A method and device for inspecting a precast beam yard

Abstract

The present application relates to the technical field of intelligent construction, in particular to a kind of inspection method and device for precast beam field, method obtains the task containing the position of component to be detected, detection item and design threshold;Equipment moves to detection area, real-time obstacle avoidance and accurate positioning;Adjust the posture of detection component and aim at target;Pretreat data after acquisition and calculate detection index;Compare index and determine quality;Generate report upload platform and push terminal;Device includes self-propelled omnidirectional mobile chassis, multi-degree-of-freedom robotic arm system, visual detection module, sensor system, control and communication system, power supply and charging system.Self-propelled omnidirectional mobile chassis carries each module and realizes movement, multi-degree-of-freedom robotic arm system adjusts the posture of detection component, visual detection module collects image and three-dimensional point cloud data, sensor system collects mechanical and geometric parameters, control and communication system coordinates the work of each module and processes data, power supply and charging system provides power, realizes automated accurate inspection.

Description

Technical Field

[0001] This invention belongs to the field of intelligent construction technology, specifically relating to an inspection method and device for precast beam yards. Background Technology

[0002] Precast beam yards, as core precast sites for large-scale infrastructure projects such as bridges and high-speed railways, primarily undertake the production of precast concrete components such as T-beams and box girders. Their construction process encompasses key stages including rebar tying, embedded pipe installation, concrete pouring, and component curing. Among these, the quality of rebar tying directly affects the structural performance of the components, the accuracy of embedded pipe placement influences the subsequent prestressing effect, and the elastic modulus and dimensions of the concrete determine the durability and assembly compatibility of the components. Quality control in these stages is a core prerequisite for ensuring the safety, durability, and service life of the engineering structure. However, existing quality inspection methods in precast beam yards still suffer from numerous technical limitations, making it difficult to meet the demands for efficient and precise quality control. (1) Manual inspection is inefficient and has significant errors. The measurement of the spacing between steel bars and the judgment of the firmness of the binding points mostly rely on manual operation with a measuring tape and calipers. On average, a single person can only complete the full-item inspection of a few precast beams per day, which is difficult to adapt to the pace of mass production. Moreover, the inspection results are easily affected by the operator's experience, on-site lighting and dust environment. The judgment of the firmness of the binding points lacks standardized basis. (2) The detection range is limited. Traditional detection tools such as laser scanners need to be fixed and can only cover open areas such as the front and top of the component. They cannot reach hidden parts such as the bottom of the beam and corners. Static sensors can only achieve single-point data acquisition, making it difficult to perform full-surface detection on large components. In particular, in the detection of the location of pre-embedded pipes, it is difficult to accurately obtain three-dimensional coordinates, and it is easy to miss hidden dangers such as pipe offset and breakage. (3) Insufficient data integration capability. The data for testing items such as steel bars, pipes, and concrete come from different devices such as manual record sheets, laser scanners, and sensor acquisition instruments. The formats are not uniform and the sources are scattered. It is necessary to manually summarize them into tables for analysis. It takes a long time to generate a complete quality report on average. It is impossible to realize the real-time association between the test data and the component location and defect information, which greatly increases the complexity of construction management and delays the efficiency of quality problem rectification.

[0003] In view of this, the present invention is hereby proposed. Summary of the Invention

[0004] To address the aforementioned technical problems in the existing technology, this invention provides an inspection device and method for precast beam yards, solving the problems of low efficiency, limited scope, and insufficient data integration in existing detection methods.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: Firstly, a quality inspection method for precast beam yards includes: S1. Obtain inspection task information, which includes the location parameters of the precast component to be inspected, at least one inspection item, and the design standard threshold of the corresponding inspection item. S2. Control the inspection equipment to move to the preset inspection area of ​​the precast component to be inspected, and identify and avoid obstacles in the construction environment in real time during the movement. S3. According to the detection items, adjust the spatial posture of the detection execution component so that the detection execution component is aligned with the target detection part of the prefabricated component to be detected, and maintains a preset safe distance. S4. The detection data of the target detection area is collected by the detection execution component. The detection data includes image data, three-dimensional point cloud data, mechanical parameter data and geometric parameter data. S5. Preprocess and extract features from the collected detection data, and call the preset algorithm to calculate the detection index value; S6. Compare the test index value with the corresponding design standard threshold to determine whether the prefabricated component to be tested meets the quality requirements. S7. Generate a test report containing test index values, judgment results, and anomaly markers, and transmit it to the data management platform.

[0006] Furthermore, the acquisition of detection task information specifically includes: The system receives structured task instructions from the cloud via a wireless communication module. These instructions include the unique identifier code of the prefabricated component to be inspected, BIM model coordinate data, and design standard thresholds for each inspection item.

[0007] Furthermore, the process of controlling the inspection equipment to move to the preset inspection area of ​​the prefabricated component to be inspected specifically includes: A three-dimensional map of the precast beam yard environment is constructed based on autonomous navigation system technology, and the optimal movement path is planned according to the location parameters of the precast components to be detected. The system uses LiDAR to scan the surrounding environment in real time. When an obstacle is detected, the straight-line distance between the obstacle and the inspection equipment is calculated. ; like The inspection equipment is controlled to move along the planned path at a speed of 0.8-1.5 m / s; like Control the inspection equipment to reduce speed to ; like The inspection equipment is stopped and a warning signal is issued. It is restarted after the obstacle is removed. The inspection equipment is precisely positioned and aligned with the prefabricated components to be inspected by using QR code recognition technology.

[0008] Furthermore, the process of adjusting the spatial orientation of the detection execution component specifically includes: Based on the BIM model data of the prefabricated component to be inspected, determine the three-dimensional coordinates of the target inspection location; The RRT* motion planning algorithm is invoked to generate an obstacle avoidance trajectory from the initial posture to the target posture, and the obstacle avoidance trajectory satisfies the kinematic constraints of the detection execution component; The multi-degree-of-freedom actuator is driven to move along the obstacle avoidance trajectory, so that the detection center of the detection actuator is aligned and the distance between the detection actuator and the target detection part is maintained within the range of 0.3-1.0m; Once the attitude adjustment is complete, a ready signal is output to trigger the data acquisition process.

[0009] Furthermore, the process of generating the test report specifically includes: It automatically summarizes the raw data, calculation process, and test index values ​​of each test item to form a structured data table; On the 3D model of the prefabricated component to be inspected, abnormal areas are marked according to the world coordinate system, and the abnormal areas are highlighted in red. The system automatically matches the rectification suggestion library based on the anomaly type and generates a text description that includes rectification measures, standard basis, and acceptance methods. The data tables, 3D models, and rectification suggestions are integrated into a PDF report, which is then uploaded to the data management platform via a 5G communication module and simultaneously pushed to designated terminal devices.

[0010] Furthermore, the testing items include: rebar binding quality testing, embedded pipe location testing, component dimensional testing, and concrete elastic modulus testing.

[0011] Furthermore, during the inspection of the rebar binding quality, the specific processes of steps S4 to S5 include: Color images of the rebar binding area are acquired using a high-definition camera; the color images are then weighted and converted to grayscale, transforming the RGB images into grayscale images; a 3×3 window midpoint filtering algorithm is used to remove image noise according to the following formula:

[0012] in, After median filtering, the coordinates The grayscale value of the pixel at that location; Within the filtering window, coordinates The original grayscale value of the pixel; This is a median calculation function; The straight line features of the reinforcing bars are extracted by Hough transform, and the peak points are detected in the parameter space according to the straight line equation to extract the straight line profile of the reinforcing bars.

[0013] Furthermore, the specific process of calling the preset algorithm to calculate the detection index value includes: For the extracted straight lines of adjacent parallel reinforcing bars, calculate the actual spacing according to the formula, which is as follows:

[0014] Where, is the vertical distance between two adjacent parallel steel bars; This is the common coefficient for parallel straight reinforcing bars; is the constant term for two parallel reinforcing bars; is the coefficient vector. The modulus length; The YOLOv3 deep learning model is used to identify ligation points, and the detection accuracy is optimized through a loss function. The specific formula is as follows:

[0015] in, The value of the loss function; Weights for coordinate loss; Weights for target confidence loss; Weights for the no-target confidence loss; The coordinates of the center of the prediction box; Width and height; Confidence in the detection frame; Be confident in the true frame; For classification probability; This represents the true classification probability. The number of missing binding points within each preset length is counted, and the deviation rate of rebar spacing and the pass rate of binding points are calculated.

[0016] Furthermore, when the test item is the concrete elastic modulus test, the specific processes of steps S4 to S5 include: The applied force is recorded in real time using pressure sensors and displacement gauges. With deformation Calculate the nominal stress and strain The specific formula is as follows:

[0017]

[0018] in, The cross-sectional area of ​​the specimen; Gauge length; The least squares method was used to fit the stress and strain data in the linear stage, and the elastic modulus was... The following formula is used to derive:

[0019] in, Number of data sets; The stress value is the value of the i-th data set. The strain value is the value of the i-th data set. for The sum of the products of stress and strain; It is the sum of stress values; It is the sum of strain values; The sum of the squares of the strain values; Simultaneously calculate the coefficient of determination. Verify linearity:

[0020] in, For the first The actual stress values ​​of the set of data; To fit the stress value; This represents the average value of the actual stress. like or If the value exceeds the design range, the test is deemed invalid and will be automatically retested.

[0021] Furthermore, when the detection item is the detection of the location of the pre-buried pipeline, the specific processes of steps S4 to S6 include: A 3D point cloud dataset of the pipeline was acquired and reconstructed using a structured light depth camera. The point cloud density is no less than 100 points / cm²; the RANSAC algorithm is used to robustly estimate the pipe axis, specifically including: A predetermined number of points are randomly selected to construct a hypothetical straight line model. The Euclidean distance from all points to this straight line is calculated using the following formula:

[0022] in, The coordinates of the reference point; For the first The spatial coordinates of the points; It is the direction vector; Statistical satisfaction The number of interior points, when the proportion of interior points The iteration stops when the time is right, and the total number of iterations is determined by the following formula:

[0023] in, Minimum number of iterations; The preset expected confidence level, The ratio of interior points; The fitted pipe axis is spatially aligned with the theoretical axis in the BIM design model, and the maximum Euclidean distance between the axes is calculated. ,like Generate a position offset alarm; The RANSAC algorithm is used to robustly extract the pipeline axis from the noisy point cloud, and the position is compared with the BIM model.

[0024] Furthermore, when the inspection item is the inspection of the component's external dimensions, the specific processes of steps S4 to S6 include: The entire surface contour data of the component is acquired using a lidar scanner at a scanning speed of 1000 points / second, obtaining at least 5×10⁻⁶ points. 5 One outline point; Contour extraction is performed using the Canny edge detection algorithm: the image is smoothed using a 5×5 Gaussian filter with a standard deviation σ=1.5, and the gradient is calculated using the Sobel operator. The specific formula is as follows:

[0025] Nonmaximum suppression and double thresholding ( , Connect the edges; perform principal component analysis on the contour point set, solve for the covariance matrix, take the eigenvector corresponding to the largest eigenvalue as the major axis direction, and calculate the minimum bounding rectangle size. The specific formula is as follows:

[0026] in, The number of contour points, Let be the mean vector of the point set. Indicates matrix transpose; For the first Two-dimensional coordinates of a point; Calculate the actual length of the component With design value The deviation rate is calculated using the following formula: If the deviation rate is ≤3%, the external dimensions are deemed acceptable; if the deviation rate is >3%, the out-of-tolerance area is marked in the 3D point cloud.

[0027] Secondly, an inspection device for a precast beam yard includes: Self-propelled omnidirectional mobile chassis: used to carry multi-degree-of-freedom robotic arm systems, vision inspection modules, sensor systems, control and communication systems, and power supply and charging systems, and to move in the precast beam yard environment; Multi-degree-of-freedom robotic arm system: mounted on the self-propelled omnidirectional mobile chassis, with a detachable actuator at its end; used to move the detection components of the vision inspection module and the sensor system to the area to be inspected of the component; Visual inspection module: Located on the detachable actuator, it includes a high-definition RGB camera and a structured light depth camera, used to acquire component images and 3D point cloud data; Sensor system: connected to the control and communication system, including pressure sensors, ultrasonic sensors, lidar and displacement gauges, used to collect mechanical and geometric parameters; Control and communication system: Establishes signal connections with the self-propelled omnidirectional mobile chassis, multi-degree-of-freedom robotic arm system, vision inspection module and sensor system respectively, to coordinate the collaborative work of each module, and to receive data collected by the vision inspection module and sensor system, process it and output the inspection results; Power supply and charging system: used to provide power to the self-propelled omnidirectional mobile chassis, multi-degree-of-freedom robotic arm system, vision inspection module, sensor system and control and communication system.

[0028] The present invention has the following advantages: (1) Improve detection efficiency: Automatic inspection replaces manual measurement, significantly reducing detection time and adapting to the needs of large-scale construction; (2) Ensure detection accuracy: Combine high-precision sensors with AI algorithms to provide highly reliable detection results and reduce human error; (3) Optimize quality management: Realize the real-time collection, integration and analysis of test data, generate standardized reports, and improve the level of construction quality management. Attached Figure Description

[0029] Figure 1 A flowchart of an inspection method for a precast beam yard provided in an embodiment of the present invention; Figure 2 This is a structural diagram of an inspection device for a precast beam yard provided in an embodiment of the present invention.

[0030] Explanation of reference numerals in the attached figures: 1. Visual inspection module; 2. Multi-degree-of-freedom robotic arm system; 3. Self-propelled omnidirectional mobile chassis; 4. Control and communication system; 5. Sensor system. Detailed Implementation

[0031] The technical solution of the present invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are not all embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0032] It should be noted that, unless otherwise specifically stated, the relative arrangement and numerical expressions of the components and steps described in these embodiments should not be construed as limiting the scope of the invention.

[0033] The following description of exemplary embodiments is merely illustrative and is not intended to limit the invention or its application or use in any way. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail herein, but where applicable, such techniques, methods, and apparatus should be considered part of this specification.

[0034] Example 1 See Figure 1 , Figure 1 This is a flowchart of an inspection method for precast beam yards proposed in this invention. The specific steps include: S1. Obtain inspection task information, which includes the location parameters of the precast component to be inspected, at least one inspection item, and the design standard threshold of the corresponding inspection item. The control and communication system receives structured task instructions from the cloud via its 5G / Wi-Fi module. These instructions include basic information about the component to be inspected, a list of inspection items, design standard thresholds, and auxiliary data. The basic information of the component to be inspected includes a unique identifier code and spatial coordinates from the precast beam yard BIM model. The list of inspection items can select one or more inspection items (reinforcement binding quality, embedded pipe location, component dimensions, concrete elastic modulus). The design standard thresholds are the qualification criteria for each item. The auxiliary data includes the component BIM model (IFC format) and the parameters required for inspection. After parsing the instructions, the control and communication system stores the parameters in a local database and generates a visual task list, providing a data benchmark for subsequent steps.

[0035] Obtaining inspection task information specifically includes: receiving structured task instructions sent from the cloud via a wireless communication module. The structured task instructions include the unique identifier code of the prefabricated component to be inspected, BIM model coordinate data, and design standard thresholds for each inspection item.

[0036] S2. Control the inspection equipment to move to the preset inspection area of ​​the precast component to be inspected, and identify and avoid obstacles in the construction environment in real time during the movement; the process of controlling the inspection equipment to move to the preset inspection area of ​​the precast component to be inspected specifically includes: A 3D map of the precast beam yard environment is constructed based on autonomous navigation system technology, and the optimal movement path is planned according to the position parameters of the precast components to be inspected. The surrounding environment is scanned in real time by LiDAR, and when an obstacle is identified, the straight-line distance between the obstacle and the inspection equipment is calculated. ;like Control the inspection equipment to move along the planned path at a speed of 0.8-1.5 m / s; if Control the inspection equipment to reduce speed to ;like The system controls the inspection equipment to stop and issue a warning signal, and restarts movement after the obstacle is removed; it uses QR code recognition technology to achieve precise positioning and alignment between the inspection equipment and the prefabricated components to be inspected.

[0037] Specifically, a 16-line LiDAR sensor scans the precast beam yard to construct a 3D environmental map (accuracy ≤ 50mm), identifying obstacles such as cranes and tool racks. Then, based on the coordinates of the component to be inspected, the optimal path from "charging station to inspection point" is planned to avoid obstacles. When there are no obstacles during the movement, the machine moves at a speed of 0.8-1.5m / s. When an obstacle is detected at a distance of 0.3m < d ≤ 0.5m, the speed is reduced to 0.2-0.5m / s. When the distance d ≤ 0.3m, the machine stops and issues a warning. Finally, the chassis and component are aligned (positioning error ≤ 30mm) by matching the QR code with the component positioning QR code using QR code recognition technology.

[0038] S3. According to the detection items, adjust the spatial orientation of the detection execution component so that the detection execution component is aligned with the target detection area of ​​the prefabricated component to be detected, while maintaining a preset safe distance; specifically including: The process of adjusting the spatial orientation of the detection execution component specifically includes: Based on the BIM model data of the prefabricated component to be inspected, the three-dimensional coordinates of the target inspection location are determined; the RRT* motion planning algorithm is called to generate an obstacle avoidance trajectory from the initial posture to the target posture, the obstacle avoidance trajectory satisfying the kinematic constraints of the inspection execution component; the multi-degree-of-freedom actuator is driven to move along the obstacle avoidance trajectory, so that the inspection center of the inspection execution component is aligned, and the distance between the inspection execution component and the target inspection location is maintained within the range of 0.3-1.0m; after the posture adjustment is completed, a ready signal is output to trigger the data acquisition process.

[0039] The testing items include: rebar tying quality inspection, embedded pipe location inspection, component shape and size inspection, and concrete elastic modulus inspection; S4. The detection data of the target detection area is collected by the detection execution component. The detection data includes image data, three-dimensional point cloud data, mechanical parameter data and geometric parameter data. Multiple types of data are collected, including image data acquired by a high-definition RGB camera to capture color images of the component surface, suitable for rebar tying and surface defect detection; 3D point cloud data acquired by a structured light depth camera, suitable for pipe location and dimension detection; mechanical parameter data acquired by pressure sensors and displacement gauges to capture the applied force F and deformation ΔL, suitable for concrete elastic modulus detection; and geometric parameter data acquired by LiDAR to capture component contour points (at a speed of 1000 points / s), suitable for dimension detection. All data are transmitted in real time to the control and communication system via Ethernet or RS485 bus.

[0040] S5. Preprocess and extract features from the collected detection data, and call the preset algorithm to calculate the detection index value; In the preprocessing stage, weighted grayscale and 3×3 median filtering are performed on the image data (the median of the pixel grayscale values ​​within a 3×3 window is taken as the filtered pixel value) to remove outliers from the point cloud data; in the feature extraction stage, straight line features are extracted using Hough transform, contour features are extracted using the Canny algorithm, and the axis is fitted using the RANSAC algorithm; in the index calculation stage, the least squares method and YOLOv3 model are used to calculate the detection index values ​​such as rebar spacing, pipe offset, dimensional deviation, and elastic modulus.

[0041] S6. Compare the test index value with the corresponding design standard threshold to determine whether the prefabricated component to be tested meets the quality requirements. The test index values ​​are compared with the corresponding design standard thresholds to determine whether the prefabricated components to be tested meet the quality requirements; the test index values ​​are compared with the design standard thresholds to determine whether the components are qualified; if there are unqualified items, the coordinates of the abnormal parts are recorded.

[0042] S7. Generate a test report containing detection index values, judgment results, and anomaly markers, and transmit it to the data management platform. The test report generation process specifically includes: The process of generating the inspection report specifically includes: automatically summarizing the original data, calculation process, and inspection index values ​​of each inspection item to form a structured data table; marking abnormal areas on the 3D model of the prefabricated component to be inspected according to the world coordinate system, with abnormal areas highlighted in red; automatically matching the rectification suggestion library according to the anomaly type to generate a text description containing rectification measures, standard basis, and acceptance methods; integrating the data table, 3D model, and rectification suggestions into a PDF report, uploading it to the data management platform via a 5G communication module, and simultaneously pushing it to designated terminal devices.

[0043] The report includes structured data tables (inspection items, design values, measured values, deviations), 3D model anomaly markers, and rectification suggestions. It is then transmitted to the cloud data management platform and construction personnel's terminals via 5G or Wi-Fi. The report generation time is ≤20 seconds.

[0044] Specifically, depending on the specific testing item, the steps include: (1) When the inspection item is the quality inspection of rebar tying, the specific process of steps S4 to S6 includes: Color images of the rebar binding area are acquired using a high-definition camera; the color images are then weighted and converted to grayscale, transforming the RGB images into grayscale images; a 3×3 window midpoint filtering algorithm is used to remove image noise according to the following formula:

[0045] in, After median filtering, the coordinates The grayscale value of the pixel at that location; Within the filtering window, coordinates The original grayscale value of the pixel; This is a median calculation function; The straight line features of the reinforcing bars are extracted by Hough transform, and the peak points are detected in the parameter space according to the straight line equation to extract the straight line profile of the reinforcing bars.

[0046] The specific process of calling the preset algorithm to calculate the detection index value includes: For the extracted straight lines of adjacent parallel reinforcing bars, calculate the actual spacing according to the formula, which is as follows:

[0047] Where, is the vertical distance between two adjacent parallel steel bars; This is the common coefficient for parallel straight reinforcing bars; is the constant term for two parallel reinforcing bars; is the coefficient vector. The modulus length; The YOLOv3 deep learning model is used to identify ligation points, and the detection accuracy is optimized through a loss function. The specific formula is as follows:

[0048] in, The value of the loss function; Weights for coordinate loss; Weights for target confidence loss; Weights for the no-target confidence loss; The coordinates of the center of the prediction box; Width and height; Confidence in the detection frame; Be confident in the true frame; For classification probability; This represents the true classification probability. The number of missing binding points within each preset length is counted, and the deviation rate of rebar spacing and the pass rate of binding points are calculated.

[0049] (2) When the test item is the elastic modulus of concrete, the specific process of steps S4 to S6 includes: The applied force is recorded in real time using pressure sensors and displacement gauges. With deformation Calculate the nominal stress and strain The specific formula is as follows:

[0050]

[0051] in, The cross-sectional area of ​​the specimen; Gauge length; The least squares method was used to fit the stress and strain data in the linear stage, and the elastic modulus was... The following formula is used to derive:

[0052] in, Number of data sets; The stress value is the value of the i-th data set. The strain value is the value of the i-th data set. for The sum of the products of stress and strain; It is the sum of stress values; It is the sum of strain values; The sum of the squares of the strain values; Simultaneously calculate the coefficient of determination. Verify linearity:

[0053] in, For the first The actual stress values ​​of the set of data; To fit the stress value; This represents the average value of the actual stress. like or If the value exceeds the design range, the test is deemed invalid and will be automatically retested.

[0054] (3) When the inspection item is the location inspection of the pre-buried pipeline, the specific process of steps S4 to S6 includes: A 3D point cloud dataset of the pipeline was acquired and reconstructed using a structured light depth camera. The point cloud density is no less than 100 points / cm²; the RANSAC algorithm is used to robustly estimate the pipe axis, specifically including: A predetermined number of points are randomly selected to construct a hypothetical straight line model. The Euclidean distance from all points to this straight line is calculated using the following formula:

[0055] in, The coordinates of the reference point; For the first The spatial coordinates of the points; It is the direction vector; Statistical satisfaction The number of interior points, when the proportion of interior points The iteration stops when the time is right, and the total number of iterations is determined by the following formula:

[0056] in, Minimum number of iterations; The preset expected confidence level, The ratio of interior points; The fitted pipe axis is spatially aligned with the theoretical axis in the BIM design model, and the maximum Euclidean distance between the axes is calculated. ,like Generate a position offset alarm; The RANSAC algorithm is used to robustly extract the pipeline axis from the noisy point cloud, and the position is compared with the BIM model.

[0057] (4) When the inspection item is the external dimensions of the component, the specific processes of steps S4 to S6 include: The entire surface contour data of the component is acquired using a lidar scanner at a scanning speed of 1000 points / second, obtaining at least 5×10⁻⁶ points. 5 One outline point; Contour extraction is performed using the Canny edge detection algorithm: the image is smoothed using a 5×5 Gaussian filter with a standard deviation σ=1.5, and the gradient is calculated using the Sobel operator. The specific formula is as follows:

[0058] Nonmaximum suppression and double thresholding ( , Connect the edges; perform principal component analysis on the contour point set, solve for the covariance matrix, take the eigenvector corresponding to the largest eigenvalue as the major axis direction, and calculate the minimum bounding rectangle size. The specific formula is as follows:

[0059] in, The number of contour points, Let be the mean vector of the point set. Indicates matrix transpose; For the first Two-dimensional coordinates of a point; Calculate the actual length of the component With design value The deviation rate is calculated using the following formula: If the deviation rate is ≤3%, the external dimensions are deemed acceptable; if the deviation rate is >3%, the out-of-tolerance area is marked in the 3D point cloud.

[0060] Example 2 See Figure 2 , Figure 2This is a structural diagram of an inspection device for precast beam yards proposed in this invention. It consists of a self-propelled omnidirectional mobile chassis 3, a multi-degree-of-freedom robotic arm system 2, a vision inspection module 1, a sensor system 5, a control and communication system 4, and a power supply and charging system. The modules are connected via a CAN bus (control commands), Ethernet (data transmission), and power lines (power supply). The overall dimensions are 1200mm × 800mm × 1800mm (including the extended height of the robotic arm), and the total weight is approximately 150kg. It is suitable for use in precast beam yard operations. Specifically, it may include: M1, self-propelled omnidirectional mobile chassis 3: used to carry the multi-degree-of-freedom robotic arm system 2, vision inspection module 1, sensor system 5, control and communication system 4, and power supply and charging system, and to move in the precast beam yard environment; The self-propelled omnidirectional mobile chassis 3 adopts a Q235 steel frame with a wall thickness of 5mm for structural load-bearing. The top has six pre-drilled M8 mounting holes with a hole spacing of 150mm×150mm, providing a maximum load capacity of 200kg. A 5mm thick rubber shock-absorbing pad is added to the bottom to control vibration amplitude to ≤0.1mm, preventing interference with precision testing components. The drive and navigation system is equipped with a dual-steering wheel drive, model AGV-200, with a speed of 500r / min and a torque of 15N. m, supporting omnidirectional movement from 0-1.5m / s; It can flexibly adapt to the narrow passages and rough ground of precast beam yards, and integrates a 16-line LIDAR sensor (model RS-LIDAR-M1) and SLAM navigation system with a positioning accuracy of ≤50mm, which can accurately lock the position of the component to be inspected. The obstacle avoidance and positioning functions are realized through 4 sets of ultrasonic sensors (model UCM-05, detection range 0.1-5m) and 2 sets of infrared sensors (model IR-300, detection range 0.1-3m). When an obstacle is detected, it can reduce speed or stop in time. With the addition of a QR code recognition module with an accuracy of ≤10mm, it can complete the precise alignment of the chassis and the component. This module mainly realizes the functions of equipment movement, obstacle avoidance and positioning.

[0061] M2, Multi-degree-of-freedom robotic arm system 2: mounted on the self-propelled omnidirectional mobile chassis 3, with a detachable actuator at its end; used to drive the detection components of the vision inspection system and the detection components of the sensor system 5 to move to the area of ​​the component to be inspected; The multi-DOF robotic arm system 2 adopts a 6-DOF serial structure, model FR50, and is made of aerospace aluminum alloy. While ensuring strength, it reduces the overall weight. The arm span ranges from 0.3 to 1.8m, and can cover the inspection areas such as the sides, top, and bottom of precast beams. The repeatability is ≤0.1mm, the rated load is 5kg, and it can stably carry the vision inspection module 1 and sensors. The verticality error of the base during installation is controlled within ≤0.05° to ensure the accuracy of posture adjustment. Its end effector adopts a detachable snap-on design, with a replacement time of ≤2min, and integrates vision and sensor interfaces, allowing for quick replacement of corresponding components according to different inspection items. In terms of motion control, the built-in RRT* algorithm can automatically generate obstacle avoidance trajectories, meet the constraints of joint rotation angle -180°~180° and movement speed ≤5° / s, and the time for a single attitude adjustment is ≤10s. It is mainly used to detect the spatial attitude adjustment of the execution component and ensure that the detection component is accurately aligned with the target area.

[0062] M3, Visual Inspection Module 1: Located on the actuator, including a high-definition RGB camera and a structured light depth camera, used to acquire component images and three-dimensional point cloud data; The hardware configuration of the visual inspection module 1 includes a 4K high-definition RGB camera and a structured light depth camera. The RGB camera is a GC5033 with a resolution of 4096×2160 and a lens focal length of 8mm, which can clearly capture detailed images of the component surface. The structured light depth camera is a D455 with a ranging range of 0.3-3m and a depth accuracy of ±2mm, which can generate high-density 3D point cloud data. Data processing relies on the NVIDIA Jetson AGXXavier AI chip, with a computing power of 32TOPS, supporting real-time preprocessing and feature extraction of the acquired images and point cloud data without relying on external computing equipment. To adapt to the dusty construction environment of the precast beam yard, the module is also designed with a dustproof glass cover (transmittance ≥95%), with an overall protection level of IP65, effectively preventing dust from entering and affecting the inspection accuracy. This module mainly realizes image and point cloud data acquisition and data preprocessing functions, laying the foundation for subsequent feature extraction and index calculation.

[0063] M4, Sensor System 5: Connected to the control and communication system 4, including pressure sensor, ultrasonic sensor, lidar and displacement gauge, used to collect mechanical and geometric parameters; Sensor system 5 includes various types of sensors to collectively acquire multi-dimensional detection data. Among them, the pressure sensor (model PT124G-111, range 0-50MPa, accuracy ±0.2%FS) is used to acquire the applied force F during concrete elastic modulus testing, and then calculate the stress σ, corresponding to the concrete elastic modulus testing and data processing steps; the ultrasonic sensor (model UCM-05, detection range 0.02-2m, accuracy ±0.1mm) is mainly used to detect the surface roughness of components, providing auxiliary data for dimensional inspection; the lidar (model RS-LIDAR-M1, detection range 0.1-20m, accuracy ±0.5mm) can acquire component contour point clouds at a rate of 1000 points / s for dimensional inspection, and can also scan the surrounding environment during equipment movement to assist in obstacle avoidance. The displacement gauge, model DT3800, has a range of 0-50mm and an accuracy of ±0.01mm. It is used to collect the deformation ΔL during concrete loading and then calculate the strain ε, corresponding to concrete elastic modulus detection and data processing. The force control sensor, model FT300, has a range of 0-100N and an accuracy of ±0.1N. It detects the firmness of the rebar binding points by applying a preset pressure, corresponding to rebar binding quality detection. All sensors are connected via RS485 bus with a sampling frequency of 10Hz to ensure real-time data transmission to the control and communication system 4, jointly realizing the acquisition of mechanical and geometric parameters.

[0064] M5 and Control and Communication System 4: respectively establish signal connections with the self-propelled omnidirectional mobile chassis 3, multi-degree-of-freedom robotic arm system 2, vision inspection module 1 and sensor system 5, to coordinate the collaborative work of each module, and receive data collected by vision inspection module 1 and sensor system 5, process it, and output the inspection results. The control and communication system 4 is the "central hub" of the device. Its hardware core is an industrial computer, model IPC-610L, equipped with an Intel Core i7-10700 processor, 16GB of memory and 512GB of SSD storage. It also has multiple interfaces such as CAN bus, RS485, and Ethernet, and can stably connect to the self-propelled chassis, robotic arm, vision module and sensors. The software system adopts the ROSNoetic operating system based on ARM architecture, and is pre-installed with all the algorithms in the supplementary materials, including Hough transform (for straight line extraction of rebar), YOLOv3 (for tie point identification), RANSAC (for pipe axis fitting), least squares method (for elastic modulus calculation), etc., which can efficiently complete data processing and detection index calculation. The communication function is achieved through a 5G module (model Huawei MH5000-31, upload speed ≥100Mbps) and a Wi-Fi 6 module (transmission speed ≥1.2Gbps). The 5G module is used to transmit large-capacity test reports to the cloud platform, and the Wi-Fi 6 module is used to communicate with the construction personnel's terminals. The system is mainly responsible for receiving test tasks, processing data, judging results, and transmitting test reports, and coordinating the collaborative work of various modules.

[0065] M6, Power Supply and Charging System: Used to provide power to the self-propelled omnidirectional mobile chassis 3, multi-degree-of-freedom robotic arm system 2, vision inspection module 1, sensor system 5 and control and communication system 4.

[0066] The power supply and charging system provides power assurance for the entire process of the device operation. It adopts two sets of 24V / 100Ah lithium iron phosphate batteries, model LP100-24, with a cycle life of ≥2000 cycles and a runtime of ≥8 hours when fully charged. It can continuously complete the full-item inspection of 10 30m long precast beams. To avoid interruption due to insufficient power during the inspection process, the backup battery adopts a snap-on design, with a replacement time of ≤5 minutes, allowing staff to quickly replace the battery. When the battery level is ≤15%, the control and communication system 4 will automatically trigger a charging command and guide the chassis to move to a dedicated charging station via SLAM navigation. The charging station uses 150W electromagnetic induction wireless charging technology with a charging efficiency of ≥85%. The time required to charge the battery from 15% to full charge is ≤2 hours. During the charging process, the charging status can also be fed back to the cloud in real time for easy remote monitoring.

[0067] The above specific embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to examples, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A quality inspection method for precast beam yards, characterized in that, include: S1. Obtain inspection task information, which includes the location parameters of the precast component to be inspected, at least one inspection item, and the design standard threshold of the corresponding inspection item. S2. Control the inspection equipment to move to the preset inspection area of ​​the precast component to be inspected, and identify and avoid obstacles in the construction environment in real time during the movement. S3. According to the detection items, adjust the spatial posture of the detection execution component so that the detection execution component is aligned with the target detection part of the prefabricated component to be detected, and maintains a preset safe distance. S4. The detection data of the target detection area is collected by the detection execution component. The detection data includes image data, three-dimensional point cloud data, mechanical parameter data and geometric parameter data. S5. Preprocess and extract features from the collected detection data, and call the preset algorithm to calculate the detection index value; S6. Compare the test index value with the corresponding design standard threshold to determine whether the prefabricated component to be tested meets the quality requirements. S7. Generate a test report containing test index values, judgment results, and anomaly markers, and transmit it to the data management platform.

2. The quality inspection method for precast beam yards according to claim 1, characterized in that, The acquisition of detection task information specifically includes: The system receives structured task instructions from the cloud via a wireless communication module. These instructions include the unique identifier code of the prefabricated component to be inspected, BIM model coordinate data, and design standard thresholds for each inspection item.

3. The quality inspection method for precast beam yards according to claim 1, characterized in that, The process of controlling the inspection equipment to move to the preset inspection area of ​​the precast component to be inspected specifically includes: A three-dimensional map of the precast beam yard environment is constructed based on autonomous navigation system technology, and the optimal movement path is planned according to the location parameters of the precast components to be detected. The system uses LiDAR to scan the surrounding environment in real time. When an obstacle is detected, the straight-line distance between the obstacle and the inspection equipment is calculated. ; like The inspection equipment is controlled to move along the planned path at a speed of 0.8-1.5 m / s; like Control the inspection equipment to reduce speed to ; like The inspection equipment is stopped and a warning signal is issued. It is restarted after the obstacle is removed. The inspection equipment is precisely positioned and aligned with the prefabricated components to be inspected by using QR code recognition technology.

4. The quality inspection method for precast beam yards according to claim 1, characterized in that, The process of adjusting the spatial orientation of the detection execution component specifically includes: Based on the BIM model data of the prefabricated component to be inspected, determine the three-dimensional coordinates of the target inspection location; The RRT* motion planning algorithm is invoked to generate an obstacle avoidance trajectory from the initial posture to the target posture, and the obstacle avoidance trajectory satisfies the kinematic constraints of the detection execution component; The multi-degree-of-freedom actuator is driven to move along the obstacle avoidance trajectory, so that the detection center of the detection actuator is aligned and the distance between the detection actuator and the target detection part is maintained within the range of 0.3-1.0m; Once the attitude adjustment is complete, a ready signal is output to trigger the data acquisition process.

5. The quality inspection method for precast beam yards according to claim 1, characterized in that, The process of generating the test report specifically includes: It automatically summarizes the raw data, calculation process, and test index values ​​of each test item to form a structured data table; On the 3D model of the prefabricated component to be inspected, abnormal areas are marked according to the world coordinate system, and the abnormal areas are highlighted in red. The system automatically matches the rectification suggestion library based on the anomaly type and generates a text description that includes rectification measures, standard basis, and acceptance methods. The data tables, 3D models, and rectification suggestions are integrated into a PDF report, which is then uploaded to the data management platform via a 5G communication module and simultaneously pushed to designated terminal devices.

6. The quality inspection method for precast beam yards according to claim 1, characterized in that, The testing items include: rebar binding quality testing, embedded pipe location testing, component shape and size testing, and concrete elastic modulus testing.

7. The quality inspection method for precast beam yards according to claim 6, characterized in that, The specific processes of steps S4 to S6 during the inspection of the rebar binding quality include: Color images of the rebar binding area are acquired using a high-definition camera; the color images are then weighted and converted to grayscale, transforming the RGB images into grayscale images; a 3×3 window midpoint filtering algorithm is used to remove image noise according to the following formula: in, After median filtering, the coordinates The grayscale value of the pixel at that location; Within the filtering window, coordinates The original grayscale value of the pixel; This is a median calculation function; The straight line features of the reinforcing bars are extracted by Hough transform, and the peak points are detected in the parameter space according to the straight line equation to extract the straight line profile of the reinforcing bars.

8. The quality inspection method for precast beam yards according to claim 6, characterized in that, The specific process of calling the preset algorithm to calculate the detection index value includes: For the extracted straight lines of adjacent parallel reinforcing bars, calculate the actual spacing according to the formula, which is as follows: Where, is the vertical distance between two adjacent parallel steel bars; This is the common coefficient for parallel straight reinforcing bars; is the constant term for two parallel reinforcing bars; is the coefficient vector. The modulus length; The YOLOv3 deep learning model is used to identify ligation points, and the detection accuracy is optimized through a loss function. The specific formula is as follows: in, The value of the loss function; Weights for coordinate loss; Weights for target confidence loss; Weights for the no-target confidence loss; The coordinates of the center of the prediction box; Width and height; Confidence in the detection frame; Believe the truth box; For classification probability; This represents the true classification probability. The number of missing binding points within each preset length is counted, and the deviation rate of rebar spacing and the pass rate of binding points are calculated.

9. The quality inspection method for precast beam yards according to claim 6, characterized in that, When the test item is the elastic modulus of concrete, the specific processes of steps S4 to S6 include: The applied force is recorded in real time using pressure sensors and displacement gauges. With deformation Calculate the nominal stress and strain The specific formula is as follows: in, The cross-sectional area of ​​the specimen; Gauge length; The least squares method was used to fit the stress and strain data in the linear stage, and the elastic modulus was... The following formula is used to derive: in, Number of data sets; The stress value is the value of the i-th data set. The strain value is the value of the i-th data set. for The sum of the products of stress and strain; It is the sum of stress values; It is the sum of strain values; The sum of the squares of the strain values; Simultaneously calculate the coefficient of determination. Verify linearity: in, For the first The actual stress values ​​of the set of data; To fit the stress value; This represents the average value of the actual stress. like or If the value exceeds the design range, the test is deemed invalid and will be automatically retested.

10. The quality inspection method for precast beam yards according to claim 6, characterized in that, When the detection item is the location detection of pre-buried pipes, the specific processes of steps S4 to S6 include: A 3D point cloud dataset of the pipeline was acquired and reconstructed using a structured light depth camera. The point cloud density is no less than 100 points / cm²; the RANSAC algorithm is used to robustly estimate the pipe axis, specifically including: A predetermined number of points are randomly selected to construct a hypothetical straight line model. The Euclidean distance from all points to this straight line is calculated using the following formula: in, The coordinates of the reference point; For the first The spatial coordinates of the points; It is the direction vector; Statistical satisfaction The number of interior points, when the proportion of interior points The iteration stops when the time is right, and the total number of iterations is determined by the following formula: in, Minimum number of iterations; The preset expected confidence level, The ratio of interior points; The fitted pipe axis is spatially aligned with the theoretical axis in the BIM design model, and the maximum Euclidean distance between the axes is calculated. ,like Generate a position offset alarm; The RANSAC algorithm is used to robustly extract the pipeline axis from the noisy point cloud, and the position is compared with the BIM model.

11. The quality inspection method for precast beam yards according to claim 6, characterized in that, When the inspection item is the external dimensions of the component, the specific processes of steps S4 to S6 include: The entire surface contour data of the component is acquired using a lidar scanner at a scanning speed of 1000 points / second, obtaining at least 5×10⁻⁶ points. 5 One outline point; Contour extraction is performed using the Canny edge detection algorithm: the image is smoothed using a 5×5 Gaussian filter with a standard deviation σ=1.5, and the gradient is calculated using the Sobel operator. The specific formula is as follows: Nonmaximum suppression and double thresholding ( , Connect the edges; perform principal component analysis on the contour point set, solve for the covariance matrix, take the eigenvector corresponding to the largest eigenvalue as the major axis direction, and calculate the minimum bounding rectangle size. The specific formula is as follows: in, The number of contour points, Let the mean vector of the point set be... Indicates matrix transpose; For the first Two-dimensional coordinates of a point; Calculate the actual length of the component With design value The deviation rate is calculated using the following formula: If the deviation rate is ≤3%, the external dimensions are deemed acceptable; if the deviation rate is >3%, the out-of-tolerance area is marked in the 3D point cloud.

12. An inspection device for a precast beam yard, characterized in that, include: Self-propelled omnidirectional mobile chassis: used to carry multi-degree-of-freedom robotic arm systems, vision inspection modules, sensor systems, control and communication systems, and power supply and charging systems, and to move in the precast beam yard environment; Multi-degree-of-freedom robotic arm system: mounted on the self-propelled omnidirectional mobile chassis, with a detachable actuator at its end; Used to move the detection components of the vision inspection module and the sensor system to the area of ​​the component to be inspected; Visual inspection module: Located on the detachable actuator, it includes a high-definition RGB camera and a structured light depth camera, used to acquire component images and 3D point cloud data; Sensor system: connected to the control and communication system, including pressure sensors, ultrasonic sensors, lidar and displacement gauges, used to collect mechanical and geometric parameters; Control and communication system: Establishes signal connections with the self-propelled omnidirectional mobile chassis, multi-degree-of-freedom robotic arm system, vision inspection module and sensor system respectively, to coordinate the collaborative work of each module, and to receive data collected by the vision inspection module and sensor system, process it and output the inspection results; Power supply and charging system: used to provide power to the self-propelled omnidirectional mobile chassis, multi-degree-of-freedom robotic arm system, vision inspection module, sensor system and control and communication system.

Images