enables the model to simultaneously "see the details of bolts" and "understand the structure of components", thereby providing high-quality input for subsequent bolt detection and component region recognition.

[0057] In this step, the convolutional layer scans the image using a sliding window. Since the bolt shape has obvious and fixed geometric features, it can effectively extract local information such as the bolt's "hexagonal geometric edges", "surface roughness differences", and "shadow drop contours". However, due to the limited receptive field of the convolutional network, it cannot capture the "structural layout". In order to identify the component to which the bolt belongs, the model must "understand the result relationship of the entire region". Therefore, this invention introduces a lightweight attention mechanism to capture long-distance relationships between pixels and form a global correlation capability.

[0058] The formula for extracting image features using the fusion method is defined as follows:

[0059] (1);

[0060] In the above formula (1), To integrate image features, For the preprocessed second image The convolution operation performed As a weighting factor, The probability that each pixel belongs to each component category. For querying the matrix, Key matrix The transpose of the matrix, As a scale factor, It is a value matrix.

[0061] In this step, the weighting factor Used to control the fusion ratio of global semantics and local textures; query matrix Indicates what "needs to focus on" at the current feature point; key matrix Indicates what kind of structural information the feature points provide; scale factor Used to stabilize attention weights; value matrix Semantic representation of location.

[0062] For semantic segmentation, since different component regions correspond to different torque standards, it is necessary to determine which structural component region the bolt belongs to. Furthermore, the spatial relationship between the bolt and the component is continuous and regional, and this structural information must be obtained through "pixel-by-pixel classification." Therefore, this invention employs a lightweight semantic segmentation method that integrates image features. Output component label diagram .

[0063] The formula for the semantic segmentation output method is as follows:

[0064] (2);

[0065] In the above formula (2), This is the output component area diagram. For the final label of the pixel, The probability that each pixel belongs to each component category. For the segmentation head convolution parameters, This is to integrate image features.

[0066] In this step, semantic segmentation generates a pixel-level map of "region" and "component," providing regional basis for bolt attribution determination and structural input for automatic torque matching. Finally, the center point of the bolt detection results is... Projected onto component area diagram In this way, the region to which it belongs can be determined. , The first output of semantic segmentation Each component area This formula establishes a crucial bridge between "visual recognition -> torque retrieval" by defining the component category to which the bolt belongs. Through this process, the system can automatically determine which part or region the bolt belongs to and then query the torque standard in the database.

[0067] To clarify the specific preprocessing procedure for the second image, step S302 includes S3021 to S3023, specifically as follows:

[0068] S3021: Based on BIM and 3DMax, construct a controllable virtual world for the application environment of trackside equipment to obtain the complete application environment of all trackside equipment bolts.

[0069] In this step, to compensate for the limitations of perspective and the impact of extreme environmental conditions that are difficult to handle on-site, this invention introduces BIM and 3DMax to construct a controllable virtual world, enabling the generative acquisition of bolt images. The virtual scene construction includes: 3D models of tracks, turnouts, switch machines, locking devices, signal boxes, track circuit boxes, etc.; randomizable parameterized stain maps are set around the bolts to cover rust, oil, mud, fallen leaves, water, etc.; a multi-light source environment is constructed: afternoon low-angle light, backlight, diffused light, point light, LED fill light, etc.; and the rendering engine outputs depth of field, dynamic blur, noise model, CMOS characteristics, etc., similar to those of a real camera.

[0070] Specifically, based on the actual needs of bolt visual recognition, the training set, validation set, and test set are all constructed with real data and virtual data in equal proportions to improve the model's generalization ability. This data system ensures that the model has both robustness to real-world scenarios and completeness in controllable generation.

[0071] S3022: Based on the complete application environment of all trackside equipment bolts, the second image is processed using adaptive gamma correction and the Retinex algorithm to obtain a third image, which is the second image after adaptive illumination enhancement.

[0072] In this step, the present invention designs a preprocessing task scheme that can maintain lightweight and high efficiency under the condition of limited computing power, specifically including illumination enhancement, noise suppression, local super-resolution and occlusion repair, so as to improve the visual quality in complex trackside scenes.

[0073] Specifically, the environment beside the railway track often has problems such as strong backlight, shadows, and oil stains that cause the visual features of bolts to be obscured. Therefore, the image must be adaptively enhanced before being input into the recognition model.

[0074] The Retinex model is used to separate the "texture of the object itself" from the "effect of lighting." Adaptive gamma correction is integrated into the Retinex algorithm to achieve "dark area enhancement + highlight suppression," and adaptive gamma correction based on the brightness histogram enhances the texture of bolts in dark areas. The original image is decomposed into illumination estimation components and reflection enhancement components, separating the influence of external lighting from the image. This ensures that the true geometric texture of the bolts remains stable and recognizable under conditions of lighting changes, shadow occlusion, or strong reflections, thereby improving the robustness of the visual algorithm in outdoor trackside environments.

[0075] The second image is defined as follows:

[0076] (3);

[0077] In the above formula (3), For the second image, For the reflection enhancement component, Estimate the components of illumination.

[0078] The illumination estimation component is defined as:

[0079] (4);

[0080] In the above formula (4), To estimate the components of illumination, For Gaussian kernel function, For the second image, These are empirical parameters.

[0081] The reflection enhancement component is defined as:

[0082] (5);

[0083] In the above formula (5), To estimate the components of illumination, To prevent small constants from being divided by zero.

[0084] The dark area enhancement image is defined as:

[0085] (6);

[0086] In the above formula (6), This is the second image after preprocessing. For the second image, This is the adaptive adjustment coefficient for illumination. This represents the average brightness of the image.

[0087] Specifically, the illumination estimation component represents the intensity of external light sources illuminating the scene, including large-scale brightness variations such as shadows, backlighting, fill light, and high-contrast areas. These are low-frequency variation components, and a Gaussian kernel is used to estimate the large-scale, gradually changing illumination background. The standard deviation of the Gaussian kernel... A larger value indicates a slower change in illumination. The reflection enhancement component represents the intrinsic reflectivity of the object's surface, i.e., the reflection information such as the geometric texture of the component, including structured textures such as the hexagonal head edge of the bolt, metal texture, and component outline, which are high-frequency components. By separating the illumination estimation component, the reflection information such as the geometric texture of the component can be highlighted.

[0088] S3023: Based on the complete application environment of all trackside equipment bolts and PatchCNN, the third image is repaired, wherein the local damaged areas in the third image are reconstructed based on PatchCNN to obtain the preprocessed second image.

[0089] Specifically, by reconstructing locally damaged areas based on PatchCNN, the edges of bolts and metal textures that are obscured by dirt can be effectively restored, thereby solving the problem of local texture loss caused by oil and rust.

[0090] The repair model is defined as follows:

[0091] (7);

[0092] In the above formula (7), To repair the model, This is a function for the PatchCNN model. This refers to the context region of the texture-missing area.

[0093] Specifically, areas lacking texture, such as oil stains, rust, and mud spots, are defined as... Its context area is ,in, This is the second image.

[0094] S303: Based on the preprocessed second image and the fourth information, train the image recognition model of all trackside equipment bolts to obtain the preset image recognition model of all trackside equipment bolts.

[0095] Example 3:

[0096] This embodiment provides an intelligent control system for bolt tightening operations on trackside equipment that integrates visual AI. The system includes: an electronic torque limiter, an edge vision processing module, a handheld mobile terminal, and an intelligent management platform. The electronic torque limiter integrates a high-precision torque measurement sensor and is equipped with a wired communication interface. The edge vision processing module is embedded in the end of the electronic torque limiter and is equipped with a binocular camera or a wide-angle fisheye camera. The edge vision processing module is connected to the electronic torque limiter via the wired communication interface. The handheld mobile terminal communicates wirelessly with the edge vision processing module. The intelligent management platform can store the operation results acquired by the handheld mobile terminal on a server.

[0097] Specifically, when the edge vision processing module uses a binocular camera, the algorithm should support image stitching; when it uses a wide-angle fisheye camera, the algorithm should support wide-angle imaging distortion correction.

[0098] Example 4:

[0099] This embodiment is a further optimization based on embodiment 3. Specifically,

[0100] The torque range of the electronic force limiter is set to be greater than or equal to 15nm and less than or equal to 300nm.

[0101] The angle resolution of the electronic force limiter wrench is set to 0.01 degrees;

[0102] The electronic force limiter wrench is set to an accuracy of 1° / 90°.

[0103] Example 5:

[0104] This embodiment further optimizes upon embodiment 3. Specifically, the edge vision processing module has a pixel count greater than or equal to 1 million and uses a fixed focal length; the edge vision processing module is equipped with an ARM architecture; the edge vision processing module has a CPU and an NPU; the edge vision processing module supports running the Ubuntu Linux operating system; the edge vision processing module has a computing power greater than or equal to 6 TOPs; and the edge vision processing module has a dual-band Wi-Fi module that supports AP working mode.

[0105] Specifically, the hardware of the edge vision processing module needs to meet the visual recognition requirements of switch machine bolts and the wireless transmission requirements of handheld terminal data and electronic force limiter wrench.

[0106] Example 6:

[0107] This embodiment further optimizes Embodiment 3. Specifically, the outer shell of the edge vision processing module is made of aluminum alloy CNC machined; the top surface of the edge vision processing module has a heat dissipation grille; and the size of the edge vision processing module is set to be less than or equal to 100mm. 70mm 30mm.

[0108] Specifically, the outer casing of the edge vision processing module meets the requirements of being lightweight and having good heat dissipation, and also has a certain degree of waterproof and dustproof capability.

Claims

1. A method for intelligent control of bolt tightening operations on trackside equipment integrating visual AI, characterized in that, include: Acquire first information and first image, wherein the first information is the equipment information to be operated and the first image is the tightening operation image of the bolt currently being operated; The torque value of the currently working bolt is collected based on a preset torque measurement sensor to obtain the second information, which is the real-time torque value of the currently working bolt. The first information is matched based on the preset image recognition model of all trackside equipment bolts, and the first image is recognized based on the matching result to obtain the third information, which is the torque limit of the currently working bolt. The result of the task is obtained by comparing the values ​​in the second information with the values ​​in the third information.

2. The intelligent control method for trackside equipment bolt tightening operations integrating visual AI as described in claim 1, characterized in that, Acquire first information and a first image, wherein the first information is the equipment information to be operated on, and the first image is an image of the current bolt tightening operation, including: Obtain the QR code information of the equipment to be operated; The device ID to be operated is obtained by scanning and recognizing the QR code information. The first information is obtained by matching the ID of the equipment to be operated with the preset trackside equipment information; If the QR code information of the equipment to be operated cannot be obtained, the latitude and longitude coordinates of the equipment to be operated are obtained based on the positioning system, and the number plate of the equipment to be operated is recognized based on the OCR algorithm to obtain the number of the equipment to be operated. The first information is obtained by matching the latitude and longitude coordinates of the equipment to be operated with the equipment number together with the preset trackside equipment information; The first image is obtained by acquiring an image of the working position of the bolt based on a preset camera module.

3. The intelligent control method for trackside equipment bolt tightening operations integrating visual AI according to claim 1, characterized in that, The first information is matched based on a preset image recognition model for all trackside equipment bolts, and the first image is then processed based on the matching results to obtain third information, which is the torque limit of the currently operating bolt, including: The first information is matched with all preset trackside equipment bolt image recognition models to determine the trackside equipment bolt image recognition model corresponding to the first information. Based on the trackside equipment bolt image recognition model corresponding to the first information, the first image is recognized and processed to obtain the data information corresponding to the currently working bolt. The third information is determined based on the data information corresponding to the currently working bolt.

4. The intelligent control method for trackside equipment bolt tightening operations integrating visual AI according to claim 1, characterized in that, The first information is matched based on a preset image recognition model of all trackside equipment bolts, and the first image is then processed based on the matching results to obtain the third information. Before the torque limit of the currently operating bolt, the third information also includes: Based on a preset camera module, images of the working positions of all bolts on the trackside equipment are acquired from multiple preset shooting angles to obtain a second image. The second image is a fastening operation image of all bolts on the trackside equipment under different shooting angles. Based on the second image, preprocessing is performed, and feature extraction and semantic segmentation are performed on the preprocessed second image to obtain the fourth information, which is the application position information of all bolts; Based on the preprocessed second image and the fourth information, the image recognition model of all trackside equipment bolts is trained to obtain the preset image recognition model of all trackside equipment bolts.

5. The intelligent control method for trackside equipment bolt tightening operations integrating visual AI according to claim 4, characterized in that, Preprocessing is performed based on the second image, including: Based on BIM and 3DMax, a controllable virtual world is constructed for the application environment of trackside equipment to obtain the complete application environment of all trackside equipment bolts. Based on the complete application environment of all trackside equipment bolts, adaptive gamma correction and Retinex algorithm are used to process the second image to obtain a third image, which is the second image after adaptive illumination enhancement. Based on the complete application environment of all trackside equipment bolts, the third image is repaired using PatchCNN. Specifically, the local damaged areas in the third image are reconstructed using PatchCNN to obtain the preprocessed second image.

6. The intelligent control method for trackside equipment bolt tightening operations integrating visual AI according to claim 1, characterized in that, The task result is obtained by comparing the values ​​in the second information with the values ​​in the third information, including: Determine whether the value in the second information is greater than or equal to the value in the third information. If the value in the second information is less than the value in the third information, then reacquire the second information until the value in the second information is greater than or equal to the value in the third information, then stop the operation and output the operation result. If the value in the second information is greater than or equal to the value in the third information, then stop the operation and output the operation result.

7. A smart control system for bolt tightening operations on trackside equipment integrating visual AI, characterized in that, include: An electronic force limiting wrench, wherein the electronic force limiting wrench integrates a high-precision torque measurement sensor and is equipped with a wired communication interface; An edge vision processing module is embedded in the end of the electronic force limiting wrench. The edge vision processing module is equipped with a binocular camera or a wide-angle fisheye camera. The edge vision processing module is connected to the electronic force limiting wrench through the wired communication interface. A handheld mobile terminal, wherein the handheld mobile terminal and the edge vision processing module communicate wirelessly; An intelligent management platform, which can store the job results obtained by the handheld mobile terminal through a server.

8. The intelligent control system for trackside equipment bolt tightening operations integrating visual AI as described in claim 7, characterized in that, include: The torque range of the electronic force limiter is set to be greater than or equal to 15nm and less than or equal to 300nm. The angle resolution of the electronic force limiter wrench is set to 0.01 degrees; The electronic force limiter wrench is set to an accuracy of 1° / 90°.

9. The intelligent control system for trackside equipment bolt tightening operations integrating visual AI as described in claim 7, characterized in that, include: The edge vision processing module has a pixel count of 1 million or more and uses a fixed focal length. The edge vision processing module is equipped with an ARM architecture; The edge vision processing module has a CPU and an NPU; The edge vision processing module supports running the Ubuntu Linux operating system; The edge vision processing module has a computing power greater than or equal to 6 TOPs; The edge vision processing module is equipped with a dual-band Wi-Fi module and supports AP working mode.

10. The intelligent control system for trackside equipment bolt tightening operations integrating visual AI according to claim 7, characterized in that, include: The outer shell of the edge vision processing module is made of aluminum alloy CNC machined. The top surface of the edge vision processing module has a heat dissipation grille; The size of the edge vision processing module is set to be less than or equal to 100mm. 70mm 30mm.

Images