Bridge coating intelligent maintenance method and system based on unmanned aerial vehicle

Abstract

The invention discloses a bridge coating intelligent maintenance method and system based on an unmanned aerial vehicle, and aims to improve the efficiency and precision of bridge coating disease detection and maintenance decision through the combination of an unmanned aerial vehicle technology and a large language model. The method comprises the following steps: automatically scanning the surface of a target bridge through an unmanned aerial vehicle carrying a high-resolution camera, and collecting high-definition image data of a bridge coating; processing the image data by using a deep learning model, and automatically identifying the disease type and the disease area of the bridge coating; and according to a disease identification result, obtaining related maintenance knowledge and meteorological data from a knowledge base through a coating maintenance large language model, and generating a personalized coating maintenance scheme in combination with reasoning analysis. According to the method, efficient and accurate detection of bridge coating diseases can be realized, a scientific and reasonable maintenance scheme is generated based on historical data and environmental factors, the bridge coating maintenance accuracy and maintenance efficiency can be effectively improved, the service life of a bridge is prolonged, and the method has a wide application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge monitoring and maintenance, and particularly relates to an intelligent maintenance method and system for bridge coatings based on unmanned aerial vehicles (UAVs). Background Art

[0002] The bridge anti-corrosion coating system is an important part of bridge maintenance and protection. Its main purpose is to isolate the bridge surface from the external environment through coating technology, prevent the erosion of corrosion factors (such as oxygen, water, salt spray, etc.) on the bridge structure, thereby extending the service life of the bridge, ensuring traffic safety, and enhancing the aesthetics of the bridge and the urban image. The bridge anti-corrosion coating system usually consists of a primer coat, an intermediate coat, and a top coat, and each coat has its specific functions and performance requirements. The primer coat is mainly used to enhance the adhesion of the coating and provide preliminary anti-corrosion protection; the intermediate coat can enhance the mechanical strength of the coating while providing additional anti-corrosion effects; the top coat plays the roles of waterproofing, anti-fouling, and aesthetics. The quality and construction technology of each coat directly affect the durability and safety of the bridge.

[0003] During the long-term use of bridges, affected by various factors such as climate change, environmental pollution, and traffic loads, the coating protection function gradually weakens, which may then lead to the corrosion and deterioration of the bridge structure, and even affect the safety of the bridge. In order to ensure the safe operation of the bridge, the coating protection of the bridge has become an important part of bridge maintenance.

[0004] Currently, the detection of bridge coatings mainly relies on manual inspections, traditional visual inspections, or the use of conventional detection equipment. These methods generally have some problems. First, manual inspections have certain limitations. Inspectors often have difficulty covering a large area of the bridge surface, and are restricted by factors such as weather and time. The inspection cycle is long and the efficiency is low. Second, it is difficult for traditional detection methods to achieve high-precision measurement of the disease area. The detection results are highly subjective and often rely on the experience and judgment of inspectors, which makes the diagnostic results prone to deviation. In addition, traditional detection methods cannot monitor the dynamic changes of bridge coatings in real time, and it is difficult to detect early diseases of coatings in a timely manner, resulting in the damage of coatings not being repaired in time, thus affecting the long-term safety of the bridge.

[0005] With the continuous development of UAV technology and image recognition technology, automated bridge coating detection based on UAVs has gradually become an emerging means to solve the above problems. With its advantages of high efficiency, precision, and flexibility, UAVs can quickly cover a large area of the bridge surface, collect high-definition images or videos, and cooperate with advanced image recognition and machine learning technologies to automatically identify and analyze the types and areas of diseases of bridge coatings. Through this automated means, the damage of bridge coatings can be discovered in the shortest time, thus providing timely data support for maintenance work.

[0006] However, despite the significant advantages of drone technology in bridge coating disease detection, its application still faces some challenges. Most existing technologies focus on image recognition and disease classification, but lack a system or method for generating intelligent and personalized maintenance plans for bridge coatings. Existing coating maintenance mostly relies on manual experience to formulate maintenance plans, and lacks an intelligent decision-making support system based on big data analysis and environmental factor reasoning. How to provide scientific, reasonable and personalized maintenance plans for different types of bridge coating diseases, different environmental conditions, and different coating materials is still a difficult problem that needs to be solved in the field of bridge maintenance. Summary of the invention

[0007] In order to solve the above technical problems, the present invention provides a method and system for intelligent maintenance of bridge coating based on drones.

[0008] The technical solution provided by the present invention is as follows:

[0009] A method for intelligent maintenance of bridge coating based on drones, comprising the following steps:

[0010] S1, collecting coating image data of the target bridge through a drone;

[0011] S2, input the coating image data into the detection model to obtain the coating disease type and calculate the disease area;

[0012] S3. Input the coating damage type and damage area into the coating maintenance LLM to obtain the corresponding maintenance plan;

[0013] Among them, the detection model is a target detection model trained using a bridge coating disease image dataset, which is used to detect the type of coating disease in the input image and output pixel-level mask information of the diseased area; the coating maintenance LLM is a bridge coating maintenance large language model obtained by fine-tuning the open source large language model, which is used to output the corresponding maintenance plan according to the input coating disease type and disease area; the maintenance plan includes coating material, thickness, construction method and subsequent maintenance cycle, and an explanation of the reasons for the maintenance plan.

[0014] Furthermore, the drone is equipped with a high-precision positioning module for recording the spatial coordinates of the shooting position, shooting angle and shooting distance in real time; the drone's route is planned based on the three-dimensional spatial coordinates of the target bridge to ensure that the spatial size of a single pixel of the collected coating image data on the bridge coating surface meets the set accuracy.

[0015] Further, the calculation method of the disease area is as follows: Using the mask information of the disease area output by the target detection model, according to the spatial coordinates of the shooting position, the shooting distance, the shooting angle, and the actual resolution parameters of the camera, through spatial geometric transformation and three-dimensional point cloud reconstruction, the pixel mask area of the disease area is converted into the actual disease area on the surface of the bridge.

[0016] Further, the construction process of the coating maintenance LLM includes the steps of:

[0017] Construct a knowledge base in the field of bridge coating maintenance,

[0018] Construct a knowledge graph in the field of bridge coating maintenance,

[0019] Based on the knowledge base and the knowledge graph, construct a corpus, and fine-tune the selected open-source large language model to ensure that the model can understand the semantics and knowledge relationship of the characteristics of bridge coating diseases and disease treatment methods, and obtain the coating maintenance LLM; the corpus form during fine-tuning includes "question-answer" dialogues and "disease description-solution output" dialogues to enhance the model's intelligent reasoning and decision-making generation ability in the field of bridge coating maintenance.

[0020] Preferably, the knowledge base includes:

[0021] Initial bridge coating information: Bridge design drawings, the material, thickness and construction method of the initial coating, painting construction standard specifications and technical process documents;

[0022] Bridge historical maintenance information: All records of the annual maintenance and repair of the target bridge and its similar bridges, including disease types, disease areas, the material, thickness and construction method of the coating, maintenance time, maintenance effect tracking data; Relationship data between the service life of the bridge and the durability of the coating;

[0023] Meteorological environment data: Meteorological environment monitoring data in the area where the target bridge is located, including air composition, air pH, air temperature and humidity, ultraviolet radiation intensity, and monitoring data of acid rain and salt fog;

[0024] Coating material performance data: Technical parameters of the relationship between the weather resistance, corrosion resistance, adhesion, coating thickness and service life of different coating materials, and long-term performance comparison data of different coating combinations.

[0025] Further, the construction steps of the knowledge graph include:

[0026] Determine the nodes of the knowledge graph, including disease types, environmental parameters, coating materials, and coating performance parameters;

[0027] Define node association relationships, including the association between disease types and causal factors, the association between diseases and the selection of corresponding maintenance coatings, the causal association between environmental factors and coating lifespan, and the association between coating combinations and disease repair effects;

[0028] Verification and update of the knowledge graph: Regularly evaluate and optimize the effectiveness of the knowledge graph, and continuously update the node relationship weights based on actual maintenance feedback data.

[0029] Furthermore, the operating mechanism of the coating maintenance LLM is as follows:

[0030] The coating maintenance LLM obtains the input disease type and disease area information;

[0031] The coating maintenance LLM automatically combines the knowledge graph and invokes the knowledge base data according to the input disease characteristics;

[0032] Based on the causal association relationships of the knowledge graph, the coating maintenance LLM first identifies the potential causes of diseases, infers the development trend of such diseases in the current environment based on historical data, and then conducts multi-dimensional reasoning based on existing maintenance data and coating material performance parameters to select appropriate coating materials, coating thicknesses, and construction methods;

[0033] The coating maintenance LLM generates and outputs a specific interpretive coating maintenance plan in natural language form.

[0034] Furthermore, the drone is also used to monitor the construction process in real time during the coating disease maintenance process, and collect image data containing the original disease area again after the maintenance is completed for input into the detection model to evaluate the maintenance effect.

[0035] A bridge coating intelligent maintenance system based on the above method includes the following modules:

[0036] Drone information collection module: Used to obtain the coating image data of the target bridge, as well as the spatial coordinates, shooting angles, and shooting distance data of the corresponding shooting positions;

[0037] Coating disease intelligent detection module: Includes a target detection unit and an area calculation unit; the target detection unit is used to detect the coating disease type in the input image and output the pixel-level mask information of the disease area; the area calculation unit is used to utilize the mask information of the disease area and convert the pixel mask area of the disease area into the actual area according to the spatial coordinates, shooting distance, shooting angle of the shooting position, and the actual resolution parameters of the camera;

[0038] Coating maintenance intelligent decision-making module: Deployed with a coating maintenance LLM, used to generate an interpretive coating maintenance plan in natural language form according to the input disease type and area information.

[0039] Furthermore, it also includes a UAV flight path planning module, which is used to plan the flight path of the UAV according to the three-dimensional spatial distribution of the target bridge and the set accuracy of the coating image data.

[0040] Compared with the prior art, the present invention has at least the following beneficial effects:

[0041] First of all, the UAV can efficiently cover a large area of the bridge surface, avoiding the time and space limitations of traditional manual inspections and significantly improving the detection efficiency. The high-resolution camera and infrared sensor equipped on the UAV can accurately obtain high-definition image data of the bridge coating, and through automated image processing and detection models, real-time identification and analysis of coating diseases can be carried out, thus overcoming the common problems of missed inspections and misjudgments in the process of manual detection. Different from the traditional method that relies on manual experience judgment, the intelligent maintenance solution based on the UAV can provide quantitative and accurate disease diagnosis results, further improving the objectivity and accuracy of the detection. In addition, by introducing big data analysis and environmental meteorological factor reasoning, the present invention can automatically generate personalized maintenance plans in combination with historical maintenance records, coating material characteristics, and regional climate environment changes. This not only ensures the scientificity and rationality of the maintenance plan, but also can predict the development trend of diseases, take measures in advance, avoid over-maintenance or delayed maintenance, reduce the maintenance cost, and extend the service life of the bridge.

[0042] Generally speaking, the present invention greatly improves the intelligent, precise and efficient level of bridge coating maintenance and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings are used to provide a further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation to the present invention.

[0044] Figure 1 It is a schematic diagram of the technical framework for intelligent maintenance of bridge coatings provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, rather than all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] Embodiment 1

[0047] As Figure 1As shown in the figure, this embodiment provides an intelligent maintenance method for bridge coatings based on drones to achieve automatic detection, diagnosis, and intelligent maintenance decision-making of the state of bridge anti-corrosion coatings, improving the efficiency and effectiveness of bridge maintenance.

[0048] This method includes the following specific steps:

[0049] Step S1: Acquisition of bridge coating image data

[0050] First, use a high-resolution visible light camera (such as an industrial digital camera with more than 50 million pixels) and an infrared camera (resolution ≥ 1024×768 pixels) carried by the drone to automatically scan the surface of the target bridge according to a preset flight path, and obtain high-precision image data of the target bridge coating.

[0051] The preset flight path is planned in advance based on the three-dimensional spatial coordinates of the bridge. The flight height of the drone is set to be 20 - 50 meters from the bridge surface, so that the spatial size of a single pixel on the bridge coating surface is controlled within 1 millimeter to ensure that disease details can be fully presented. The drone carries an RTK-GPS high-precision positioning module to record the spatial coordinates, shooting angles (pitch angle, tilt angle), and shooting distance data of the shooting position in real time, and transmits them together with the captured image data to the ground station in real time via wireless transmission for subsequent accurate calculation of the disease area.

[0052] Step S2: Intelligent detection of bridge coating diseases

[0053] The intelligent detection model for bridge coating diseases in this embodiment is an object detection model obtained through deep learning training, such as the YOLOv8 instance segmentation network or the Mask R-CNN network. The model uses a bridge coating disease image dataset (including different types of diseases such as rust, peeling, cracks, and bulges), and is obtained through transfer learning and model optimization and fine-tuning for the bridge disease scenarios in specific regions.

[0054] The specific detection process is as follows:

[0055] Input the high-precision image data of the bridge coating obtained in step S1 into the disease detection model, and the model automatically outputs the category information of the disease and the pixel-level mask information of the disease area;

[0056] Then, using the disease area mask information, according to the spatial coordinates, shooting distance, shooting angles, and actual resolution parameters of the drone camera, through spatial geometric transformation and three-dimensional point cloud reconstruction methods, accurately convert the pixel mask area of the disease area into the actual disease area on the bridge surface.

[0057] Among them, for the three-dimensional point cloud reconstruction, the SfM (Structure from Motion) algorithm is specifically adopted. According to the multi-angle shooting data, the accurate mapping of the disease area in the actual coordinate system of the bridge is realized, so as to improve the measurement accuracy of the disease area, and the accuracy error is controlled within ±3%.

[0058] Step S3, Intelligent maintenance decision-making for bridge coatings

[0059] In order to achieve the accurate diagnosis and intelligent maintenance decision-making of bridge coating diseases, in this embodiment, a special intelligent maintenance large language model for the field of bridge coating maintenance (Bridge Coating Maintenance LLM) is constructed. The construction of this model includes the following specific steps:

[0060] (1) Select the basic model

[0061] First of all, in this embodiment, an open-source Transformer large language model is selected as the basic model (such as LLaMA, DeepSeek, ChatGLM or GPT family models), and its general natural language understanding and reasoning capabilities are obtained to ensure the efficiency and intelligence of the subsequent decision-making process.

[0062] (2) Construct a knowledge base for the field of bridge coating maintenance

[0063] In the process of constructing a large language model for a specific domain in this embodiment, a dedicated knowledge base for the field of bridge coating maintenance is particularly built. The specific content of the knowledge base includes but is not limited to:

[0064] (1) Initial coating information of the bridge

[0065] Bridge design drawing information;

[0066] Initial anti-corrosion coating system (including coating material, thickness, construction method, etc.);

[0067] Coating construction standard specifications and technical process documents.

[0068] (2) Historical maintenance information of the bridge

[0069] Store all the records of the annual maintenance and repair of the target bridge and other similar bridges, including disease types, disease areas, coating schemes (materials, thicknesses, construction methods), repair times, and repair effect tracking data;

[0070] Relationship data between the service life of the bridge and the durability of the coating.

[0071] (3) Meteorological environment data

[0072] Meteorological environment monitoring data of the area where the target bridge is located, including air components (such as SO2, NO x 、Cl -Monitoring data on corrosive components (such as), air acidity and alkalinity, air temperature and humidity, ultraviolet radiation intensity, and the impact of acid rain or salt spray on the coating, etc.

[0073] (4) Coating material performance data

[0074] Technical parameters on the relationship between the weather resistance, corrosion resistance, adhesion, coating thickness and service life of different coating materials (such as polyurethane, fluorocarbon paint, epoxy resin, acrylic polyurethane, etc.);

[0075] Long-term performance comparison data of different coating combinations (such as primer - intermediate coat - top coat).

[0076] Through the expansion of the above knowledge base data, the knowledge base of the model is greatly enriched, enabling the model to have the ability to make accurate maintenance decisions for diverse environmental conditions and bridge characteristics.

[0077] (III) Constructing a knowledge graph in the field of bridge coating maintenance

[0078] In order to efficiently utilize the reasoning and decision-making capabilities of the bridge coating maintenance LLM, this embodiment establishes a dedicated knowledge graph for bridge coating maintenance. The specific steps include:

[0079] (1) Definition of domain knowledge ontology

[0080] Clarify the domain knowledge ontology of the bridge coating field. The nodes include "coating material", "disease type", "corrosion factor", "maintenance process", "environmental parameter", "coating performance parameter", "repair plan", etc., and define the attributes and relationships of each node.

[0081] (2) Instance construction of the knowledge graph

[0082] Normalize the knowledge base data (historical maintenance data, regional environmental data, coating technical parameters) to construct instance nodes in the knowledge graph, such as the relationship between a specific maintenance activity, a specific environmental condition and the coating combination.

[0083] (3) Definition of node association relationships

[0084] Define the semantic attributes of the relationships between nodes, including:

[0085] "Disease type - cause factor" association (such as the relationship between rust and acid rain, salt spray);

[0086] Association between diseases and the selection of the best repair coating;

[0087] Causal association between environmental factors and coating life;

[0088] Association between coating combinations and disease repair effects.

[0089] (4) Knowledge Graph Verification and Dynamic Update Mechanism

[0090] Regularly evaluate and optimize the effectiveness of the knowledge graph, continuously update the node relationship weights according to the subsequent actual maintenance feedback data, and gradually improve the accuracy of reasoning.

[0091] (4) Fine-tuning and training of large language models

[0092] Construct a corpus in the field of bridge coating maintenance, and transform the aforementioned knowledge base and knowledge graph into text forms that can be understood by large language models; fine-tune and train the selected basic model according to the corpus in the field of bridge coating maintenance to ensure that the model can understand the semantics and knowledge relationships of bridge coating disease characteristics and disease treatment methods, and obtain a large language model for bridge coating maintenance (LLM); the corpus forms during fine-tuning include "question-answer" dialogues, "disease description - solution output" dialogues, etc., to enhance the model's intelligent reasoning and decision-making generation capabilities in the field of bridge coating maintenance.

[0093] After fine-tuning, use Prompt engineering technology to automatically convert bridge coating disease information into model input prompts (Prompts), and combine the knowledge graph to call the knowledge base data. When the model generates a maintenance plan, it dynamically and real-time reasons and outputs decision-making suggestions.

[0094] Specifically, the operation mechanism of the large language model for bridge coating maintenance can be summarized as follows:

[0095] 1. Automatically identify corresponding nodes for input disease data

[0096] Encode the disease type, disease area, and spatial location information obtained in step S2 into structured Prompts and input them into the large language model for bridge coating maintenance (LLM).

[0097] 2. Automatically retrieve historical information and environmental data from the knowledge base

[0098] Based on the input disease characteristics, the large language model for bridge coating maintenance (LLM) automatically combines the knowledge graph to call knowledge base data such as historical maintenance records, initial bridge coating plans, and regional climate environment parameters, and realizes efficient data calling through a retrieval and matching algorithm.

[0099] 3. Reasoning and analysis based on the knowledge graph

[0100] The large language model for bridge coating maintenance (LLM) is based on the causal association relationship of the knowledge graph:

[0101] First, identify the potential causes of diseases (such as rust → humidity, acid rain, salt spray environment, etc.);

[0102] According to historical data, infer the development trend of such diseases in the current environment (such as the crack or rust expansion speed);

[0103] Based on existing maintenance data and coating material performance parameters, conduct multi-dimensional reasoning to select suitable coating materials, coating thicknesses, and painting schemes.

[0104] 4. Multi-round Interactive Optimization Decision-making Based on Enhanced Prompt Engineering

[0105] Using the Chain-of-Thought Prompting technology, the bridge coating maintenance LLM can self-feedback and optimize the maintenance plan. Through multiple logical inferences on the relationship between the corrosion resistance of the coating, the construction thickness, and the actual environment of the bridge, the scientificity and practicality of the output plan can be further improved.

[0106] 5. Intelligent Maintenance Plan Output

[0107] Finally, the bridge coating maintenance LLM generates a specific and explanatory coating maintenance plan, including:

[0108] Recommended coating combinations (primer, intermediate coat, topcoat) and specific materials;

[0109] The recommended construction thickness for each coating (e.g., primer coat: zinc-rich epoxy primer, thickness 80μm; intermediate coat: epoxy mica iron paint, thickness 100μm; topcoat: fluorocarbon paint, thickness 50μm);

[0110] Construction methods (e.g., sandblasting grade for surface pretreatment, construction temperature and humidity requirements, drying and curing time);

[0111] Subsequent maintenance cycles (e.g., regular maintenance inspections every spring).

[0112] The output results of the bridge coating maintenance LLM are also accompanied by "explanations", clearly stating the reasons for recommending such coating schemes, such as "It is recommended to use 100μm zinc-rich epoxy primer + 150μm epoxy mica iron intermediate coat + 50μm fluorocarbon topcoat. Reason: The air acidity and alkalinity in this area is pH = 4.5, and the humidity is relatively high. Using zinc-rich epoxy primer and fluorocarbon topcoat can effectively inhibit the erosion of the acidic environment and humid air on the bridge surface", etc.

[0113] The above operating mechanism ensures that the intelligent maintenance decision-making has high accuracy, high applicability, and the ability to effectively predict future disease trends, greatly improving the scientificity, rationality, and practicality of bridge maintenance, and achieving a breakthrough innovation in active and intelligent bridge coating maintenance.

[0114] The above is the main content of the intelligent maintenance method for bridge coatings provided in this embodiment. By adopting a high-precision disease identification and detection model, this method accurately quantifies the disease area, avoiding the uncertainty of manual evaluation. The construction of the large language model and the application of the knowledge graph enable the intelligent and customized maintenance decision-making. At the same time, the intelligent data collection is realized by using the high-efficiency scanning of drones to replace manual detection. In some embodiments, drones are also used to monitor the construction process in real time during the maintenance of coating diseases, and to collect image data including the original disease area again after the maintenance is completed, for inputting into the detection model to evaluate the maintenance effect.

[0115] Embodiment 2

[0116] Based on the above method, this embodiment provides an intelligent maintenance system for bridge coatings based on drones. The system mainly includes the following modules:

[0117] UAV information collection module: used to obtain the coating image data of the target bridge, as well as the spatial coordinates, shooting angles, and shooting distance data of the corresponding shooting positions;

[0118] Coating disease intelligent detection module: includes a target detection unit and an area calculation unit; the target detection unit is used to detect the coating disease types in the input image and output the pixel-level mask information of the disease area; the area calculation unit is used to utilize the mask information of the disease area and convert the pixel mask area of the disease area into the actual area according to the spatial coordinates, shooting distance, shooting angle of the shooting position, and the actual resolution parameters of the camera.

[0119] Coating maintenance intelligent decision-making module: deployed with a coating maintenance LLM, used to generate an explanatory coating maintenance plan in natural language according to the input disease type and area information.

[0120] In some embodiments, the above system further includes a UAV flight path planning module, used to plan the flight path of the UAV according to the three-dimensional spatial distribution of the target bridge and the set accuracy of the coating image data.

[0121] The above system can execute the intelligent maintenance method for bridge coatings based on drones described in Embodiment 1, and has the corresponding functional modules and beneficial effects of the method. For the technical details not described in detail in this embodiment, reference can be made to the intelligent maintenance method for bridge coatings based on drones provided in Embodiment 1 of the present invention.

[0122] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; under the idea of the present invention, the technical features in the above embodiments or different embodiments can also be combined, and the steps can be implemented in any order, and there are many other variations in different aspects of the present invention as described above. For the sake of brevity, they are not provided in detail; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the various embodiments of the present application.

Claims

1. An intelligent maintenance method for bridge coatings based on drones, characterized in that, Including the steps: S1. Collect the coating image data of the target bridge by using a drone; S2. Input the coating image data into the detection model to obtain the coating disease types and calculate the disease areas; S3. Input the coating disease types and disease areas into the coating maintenance LLM to obtain the corresponding maintenance plans; Among them, the detection model is a target detection model trained by using a bridge coating disease image dataset, which is used to detect the coating disease types in the input image and output the pixel-level mask information of the disease areas; The coating maintenance LLM is a large language model for bridge coating maintenance fine-tuned based on an open-source large language model, which is used to output the corresponding maintenance plans according to the input coating disease types and disease areas; the maintenance plans include coating materials, thickness, construction methods and subsequent maintenance cycles, as well as the reason explanations for giving the maintenance plans.

2. The intelligent maintenance method for bridge coatings according to claim 1, characterized in that, The drone is equipped with a high-precision positioning module, which is used to record the spatial coordinates, shooting angles and shooting distances of the shooting positions in real time; the flight path of the drone is planned based on the three-dimensional spatial coordinates of the target bridge to ensure that the spatial size of a single pixel of the collected coating image data on the bridge coating surface meets the set accuracy.

3. The intelligent maintenance method of the bridge coating according to claim 2, characterized in that, The calculation method of the disease area is as follows: using the mask information of the disease area output by the target detection model, according to the spatial coordinates, shooting distance, shooting angle of the shooting position and the actual resolution parameters of the camera, through spatial geometric transformation and three-dimensional point cloud reconstruction, convert the pixel mask area of the disease area into the disease area on the actual surface of the bridge.

4. The intelligent maintenance method of the bridge coating according to claim 1, wherein, The construction process of the coating maintenance LLM includes the steps: Construct a knowledge base for the field of bridge coating maintenance, Construct a knowledge graph for the field of bridge coating maintenance, Construct a corpus based on the knowledge base and the knowledge graph, and fine-tune the selected open-source large language model to ensure that the model can understand the semantics and knowledge relationships of the characteristics of bridge coating diseases and disease treatment methods, and obtain the coating maintenance LLM; the corpus forms during fine-tuning include "question-answer" type dialogues and "disease description - solution output" type dialogues to enhance the intelligent reasoning and decision-making generation ability of the model for the field of bridge coating maintenance.

5. The intelligent maintenance method for bridge coatings according to claim 4, characterized in that, The knowledge base includes: Initial coating information of the bridge: bridge design drawings, materials, thickness and construction methods of the initial coating, painting construction standard specifications and technical process documents; Historical maintenance information of the bridge: all records of the annual maintenance and repair of the target bridge and its similar bridges, including disease types, disease areas, materials, thickness and construction methods of the coating, maintenance time, maintenance effect tracking data; relationship data between the service life of the bridge and the durability of the coating; Meteorological environment data: meteorological environment monitoring data of the area where the target bridge is located, including air components, air acidity and alkalinity, air temperature and humidity, ultraviolet radiation intensity, as well as monitoring data of acid rain and salt fog; Coating material performance data: technical parameters of the relationship between the weather resistance, corrosion resistance, adhesion, coating thickness and service life of different coating materials, and long-term performance comparison data of different coating combinations.

6. The intelligent maintenance method for bridge coatings according to claim 4, wherein, The construction steps of the knowledge graph include: Determine the nodes of the knowledge graph, including disease types, environmental parameters, coating materials, and coating performance parameters; Define node association relationships, including the association between disease types and causal factors, the association between diseases and the selection of corresponding maintenance coatings, the causal association between environmental factors and coating lifespan, and the association between coating combinations and disease repair effects; Verification and update of the knowledge graph: Regularly evaluate and optimize the effectiveness of the knowledge graph, and continuously update the node relationship weights based on actual maintenance feedback data.

7. The intelligent maintenance method for bridge coatings according to claim 4, characterized in that: The operation mechanism of the coating maintenance LLM is as follows: The coating maintenance LLM obtains the input disease type and disease area information; The coating maintenance LLM automatically combines the knowledge graph and calls the knowledge base data according to the input disease characteristics; Based on the causal association relationships of the knowledge graph, the coating maintenance LLM first identifies the potential causes of diseases, infers the development trend of such diseases in the current environment according to historical data, and then conducts multi-dimensional reasoning based on existing maintenance data and coating material performance parameters to select appropriate coating materials, coating thicknesses, and construction methods; The coating maintenance LLM generates a specific explanatory coating maintenance plan in natural language form and outputs it.

8. The intelligent maintenance method of the bridge coating according to claim 1, characterized in that, The drone is also used to monitor the construction process in real time during the coating disease maintenance process, and collect image data including the original disease area again after the maintenance is completed, for input into the detection model to evaluate the maintenance effect.

9. A bridge coating intelligent maintenance system based on the method according to any one of claims 1 to 8, characterized in that, It includes the following modules: Drone information collection module: Used to obtain the coating image data of the target bridge, as well as the spatial coordinates, shooting angles, and shooting distance data of the corresponding shooting positions; Coating disease intelligent detection module: Includes an object detection unit and an area calculation unit; the object detection unit is used to detect the coating disease type in the input image and output the pixel-level mask information of the disease area; The area calculation unit is used to utilize the mask information of the disease area and convert the pixel mask area of the disease area into the actual area according to the spatial coordinates, shooting distance, shooting angle of the shooting position, and the actual resolution parameters of the camera; Coating maintenance intelligent decision-making module: Deployed with a coating maintenance LLM, used to generate an explanatory coating maintenance plan in natural language form according to the input disease type and area information.

10. The intelligent maintenance system for bridge coatings according to claim 9, wherein It also includes a drone flight route planning module, used to plan the flight route of the drone according to the three-dimensional spatial distribution of the target bridge and the set accuracy of the coating image data.

Images