A method and system for in-situ 3D printing repair of dam defects

Abstract

The application provides a dam defect in-situ 3D printing repair method and system, and belongs to the technical field of dam structure repair. The method comprises the following steps: obtaining a point cloud image of a to-be-repaired area through global scanning, accurately registering the point cloud image with a dam original BIM design model, extracting key parameters and generating a standardized defect repair list; completing second-order intelligent path planning based on the point cloud image, assigning multiple machine tasks in combination with the repair list and printing unmanned aerial vehicle performance parameters, and generating task slices; calling the printing unmanned aerial vehicle and the mobile stock bin to carry out repair, continuously detecting and correcting the path in a closed loop during the repair, and storing the report after the repair through global scanning acceptance. The application realizes full coverage scanning of the dam surface and underwater by using air-water integrated scanning unmanned aerial vehicles, automatically completes defect identification and positioning in combination with edge computing and deep learning, reduces the risk of high-altitude operation, improves the detection efficiency and parameter extraction accuracy, and provides stable and reliable digital support for dam repair.

Description

Technical Field

[0001] This invention belongs to the field of dam structure repair technology, and in particular relates to a method and system for in-situ 3D printing repair of dam defects. Background Technology

[0002] Existing technologies primarily address the repair of freeze-thaw erosion, scour pits, and cracks on the concrete surface of 100-meter-high dams. Traditionally, this involves erecting cantilevered scaffolding or suspended platforms on the dam crest or abutment, with workers carrying steel formwork, cement mortar guns, or spraying machines to roughen, reinforce, erect formwork, and apply layered spraying or troweling to the defective areas, followed by water curing. To shorten the project timeline, some projects utilize single DJI M300 industrial drones equipped with small 3D printing nozzles. Ground-based truck-mounted pumps deliver fast-hardening mortar through 100-meter nylon pipes, and the drone stacks the mortar layer by layer according to preset waypoints, completing the "aerial printing" repair. The steps can be summarized as follows: manual surveying and mapping using a suspended platform to establish a two-dimensional defect log; manual selection of the repair area in BIM to generate contour-level slicing paths; single-drone printing along fixed waypoints, with a layer height of 1cm, and continuous material supply from the pump truck; and a second inspection using a suspended platform after printing, with manual smoothing of any areas exceeding tolerances.

[0003] However, because this method still uses an open process of "single UAV + fixed waypoint + single-layer 1cm", the following problems still exist: 1. Single-machine constraints - The drone's payload is ≤5kg and its flight time is ≤15min. Frequent round trips are required to change batteries and refill materials. Large-area defects need to be divided into multiple days and shifts, so the overall construction period is still long. 2. Insufficient precision – There is no online detection closed loop, the layer width cannot be adjusted in real time at the curvature change of the dam surface, and step misalignment occurs after stacking. The forming accuracy is only ≈1cm, which does not meet the impact resistance and wear resistance <5mm specification. 3. Collision risk – When operating alone, the gantry crane and cable crane on the dam top operate as usual according to the production plan. The preset waypoints cannot predict dynamic obstacles, and there have been accidents where the cable scraped against the drone. 4. Path redundancy - Two-dimensional high-level slicing does not consider the curvature of the surface, resulting in local overspray / underspray up to 15%, which wastes material and requires manual secondary leveling; 5. Intermittent material supply - When the height difference between the vehicle-mounted pump and the drone is greater than 100m, the back pressure of the hose is large, requiring the machine to be stopped to clear the blockage, resulting in cold joints between layers and reducing the interlayer bonding strength.

[0004] Therefore, existing "scaffolding + template" or "single-machine aerial printing" technologies all suffer from common defects in the complex environment of high dams, such as low efficiency, poor accuracy, high safety risks, and inability to operate continuously over large areas. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a method and system for in-situ 3D printing repair of dam defects.

[0006] Firstly, the method includes, A full-area scan of the dam was performed to obtain point cloud images of the area to be repaired; the point cloud images of the area to be repaired were then precisely registered with the original BIM design model of the dam to extract key parameters of the dam and generate a standardized list of defects to be repaired. Based on the point cloud image of the area to be repaired, second-order intelligent path planning is performed, and multi-machine task allocation is carried out based on the standardized defect repair list and the performance parameters of the printing drone, generating task slices. Based on the task slices, printing drones and mobile material silos were deployed to repair the areas of the dam awaiting repair. During the repair process, repair detection and repair path correction are carried out in a closed loop iterative manner at a certain period. After the repair is completed, a full-domain scan is performed to conduct a repair acceptance test, and an acceptance report is stored to complete the repair operation.

[0007] Furthermore, the full-area scanning of the dam specifically includes using an air-water integrated scanning drone to perform a full-area scan of the dam and obtain the original point cloud image of the dam; The acquisition of point cloud images of the area to be repaired specifically includes: using an edge GPU server to perform noise reduction, filtering, and downsampling preprocessing on the original point cloud image of the dam; performing semantic segmentation through a PointRCNN++ deep learning network; and automatically identifying defects such as freeze-thaw erosion, scour pits, cracks, and cavitation pits according to a preset threshold to generate point cloud images of the area to be repaired.

[0008] Furthermore, the generation of the standardized defect repair list specifically includes using an ICP+CPD hybrid registration algorithm to accurately register the point cloud image of the area to be repaired with the original BIM design model of the dam, controlling the registration error, extracting key parameters of the dam defects, and generating a standardized defect repair list.

[0009] Furthermore, the second-order intelligent path planning specifically includes utilizing the connected Fermat spiral algorithm and RRT. The SF safe field algorithm is used for second-order intelligent path planning.

[0010] Furthermore, the continuous closed-loop iteration for repair detection and repair path correction specifically includes, When repairing at least two layers to be repaired, stop the local repair work and start the air-water integrated scanning drone to quickly scan the repaired area and obtain the point cloud image of the repair. The repaired point cloud image is compared with the original BIM design model of the dam to obtain the repair deviation value of the whole area and form a deviation distribution cloud map. Based on the repair deviation value and the deviation distribution cloud map, the deviation level is determined. If the repair deviation is greater than the preset threshold, a path correction command is immediately generated and sent to the printing drone to automatically adjust the repair trajectory of subsequent layers. Conversely, if the repair deviation is less than the preset threshold, the printing drone does not need to correct the path and continues the repair task.

[0011] Furthermore, after the repair is completed, a full-domain scan is performed to conduct a repair acceptance test, and an acceptance report is stored to complete the repair operation. Specifically, this includes... After all repairs are completed, an air-water integrated scanning drone is used to perform a final full-area scan to detect the defect filling rate, surface flatness, and contour dimensions, ensuring that there are no missed sprays, no oversprays, no voids, and no cracks; and to collect data on dam geometry, drone pose, material condition, environmental parameters, and repair process. Based on the test results and collected data, an acceptance report is generated that includes point cloud images, photographs, process parameters, and performance indicators. The hash value of the acceptance report is stored on the blockchain to ensure data immutability and remote acceptance, thus completing the entire process of repair.

[0012] Secondly, the system includes: a scanning module, a task generation module, a repair module, a path correction module, and an acceptance module; The scanning module is used to perform a full-area scan of the dam to obtain point cloud images of the area to be repaired; and to accurately register the point cloud images of the area to be repaired with the original BIM design model of the dam, extract key parameters of the dam, and generate a standardized defect repair list. The task generation module is used to perform second-order intelligent path planning based on the point cloud image of the area to be repaired, and to allocate tasks to multiple drones based on the standardized defect repair list and the performance parameters of the printing drone, thereby generating task slices. The repair module is used to call upon the printing drone and mobile silo to repair the area of ​​the dam to be repaired, based on the task slice; The path correction module is used to perform repair detection and repair path correction in a closed loop iterative manner at a certain period during the repair process; The acceptance module is used to perform a full-domain scan after the repair is completed, to conduct repair acceptance, and to store the acceptance report, thus completing the repair operation.

[0013] Furthermore, the scanning module is specifically used to perform a full-area scan of the dam using an integrated air-water scanning drone to obtain the original point cloud image of the dam; The original point cloud image of the dam is preprocessed by denoising, filtering, and downsampling using an edge GPU server, and semantic segmentation is performed by a PointRCNN++ deep learning network. Defects such as freeze-thaw erosion, scour pits, cracks, and cavitation pits are automatically identified according to preset thresholds to generate point cloud images of the area to be repaired.

[0014] Furthermore, the scanning module is specifically used to accurately register the point cloud image of the area to be repaired with the original BIM design model of the dam using an ICP+CPD hybrid registration algorithm, control the registration error, extract key parameters of dam defects, and generate a standardized defect repair list.

[0015] Furthermore, the task generation module is specifically used to utilize the connected Fermat spiral algorithm and RRT. The SF safe field algorithm is used for second-order intelligent path planning.

[0016] Furthermore, the path correction module is specifically used for: When repairing at least two layers to be repaired, stop the local repair work and start the air-water integrated scanning drone to quickly scan the repaired area and obtain the point cloud image of the repair. The repaired point cloud image is compared with the original BIM design model of the dam to obtain the repair deviation value of the whole area and form a deviation distribution cloud map. Based on the repair deviation value and the deviation distribution cloud map, the deviation level is determined. If the repair deviation is greater than the preset threshold, a path correction command is immediately generated and sent to the printing drone to automatically adjust the repair trajectory of subsequent layers. Conversely, if the repair deviation is less than the preset threshold, the printing drone does not need to correct the path and continues the repair task.

[0017] Furthermore, the acceptance module is specifically used for, After all repairs are completed, an air-water integrated scanning drone is used to perform a final full-area scan to detect the defect filling rate, surface flatness, and contour dimensions, ensuring that there are no missed sprays, no oversprays, no voids, and no cracks; and to collect data on dam geometry, drone pose, material condition, environmental parameters, and repair process. Based on the test results and collected data, an acceptance report is generated that includes point cloud images, photographs, process parameters, and performance indicators. The hash value of the acceptance report is stored on the blockchain to ensure data immutability and remote acceptance, thus completing the entire process of repair.

[0018] Thirdly, a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described in-situ 3D printing repair methods for dam defects.

[0019] Fourthly, an electronic device includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus. Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements any of the steps of the above-described in-situ 3D printing repair method for dam defects.

[0020] Compared with the prior art, the present invention has the following advantages: 1. This invention uses an air-water integrated scanning drone to achieve full coverage scanning of the dam surface and underwater area. Combined with edge computing and deep learning networks, it automatically completes defect identification, segmentation and localization, replacing the traditional manual surveying method using scaffolding and suspended platforms, and significantly reducing the safety hazards of working at heights. Furthermore, through precise registration of point cloud and BIM model, defect parameters are extracted accurately and reliably, and the detection efficiency is improved several times, providing a stable and reliable digital foundation for subsequent repairs, realizing the transformation from "human experience judgment" to "data-driven precision repair".

[0021] 2. By utilizing second-order intelligent path planning combined with the performance of UAVs for dynamic task allocation, problems such as short single-unit endurance, limited payload, and discontinuous operation can be effectively solved. This enables continuous and uninterrupted construction of large-area defects, significantly shortening the construction period. During the repair process, periodic detection and real-time path correction can be performed to correct deviations in a timely manner, reduce over-spraying and under-spraying, improve material utilization, ensure surface flatness and structural density, and make the repair layer more reliable and durable, meeting the stringent specifications for dam erosion resistance, wear resistance, frost resistance, and seepage prevention.

[0022] 3. By collecting, recording, and synchronizing data throughout the entire construction process in real time, automated quality acceptance is completed through full-domain scanning after repairs are finished. A standardized and complete acceptance report is generated and stored on the blockchain as evidence, ensuring the authenticity and immutability of the acceptance results. This model simplifies on-site management processes, improves operation and maintenance control, supports remote acceptance and long-term archive management, and significantly enhances the intelligence, standardization, and normalization of dam operation and maintenance.

[0023] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A schematic diagram of the process of an in-situ 3D printing repair method for dam defects according to the present invention is shown.

[0026] Figure 2 A schematic diagram of a dam defect in-situ 3D printing repair system module of the present invention is shown. Detailed Implementation

[0027] 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 embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] like Figure 1 As shown, this invention proposes an in-situ 3D printing repair method for dam defects using integrated air-water 3D scanning and multi-machine collaboration. High-precision point cloud acquisition is achieved through an integrated air-water ScanDrone; collaborative printing using 4-8 BuildDrone units combined with continuous pumping from a ground-based mobile silo enables rapid repair of large-area defects; and employs... Second-order path planning and online task reallocation are performed to solve the problems of large curvature variations on the dam surface and easy misalignment between layers; distributed ORCA collision avoidance and UWB-RTK centimeter-level positioning ensure safe operation in all weather conditions; and printing-inspection closed loop and real-time correction of differential point cloud are implemented. The steps include: S1. Perform a full-area scan of the dam to obtain point cloud images of the area to be repaired; then accurately register the point cloud images of the area to be repaired with the original BIM design model of the dam, extract key parameters of the dam, and generate a standardized defect repair list.

[0029] In this embodiment, the present invention utilizes a ScanDrone air-water integrated scanning drone to perform a full-area scan of the dam, the steps of which include: a. Place ScanDrone on the dam crest safety take-off and landing platform, complete the power-on self-test, confirm that the lidar, multibeam sonar, underwater camera, positioning module, and communication module are working properly, and establish a stable communication link with the edge GPU server; b. ScanDrone performs a zigzag full-area scan along the dam face to acquire the original point cloud image of the dam; among them, the dam facade area uses a 32-line lidar to collect point clouds, while the water level fluctuation area and underwater area are switched to 130kHz multibeam sonar and pressure-resistant underwater lens operation, acquiring no less than 20 million point cloud data within 5 minutes, with a point cloud density of no less than 500pts / m².

[0030] In this embodiment, the present invention utilizes an edge GPU server to perform noise reduction, filtering, and downsampling preprocessing on the original point cloud image of the dam, and completes semantic segmentation through a PointRCNN++ deep learning network. Based on preset thresholds (such as depth ≥5mm, area ≥0.02m²), it automatically identifies defects such as freeze-thaw erosion, scour pits, cracks, and cavitation pits, and generates point cloud images of the area to be repaired.

[0031] In this embodiment, the present invention uses the ICP+CPD hybrid registration algorithm to accurately register the point cloud image of the area to be repaired with the original BIM design model of the dam, control the registration error (e.g., error ≤ 3mm), extract key parameters such as the three-dimensional coordinates, normal vector, principal curvature, area, and volume of the dam defects, and generate a standardized defect repair list.

[0032] S2. Based on the point cloud image of the area to be repaired, perform second-order intelligent path planning, and based on the standardized defect repair list and the performance parameters of the BuildDrone printing drone, perform multi-machine task allocation and generate task slices.

[0033] In this embodiment, the present invention utilizes the Connected Fermat Spirals Algorithm (MCFS) and The safe field algorithm for second-order intelligent path planning includes the following steps: a. Use the MCFS algorithm to perform coarse-level planning, divide the point cloud image of the area to be repaired into several wide rings (e.g., 0.8m) according to the contour lines of the dam surface, and generate an initial trajectory that is fully covered, non-intersecting, and without omissions; b. Based on The safety field algorithm performs fine-grained optimization in the dam's 5D state space (e.g., x, y, z, yaw ψ, pitch θ), through... The protocol reads the real-time trajectory of the gantry crane and cable crane PLC on the dam top, transforms the dynamic sling into a virtual cylindrical obstacle, realizes real-time dynamic obstacle avoidance, and outputs the optimal and refined waypoints.

[0034] In this embodiment, the present invention allocates tasks to multiple printing drones according to a standardized defect repair list and BuildDrone performance parameters, based on a certain single block volume, to generate task slices. Within a single task slice, 4–8 printing drones are allocated according to the principles of balanced load and proximity of operation.

[0035] Optionally, the present invention also establishes an online task monitoring mechanism. When any BuildDrone's battery level is less than 30% or its remaining material weight is less than 2kg, it triggers a relay redistribution of neighboring machines, transferring unfinished repair tasks according to the nearest neighbor principle to ensure continuous and uninterrupted repair.

[0036] S3. Based on the task slice, call BuildDrone and mobile silos to repair the dam repair area; in a single task slice, this invention is allocated to 4-8 BuildDrone according to the principle of balanced load and nearby operation.

[0037] In this embodiment, after the BuildDrone takes off sequentially to the area to be repaired, the present invention uses UWB-RTK fusion positioning to locate the task area at the centimeter level, and maintains stable hovering after arriving at the task area to begin repair.

[0038] Optionally, BuildDrone adaptively adjusts the nozzle extrusion rate based on the real-time principal curvature K of the dam surface to ensure that the layer width / layer height remains stable at 4±0.2mm; wherein the adaptive adjustment formula is expressed as follows:

[0039] in, The speed of mouth lateral movement, For extrusion rate, This represents the cross-sectional area.

[0040] Optionally, after each layer is repaired, BuildDrone's micro-pump reverses to achieve material recovery during downtime, preventing material dripping, and the moving hopper continuously pumps during the repair process to achieve high-altitude backpressure-free and uninterrupted material supply, eliminating cold seams between layers.

[0041] In this embodiment, for underwater defect repair, the present invention uses a mobile hopper to switch to an underwater-specific formula, and adopts a variable layer thickness strategy and dual-machine alternating operation to overcome water flow disturbance and ensure underwater molding quality.

[0042] S4. During the repair process, repair detection and repair path correction are carried out in a closed loop iterative manner at a certain period.

[0043] In this embodiment, when repairing at least two layers to be repaired, the present invention stops the local repair work and starts ScanDrone to quickly scan the repaired area, obtains the repair point cloud image, and compares the repair point cloud image with the original BIM design model of the dam to obtain the repair deviation value of the whole area and form a deviation distribution cloud map.

[0044] In this embodiment, the present invention determines the deviation level based on the repair deviation value and the deviation distribution cloud map. If the repair deviation is greater than a preset threshold (e.g., 2mm), a path correction instruction is immediately generated, and a lateral offset correction is performed according to the compensation value = 0.7 × deviation value. The correction instruction is then sent to BuildDrone to automatically adjust the repair trajectory of subsequent layers. Conversely, if the repair deviation is less than the preset threshold, BuildDrone does not need to correct the path and continues the repair task.

[0045] S5. After the repair is completed, perform a full-domain scan, conduct repair acceptance, store the acceptance report, and complete the repair operation.

[0046] In this embodiment, after all repairs are completed, the present invention calls ScanDrone to perform a final full-domain scan to detect defect fill rate, surface flatness, and contour dimensions, ensuring no missed spraying, no overspraying, no voids, and no cracks; and collects dam geometry, UAV pose, material state, environmental parameters, and repair process data; then, based on the detection results and the collected data, an acceptance report containing point cloud images, photos, process parameters, and performance indicators is generated, and the hash value of the acceptance report is stored on the blockchain for evidence, realizing data immutability and remote acceptance, thus completing the entire repair process.

[0047] In another embodiment of the present invention, taking a 200-meter-class concrete gravity dam as an example, a large area of ​​freeze-thaw erosion occurred in the non-overflow section on the right bank within an elevation range of 180-220 meters, with a maximum erosion depth of 8 centimeters. The "Air-Water Integrated 3D Scanning and Multi-Machine Collaborative In-situ 3D Printing Repair Method for Dam Defects" of the present invention was used for unmanned repair. The steps include… A1 and ScanDrone took off vertically from the dam top platform. They first used a 32-line lidar to perform a zigzag scan of the dam surface at a speed of 8 m / s, acquiring high-density point cloud data (density 600 pts / m²) within 5 minutes. Then, they switched to underwater sonar mode to perform supplementary scanning of the water level fluctuation area to ensure the integrity of defect information. Meanwhile, the edge GPU server processed the point cloud in real time and identified freeze-thaw erosion areas through the PointRCNN++ network, automatically marking areas with a depth ≥5 mm as "areas requiring repair".

[0048] A2. Using the MCFS algorithm, the 350m² area is divided into 14 contour lines (0.8m wide), generating a collision-free initial trajectory; using... The algorithm is finely optimized in 5D space to avoid the operating radius of the gantry crane on the dam top and plans 6 optimal printing paths; based on the single BuildDrone's endurance (25 minutes) and load (10kg), the total task is divided into 28 4m³ task blocks.

[0049] Six BuildDrone drones took off sequentially from the dam crest landing area, achieving centimeter-level positioning through UWB-RTK fusion positioning. A ground-based mobile silo continuously pumped fast-hardening sulfoaluminate cement mortar (2h compressive strength 20MPa) and supplied it to the drones through a Φ25mm flexible hose. The drones printed layer by layer at a thickness of 4mm, with a nozzle speed of 0.3m / s. The system adjusted the extrusion rate in real time according to the curvature of the dam surface to maintain a layer width / height of 4±0.2mm. When the battery of drone #3 dropped to 30%, task reassignment was automatically triggered, and the remaining tasks were taken over by the adjacent drone #5. The handover process took only 8 seconds, and printing was uninterrupted.

[0050] A4. After every two layers (8mm thickness), ScanDrone performs a differential scan; if a local area printing deviation of 2.5mm is found, the path of the subsequent three layers is immediately corrected online with a compensation of 1.8mm; the final surface flatness test shows a maximum deviation of 2.8mm, which meets the accuracy requirement of <3mm.

[0051] A5. Generate a complete digital twin report, including 3.5 million thickness data points, temperature curves, and wind records; and use blockchain to store the evidence, forming an unalterable acceptance file.

[0052] In another embodiment of the present invention, the present invention also proposes an in-situ 3D printing repair system for dam defects using integrated air-water 3D scanning and multi-machine collaboration, which includes: a scanning module, a task generation module, a repair module, a path correction module, and an acceptance module.

[0053] 1. The scanning module is used to perform a full-area scan of the dam to obtain point cloud images of the area to be repaired; and to accurately register the point cloud images of the area to be repaired with the original BIM design model of the dam, extract key parameters of the dam, and generate a standardized defect repair list.

[0054] In this embodiment, the scanning module utilizes a ScanDrone integrated air-water scanning drone to perform a full-area scan of the dam. The steps include: a. Place ScanDrone on the dam crest safety take-off and landing platform, complete the power-on self-test, confirm that the lidar, multibeam sonar, underwater camera, positioning module, and communication module are working properly, and establish a stable communication link with the edge GPU server; b. ScanDrone performs a zigzag full-area scan along the dam face to acquire the original point cloud image of the dam; among them, the dam facade area uses a 32-line lidar to collect point clouds, while the water level fluctuation area and underwater area are switched to 130kHz multibeam sonar and pressure-resistant underwater lens operation, acquiring no less than 20 million point cloud data within 5 minutes, with a point cloud density of no less than 500pts / m².

[0055] In this embodiment, the scanning module also uses an edge GPU server to perform noise reduction, filtering, and downsampling preprocessing on the original point cloud image of the dam, and completes semantic segmentation through the PointRCNN++ deep learning network. Based on preset thresholds (such as depth ≥5mm, area ≥0.02m²), it automatically identifies defects such as freeze-thaw erosion, scour pits, cracks, and cavitation pits, and generates point cloud images of the area to be repaired.

[0056] In this embodiment, the scanning module adopts the ICP+CPD hybrid registration algorithm to accurately register the point cloud image of the area to be repaired with the original BIM design model of the dam, control the registration error (e.g., error ≤ 3mm), extract key parameters such as the three-dimensional coordinates, normal vector, principal curvature, area, and volume of the dam defects, and generate a standardized defect repair list.

[0057] 2. The task generation module is used to perform second-order intelligent path planning based on the point cloud image of the area to be repaired, and to standardize the defect repair list and the performance parameters of the printing drone (BuildDrone) to allocate tasks to multiple drones and generate task slices.

[0058] In this embodiment, the task generation module utilizes the Connected Fermat Spirals algorithm (MCFS) and The safe field algorithm for second-order intelligent path planning includes the following steps: a. Use the MCFS algorithm to perform coarse-level planning, divide the point cloud image of the area to be repaired into several wide rings (e.g., 0.8m) according to the contour lines of the dam surface, and generate an initial trajectory that is fully covered, non-intersecting, and without omissions; b. Based on The safety field algorithm performs fine-grained optimization in the dam's 5D state space (e.g., x, y, z, yaw ψ, pitch θ), through... The protocol reads the real-time trajectory of the gantry crane and cable crane PLC on the dam top, transforms the dynamic sling into a virtual cylindrical obstacle, realizes real-time dynamic obstacle avoidance, and outputs the optimal and refined waypoints.

[0059] In this embodiment, the task generation module allocates tasks to multiple BuildDrone instances based on a standardized defect repair list and BuildDrone performance parameters, with each instance having a certain single-block volume, thus generating task slices. Within a single task slice, tasks are allocated to 4–8 BuildDrone instances according to the principles of balanced load and proximity of operation.

[0060] Optionally, the task generation module also establishes an online task monitoring mechanism. When any BuildDrone's battery level is less than 30% or the remaining material weight is less than 2kg, it triggers a relay redistribution of neighboring machines, transferring unfinished repair tasks according to the nearest neighbor principle to ensure continuous and uninterrupted repair.

[0061] 3. The repair module is used to call BuildDrone and mobile silos to repair the dam area to be repaired according to the task slice; in a single task slice, the present invention is allocated to 4-8 BuildDrone according to the principle of balanced load and operation nearby.

[0062] In this embodiment, after the BuildDrone takes off sequentially to the area to be repaired, the repair module uses UWB RTK fusion positioning to locate the task area at the centimeter level, and hovers stably after arriving at the task area to begin the repair process.

[0063] Optionally, BuildDrone adaptively adjusts the nozzle extrusion rate based on the real-time principal curvature K of the dam surface to ensure that the layer width / layer height remains stable at 4±0.2mm; wherein the adaptive adjustment formula is expressed as follows: Q=K·v·A Where v is the nozzle lateral movement speed, Q is the extrusion rate, and A is the cross-sectional area.

[0064] Optionally, after each layer is repaired, BuildDrone's micro-pump reverses to achieve material recovery during downtime, preventing material dripping, and the moving hopper continuously pumps during the repair process to achieve high-altitude backpressure-free and uninterrupted material supply, eliminating cold seams between layers.

[0065] In this embodiment, for underwater defect repair, the repair module uses a moving hopper to switch to an underwater-specific formula and adopts a variable layer thickness strategy and dual-machine alternating operation to overcome water flow disturbance and ensure underwater molding quality.

[0066] 4. The path correction module is used to perform repair detection and repair path correction in a closed loop at a certain period during the repair process.

[0067] In this embodiment, when the path correction module has continuously repaired at least two layers of areas to be repaired, it stops the local repair operation and starts ScanDrone to quickly scan the repaired area, obtain the repair point cloud image, and compares the repair point cloud image with the original BIM design model of the dam to obtain the repair deviation value of the whole area and form a deviation distribution cloud map.

[0068] In this embodiment, the path correction module determines the deviation level based on the repair deviation value and the deviation distribution cloud map. If the repair deviation is greater than a preset threshold (e.g., 2mm), a path correction instruction is immediately generated, and a lateral offset correction is performed according to the compensation value = 0.7 × deviation value. The correction instruction is then sent to BuildDrone to automatically adjust the printing trajectory of subsequent layers. Conversely, if the repair deviation is less than the preset threshold, BuildDrone does not need to correct the path and continues the repair task.

[0069] 5. The acceptance module is used to perform a full-domain scan after the repair is completed, to conduct repair acceptance, and to store the acceptance report, thus completing the repair operation.

[0070] In this embodiment, after all repairs are completed, the acceptance module calls ScanDrone to perform a final full-domain scan to detect defect fill rate, surface flatness, and contour dimensions, ensuring no missed spraying, no overspraying, no voids, and no cracks; and collects data on dam geometry, UAV pose, material condition, environmental parameters, and repair process; then, based on the detection results and the collected data, it generates an acceptance report containing point cloud images, photos, process parameters, and performance indicators, and stores the hash value of the acceptance report on the blockchain for evidence, achieving data immutability and remote acceptance, thus completing the entire repair process.

[0071] The foregoing description and accompanying drawings fully illustrate embodiments of the invention to enable those skilled in the art to practice them. Other embodiments may include structural and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Some portions and features of some embodiments may be included or substituted for portions and features of other embodiments. Embodiments of the invention are not limited to the structures described above and shown in the accompanying drawings, and various modifications and changes may be made without departing from their scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for in-situ 3D printing repair of dam defects, characterized in that, The method includes, A full-area scan of the dam was performed to obtain point cloud images of the area to be repaired; the point cloud images of the area to be repaired were then precisely registered with the original BIM design model of the dam to extract key parameters of the dam and generate a standardized list of defects to be repaired. Based on the point cloud image of the area to be repaired, second-order intelligent path planning is performed, and multi-machine task allocation is carried out based on the standardized defect repair list and the performance parameters of the printing drone, generating task slices. Based on the task slices, printing drones and mobile material silos were deployed to repair the areas of the dam awaiting repair. During the repair process, repair detection and repair path correction are carried out in a closed loop iterative manner at a certain period. After the repair is completed, a full-domain scan is performed to conduct a repair acceptance test, and an acceptance report is stored to complete the repair operation.

2. The in-situ 3D printing repair method for dam defects according to claim 1, characterized in that, The full-area scanning of the dam specifically includes using an air-water integrated scanning drone to scan the dam in its entirety and obtain the original point cloud image of the dam. The acquisition of point cloud images of the area to be repaired specifically includes: using an edge GPU server to perform noise reduction, filtering, and downsampling preprocessing on the original point cloud image of the dam; performing semantic segmentation through a PointRCNN++ deep learning network; and automatically identifying defects such as freeze-thaw erosion, scour pits, cracks, and cavitation pits according to a preset threshold to generate point cloud images of the area to be repaired.

3. The in-situ 3D printing repair method for dam defects according to claim 1, characterized in that, The process of generating a standardized defect repair list specifically includes using an ICP+CPD hybrid registration algorithm to accurately register the point cloud image of the area to be repaired with the original BIM design model of the dam, controlling the registration error, extracting key parameters of the dam defects, and generating a standardized defect repair list.

4. The in-situ 3D printing repair method for dam defects according to claim 1, characterized in that, The second-order intelligent path planning specifically includes utilizing the connected Fermat spiral algorithm and RRT. The SF safe field algorithm is used for second-order intelligent path planning.

5. The in-situ 3D printing repair method for dam defects according to claim 1, characterized in that, The continuous closed-loop iteration performs repair detection and repair path correction, specifically including: When repairing at least two layers to be repaired, stop the local repair work and start an air-water integrated scanning drone to quickly scan the repaired area and obtain point cloud images of the repaired area; The repaired point cloud image is compared with the original BIM design model of the dam to obtain the repair deviation value of the whole area and form a deviation distribution cloud map. Based on the repair deviation value and the deviation distribution cloud map, the deviation level is determined; if the repair deviation is greater than the preset threshold, a path correction instruction is immediately generated and sent to the printing drone to automatically adjust the repair trajectory of subsequent layers. Conversely, if the repair deviation is less than the preset threshold, the printing drone does not need to correct the path and can continue the repair task.

6. The in-situ 3D printing repair method for dam defects according to claim 1, characterized in that, After the repair is completed, a full-domain scan is performed to conduct a repair acceptance test, and an acceptance report is stored to complete the repair operation. Specifically, this includes... After all repairs are completed, an air-water integrated scanning drone is used to perform a final full-area scan to detect the defect filling rate, surface flatness, and contour dimensions, ensuring that there are no missed sprays, no oversprays, no voids, and no cracks. It also collects data on dam geometry, UAV pose, material condition, environmental parameters, and repair process. Based on the test results and collected data, an acceptance report is generated that includes point cloud images, photographs, process parameters, and performance indicators. The hash value of the acceptance report is stored on the blockchain to ensure data immutability and remote acceptance, thus completing the entire process of repair.

7. A 3D printing repair system for dam defects in situ, characterized in that, The system includes: a scanning module, a task generation module, a repair module, a path correction module, and an acceptance module; The scanning module is used to perform a full-area scan of the dam to obtain point cloud images of the area to be repaired; and to accurately register the point cloud images of the area to be repaired with the original BIM design model of the dam, extract key parameters of the dam, and generate a standardized defect repair list. The task generation module is used to perform second-order intelligent path planning based on the point cloud image of the area to be repaired, and to allocate tasks to multiple drones based on the standardized defect repair list and the performance parameters of the printing drone, thereby generating task slices. The repair module is used to call upon the printing drone and mobile silo to repair the area of ​​the dam to be repaired, based on the task slice; The path correction module is used to perform repair detection and repair path correction in a closed loop iterative manner at a certain period during the repair process; The acceptance module is used to perform a full-domain scan after the repair is completed, to conduct repair acceptance, and to store the acceptance report, thus completing the repair operation.

8. The in-situ 3D printing repair system for dam defects according to claim 7, characterized in that, The scanning module is specifically used to perform a full-area scan of the dam using an air-water integrated scanning drone to obtain the original point cloud image of the dam. The original point cloud image of the dam is preprocessed by denoising, filtering, and downsampling using an edge GPU server, and semantic segmentation is performed by a PointRCNN++ deep learning network. Defects such as freeze-thaw erosion, scour pits, cracks, and cavitation pits are automatically identified according to preset thresholds to generate point cloud images of the area to be repaired.

9. The in-situ 3D printing repair system for dam defects according to claim 7, characterized in that, The scanning module is specifically used to accurately register the point cloud image of the area to be repaired with the original BIM design model of the dam using the ICP+CPD hybrid registration algorithm, control the registration error, extract key parameters of dam defects, and generate a standardized defect repair list.

10. The in-situ 3D printing repair system for dam defects according to claim 7, characterized in that, The task generation module is specifically used to utilize the connected Fermat spiral algorithm and RRT. The SF safe field algorithm is used for second-order intelligent path planning.

11. The in-situ 3D printing repair system for dam defects according to claim 7, characterized in that, The path correction module is specifically used for, When repairing at least two layers to be repaired, stop the local repair work and start an air-water integrated scanning drone to quickly scan the repaired area and obtain point cloud images of the repaired area; The repaired point cloud image is compared with the original BIM design model of the dam to obtain the repair deviation value of the whole area and form a deviation distribution cloud map. Based on the repair deviation value and the deviation distribution cloud map, the deviation level is determined; if the repair deviation is greater than the preset threshold, a path correction instruction is immediately generated and sent to the printing drone to automatically adjust the repair trajectory of subsequent layers. Conversely, if the repair deviation is less than the preset threshold, the printing drone does not need to correct the path and can continue the repair task.

12. The in-situ 3D printing repair system for dam defects according to claim 7, characterized in that, The acceptance module is specifically used for, After all repairs are completed, an air-water integrated scanning drone is used to perform a final full-area scan to detect the defect filling rate, surface flatness, and contour dimensions, ensuring that there are no missed sprays, no oversprays, no voids, and no cracks. It also collects data on dam geometry, UAV pose, material condition, environmental parameters, and repair process. Based on the test results and collected data, an acceptance report is generated that includes point cloud images, photographs, process parameters, and performance indicators. The hash value of the acceptance report is stored on the blockchain to ensure data immutability and remote acceptance, thus completing the entire process of repair.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the in-situ 3D printing repair method for dam defects according to any one of claims 1-6.

14. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes the program stored in the memory, it implements the steps of the in-situ 3D printing repair method for dam defects as described in any one of claims 1-6.

Images