Bridge management and maintenance decision-making system and method based on multi-agent collaborative optimization

Abstract

The invention discloses a bridge management and maintenance decision-making system and method based on multi-agent collaborative optimization, and the system comprises a monitoring agent which is deployed in a cloud server and is used for obtaining abnormal data in a bridge structure and environment data; the diagnosis agent is used for evaluating the health state of the bridge by utilizing a built-in knowledge base, a built-in machine learning model and a built-in physical model based on the abnormal data; the decision-making agent is used for generating a plurality of candidate maintenance schemes based on an evaluation result, and screening out an optimal scheme by integrating the comprehensive utility of each scheme and the resource matching degree score of the resource scheduling agent on each scheme; the resource scheduling agent generates a construction plan according to the optimal scheme; the coordination / communication agent is used for ensuring efficient cooperation among the agents through a communication protocol and a negotiation mechanism among the agents; the diagnosis report, the optimal scheme and the construction plan are integrated and presented to a bridge manager through a human-computer interaction interface; and collaborative optimization of bridge management and maintenance decisions is realized through continuous feedback and self-learning.

Description

Technical Field

[0001] This invention belongs to the field of bridge maintenance technology, and more specifically, relates to a bridge maintenance decision-making system and method based on multi-agent collaborative optimization. Background Technology

[0002] In modern transportation networks, bridges play a crucial role, acting as key hubs to maintain smooth and stable traffic flow. The safety and durability of bridges directly impact the safety of life and property for passing vehicles and pedestrians, as well as the efficiency and sustainability of traffic operations. However, current bridge maintenance decision-making models largely rely on periodic inspections. These inspections may be conducted manually on-site or utilize automated equipment for data collection. These data are then combined with subjective judgments from experts based on years of accumulated experience to formulate corresponding maintenance strategies. This traditional decision-making model has many shortcomings that urgently need to be addressed, severely restricting the scientific, efficient, and accurate nature of bridge maintenance work.

[0003] First, there are significant shortcomings in terms of collaboration. Bridge maintenance encompasses multiple key stages, including monitoring, diagnosis, decision-making, and repair. In practice, these stages are often scattered across different departments or handled independently by different systems. Poor information flow between departments creates "data silos," hindering effective data sharing and collaborative operations. This results in an inability to consider the overall picture, making it difficult to formulate maintenance strategies that truly align with optimal global goals. The various stages hinder each other, preventing the formation of an efficient collaborative maintenance effort.

[0004] Secondly, real-time performance is severely lacking. Traditional maintenance decisions mostly follow fixed cycles, such as conducting comprehensive inspections and assessments at regular intervals before making decisions. This periodic decision-making model cannot respond promptly to sudden damage to bridge structures, such as immediate damage caused by natural disasters or accidental impacts, or to rapid deterioration processes, such as the rapid decline in material performance under harsh environmental conditions. It cannot provide real-time risk warnings, and it is even more difficult to dynamically adjust decision-making plans based on real-time changes, posing significant safety hazards. Missing the optimal treatment window may lead to further damage, increasing repair costs and difficulty.

[0005] Furthermore, the decision-making process is overly subjective. It relies excessively on the personal experience of experts, which, while valuable, has limitations. Disagreements may arise between different experts, and experience is difficult to quantify and standardize. The entire decision-making process lacks transparency, making it difficult to trace the basis of decisions. For complex maintenance issues involving multiple conflicting objectives, such as finding a balance between cost control, safety assurance, minimizing traffic impact, and extending service life, human decision-making often fails to accurately pinpoint the optimal solution, easily leading to decision biases and impacting maintenance effectiveness and resource utilization efficiency.

[0006] Finally, inefficiency is a particularly prominent issue. Data processing, model analysis, and scheme comparison all require significant manpower and time. This is especially true when managing large-scale bridge groups, which are numerous, structurally diverse, and generate massive amounts of data. Traditional methods struggle to handle these challenges efficiently, leading to delays in maintenance, an inability to respond promptly to the actual needs of the bridges, impacting the overall efficiency of the transportation network, and increasing management costs.

[0007] Given the numerous shortcomings and deficiencies of the traditional bridge maintenance decision-making methods, there is an urgent need for a new, intelligent bridge maintenance decision-making method that can integrate heterogeneous information sources, achieve multi-objective collaborative optimization, and possess adaptive and real-time response capabilities. Summary of the Invention

[0008] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a bridge maintenance decision-making method and system based on multi-agent collaborative optimization. Through clear division of labor and close collaboration among multiple agents, the system achieves full-process automation and intelligence from data collection, status monitoring, health diagnosis, decision-making to resource scheduling. This effectively breaks down information silos, promotes information sharing, and enables efficient data flow and collaborative processing across all stages, making maintenance decisions more scientific, accurate, and transparent. Simultaneously, the system possesses real-time response and dynamic adaptability, enabling real-time monitoring of bridge status, rapid response to sudden damage and deterioration of bridge structures, and providing accurate risk warnings and dynamic decision adjustments, significantly reducing safety risks and maintenance costs. Based on scientific models and algorithms, it reduces subjective interference from human experience, achieving transparency, quantification, and traceability in the decision-making process. Furthermore, the automated decision-making process significantly shortens the decision-making cycle, reduces the workload of management personnel, and greatly improves the efficiency of maintenance decisions, providing strong support for the long-term stable operation of bridges and the efficient operation of transportation networks. It is particularly suitable for the efficient management of large-scale bridge groups, and can comprehensively improve the scientific nature, accuracy and efficiency of bridge maintenance work, ensuring the safety and durability of bridges and the stable operation of transportation networks.

[0009] To achieve the above objectives, one aspect of the present invention provides a bridge maintenance decision-making system based on multi-agent collaborative optimization, including a monitoring agent, a diagnostic agent, a decision-making agent, a resource scheduling agent, and a coordination / communication agent deployed on a cloud server;

[0010] The monitoring agent is used to acquire structural and environmental data of the bridge, filter out abnormal data, and send it to the coordination / communication agent.

[0011] The diagnostic agent is used to receive abnormal data forwarded by the coordination / communication agent, and to assess the health status of the bridge using a built-in knowledge base, machine learning model and physical model, identify the type, location and severity of damage, and output a structural health index to generate a diagnostic report.

[0012] The decision-making agent is used to generate multiple feasible candidate maintenance schemes based on the diagnostic report output by the diagnostic agent, comprehensively consider the preset multi-objective maintenance utility function, calculate the comprehensive utility of each candidate maintenance scheme, and select the optimal maintenance scheme by combining the comprehensive utility of each candidate maintenance scheme and the resource matching degree score given by the resource scheduling agent for each candidate maintenance scheme.

[0013] The resource scheduling agent is used to follow the negotiation instructions issued by the coordination / communication agent, perform resource feasibility assessment on each candidate maintenance scheme, and give a corresponding resource matching score; based on the optimal maintenance scheme selected by the decision agent, it uses an optimization algorithm to generate a specific maintenance construction plan and submits the construction plan to the coordination / communication agent.

[0014] The coordination / communication agent ensures efficient collaboration among its various agents through communication protocols and negotiation mechanisms; it integrates the diagnostic report, the optimal maintenance plan, and the maintenance construction plan, presenting them to bridge managers through a human-computer interaction interface; and through continuous feedback and self-learning, it achieves collaborative optimization of bridge maintenance decisions.

[0015] Furthermore, the monitoring agent includes a data acquisition module, a data cleaning module, a data preprocessing module, and a preliminary anomaly detection module;

[0016] The data acquisition module is used to collect bridge-related data from multiple data sources, including: embedded sensors for real-time monitoring of bridge strain, displacement, temperature, and humidity; a drone inspection system that uses a drone-mounted high-definition camera, thermal imager, and lidar to inspect the bridge's appearance, thermal imaging, and three-dimensional morphology; and a manual reporting terminal for inspection personnel to record and report bridge defects on-site.

[0017] The data cleaning module is used to preprocess the collected raw data;

[0018] The data preprocessing module is used to further process the cleaned data, including data format unification, data normalization, and feature extraction.

[0019] The preliminary anomaly detection module is used to perform preliminary analysis on the preprocessed data, identify and mark abnormal data.

[0020] Furthermore, the diagnostic intelligent agent includes a data receiving module, a feature extraction module, a knowledge base module, a machine learning and physical model fusion module, and a health status assessment module;

[0021] The data receiving module is responsible for receiving preliminary abnormal data forwarded by the coordinating / communication agent;

[0022] The feature extraction module is used to extract features related to the bridge's health status from the preliminary anomaly data; the features related to the bridge's health status include stress concentration areas in strain data, deformation trends in displacement data, and crack length and width in image data;

[0023] The knowledge base module is used to store various knowledge related to bridge health diagnosis, including bridge design specifications, historical case data, material performance parameters, and damage modes.

[0024] The machine learning and physical model fusion module is used to comprehensively assess the health status of the bridge; the machine learning model includes a convolutional neural network (CNN) and a long short-term memory network (LSTM); the physical model includes a finite element model (FEM) to simulate the stress and deformation of the bridge under different working conditions based on the bridge's structural parameters and physical characteristics.

[0025] The health status assessment module is used to comprehensively assess the overall health status of the bridge based on the features extracted by the feature extraction module, the knowledge in the knowledge base module, and the analysis results of the machine learning and physical model fusion module; it integrates information on various damage types, locations, and severity to calculate the structural health index.

[0026] Furthermore, the decision-making intelligent agent includes a data receiving and parsing module and a decision-making module;

[0027] The data receiving and parsing module is responsible for receiving the health status information output by the diagnostic agent, including structural health index, damage type, location and severity data; and parsing this data to convert it into a format that the decision module can understand and process.

[0028] The decision-making module is used to comprehensively analyze and process the health status information parsed by the data receiving and parsing module, and finally generate a maintenance plan; the decision-making module includes a multi-objective maintenance utility function definition module, a plan generation module, a comprehensive utility evaluation module, and an optimal maintenance plan selection module;

[0029] The multi-objective maintenance utility function definition module is used to define and manage multi-objective maintenance utility functions. The multi-objective maintenance utility function includes multiple objectives, including cost minimization, safety maximization, traffic impact minimization, and service life maximization. Each objective has a corresponding weight, which is adjusted according to the specific conditions of the bridge and maintenance requirements.

[0030] The scheme generation module is used to generate multiple feasible candidate maintenance schemes based on the parsed health status information and the preset multi-objective maintenance utility function through an algorithm.

[0031] The comprehensive utility evaluation module performs a detailed evaluation of each candidate maintenance plan generated by the plan generation module, calculates the specific performance of each plan on each objective utility function, and obtains the comprehensive utility of each candidate maintenance plan.

[0032] The optimal maintenance scheme selection module integrates the overall utility of each candidate maintenance scheme and the resource matching score given by the resource scheduling agent for each candidate maintenance scheme. Through a multi-objective optimization algorithm, it finds the optimal balance point among multiple objectives and selects the optimal maintenance scheme.

[0033] Furthermore, the resource scheduling intelligent agent includes a resource demand analysis module, a resource inventory and availability management module, a resource matching degree scoring module, and a construction plan generation module;

[0034] The resource demand analysis module is used to perform resource demand analysis on the candidate maintenance plans generated by the decision-making module, and determine the human, material, financial and time resources required to implement the optimal maintenance plan;

[0035] The resource inventory and availability management module is used to manage the currently available resource inventory, including human, material and financial resources. By updating the availability status of resources in real time, it ensures that reasonable scheduling can be carried out according to the current resource situation.

[0036] The resource matching degree scoring module is used to follow the negotiation instructions issued by the coordination / communication agent, and based on the results of the resource demand analysis module and the real-time data of the resource inventory and availability module, to evaluate the resource feasibility of each candidate maintenance scheme and give a corresponding resource matching degree score.

[0037] The construction plan generation module generates a specific maintenance construction plan based on the optimal maintenance scheme selected by the decision-making agent, using optimization algorithms to coordinate and plan the manpower, materials, equipment, and time window resources required to execute the optimal maintenance scheme.

[0038] Furthermore, the coordination / communication agent includes a data receiving and forwarding module, a task distribution module, a conflict coordination module, a communication protocol and negotiation mechanism definition module between agents, an information integration and presentation module, and a feedback and self-learning module;

[0039] The data receiving and forwarding module is responsible for receiving data from the monitoring agent, the diagnostic agent, the decision-making agent, and the resource scheduling agent, and forwarding this data to the corresponding execution agents.

[0040] The task distribution module is used to generate specific tasks according to the overall needs of the system, assign the tasks to the corresponding intelligent agents, and ensure the execution of the tasks.

[0041] The conflict coordination module is used to resolve conflicts between intelligent agents and ensure the collaborative operation of the system; it includes conflict detection, which detects conflicts between intelligent agents in real time; conflict resolution, which resolves conflicts according to preset rules and strategies; and a negotiation mechanism, which initiates a negotiation process to allow relevant intelligent agents to negotiate.

[0042] The module defining communication protocols and negotiation mechanisms between agents is used to define the communication protocols and negotiation mechanisms between agents, ensuring that the information exchange and negotiation process between agents is efficient and orderly.

[0043] The information integration and presentation module is used to integrate the information provided by various intelligent agents and present it to the bridge manager through a human-computer interaction interface.

[0044] The feedback and self-learning module is used to reassess the structural health index of the bridge based on the new data collected by the monitoring agent. Using the new monitoring data and the reassessed structural health index, the diagnostic model of the diagnostic agent is updated and optimized online. The control decision agent updates and optimizes its decision strategy model online based on the new structural health index and the new diagnostic report, forming a "decision-execution-feedback" closed loop to achieve collaborative optimization of bridge maintenance decisions.

[0045] Furthermore, the structural health index is calculated using equations (1) to (3):

[0046]

[0047] Where i represents the damage type; N represents the total number of damages; C i The contribution of the i-th injury to overall health status; w iLet D be the weight of the i-th injury, based on the severity D of the i-th injury. i and positional importance L i Confirmed; w total D is the sum of all damage weights; i This represents the severity of the i-th damage (e.g., crack width, strain value, etc.); 1-D i To represent the health contribution of the injury, let's assume D. i The value is between [0,1], where 0 represents no damage and 1 represents complete damage;

[0048] Assume weight w i With positional importance L i Proportional; directly using positional importance as the weight, i.e., w i =L i Then the total weight w total This can be expressed by equation (4):

[0049]

[0050] Furthermore, the multi-objective maintenance utility function is expressed by equation (5):

[0051] max(U total )=w1·f s (S)+w2·f c (S)+w3·f t (S)+w4·f l (S) (5)

[0052] Among them, U total The comprehensive objective function value represents the overall performance of the maintenance scheme; max(U total ) represents the maximized overall utility; S represents a specific maintenance plan; f s (S) represents the expected safety utility function after the implementation of scheme S, which is positively correlated with the improvement of the structural health index (BHI); f c (S) is the cost-utility function of option S, usually negative, and negatively correlated with maintenance costs; f t (S) is the utility function of the traffic impact of plan S, which is usually negative and negatively correlated with the duration and extent of traffic disruption; f l (S) is the utility function of the impact of scheme S on the remaining life of the bridge; w1, w2, w3, and w4 are respectively the utility function of f s (S), f c (S), f t (S), f l (S) The weight coefficients of each utility function, and w1+w2+w3+w4=1.

[0053] A second aspect of the present invention provides a bridge maintenance decision-making method based on multi-agent cooperative optimization, which is implemented by applying the aforementioned bridge maintenance decision-making method based on multi-agent cooperative optimization, and includes the following steps:

[0054] S1: Data Acquisition and Preprocessing. The monitoring agent continuously collects structural and environmental data of the bridge, performs noise reduction and normalization on the collected data, marks abnormal data, and sends it to the coordination / communication agent.

[0055] S2: The coordination / communication agent forwards the abnormal data preprocessed in step S1 to the diagnostic agent. The diagnostic agent uses its built-in knowledge base, machine learning model and physical model to assess the health status of the bridge, identify the damage type, location and severity, output the structural health index, generate a diagnostic report, and send the diagnostic report to the coordination / communication agent.

[0056] S3: The coordination / communication agent distributes the diagnostic report to the decision agent, which generates one or more candidate maintenance scheme sets according to a preset multi-objective maintenance utility function and submits them to the coordination / communication agent;

[0057] S4: The coordination / communication agent distributes the candidate maintenance scheme set to the resource scheduling agent, and initiates a negotiation process between the resource scheduling agent and the decision-making agent according to a predetermined negotiation mechanism. The resource scheduling agent evaluates the resource feasibility of each candidate maintenance scheme according to the negotiation instructions issued by the coordination / communication agent, and gives a corresponding resource matching score. The decision-making agent calculates the comprehensive utility of each candidate maintenance scheme, and combines the comprehensive utility of each candidate maintenance scheme with the resource matching score, and uses an optimization algorithm to select the optimal maintenance scheme.

[0058] S5: The resource scheduling agent, based on the optimal maintenance scheme selected by the decision-making agent, uses optimization algorithms to coordinate and plan the human, material, equipment, and time window resources required to execute the optimal maintenance scheme, generates a specific maintenance construction plan, and submits the construction plan to the coordination / communication agent;

[0059] S6: The coordination / communication agent integrates the diagnostic report, the optimal maintenance plan, and the maintenance construction plan, and presents them to the bridge manager through a human-computer interaction interface; the bridge manager can view detailed information, adjust task priorities, and approve maintenance plans through the human-computer interaction interface.

[0060] S7: After maintenance is implemented, the coordination / communication agent reassesses the structural health index of the bridge based on the new data collected by the monitoring agent. Using the new monitoring data and the reassessed structural health index, the diagnostic model of the diagnostic agent is updated and optimized online. The control decision agent updates and optimizes its decision-making strategy model online based on the new structural health index and the new diagnostic report, forming a "decision-execution-feedback" closed loop. Through continuous feedback and self-learning, the collaborative optimization of bridge maintenance decisions is achieved.

[0061] A third aspect of the present invention provides an electronic device including a processor and a memory interconnected thereto;

[0062] The memory is used to store computer programs;

[0063] The processor is configured to execute the bridge maintenance decision-making method based on multi-agent cooperative optimization when the computer program is invoked.

[0064] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that is executed by a processor to implement the bridge maintenance decision-making method based on multi-agent cooperative optimization.

[0065] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0066] (1) Addressing the problem in traditional management and maintenance models where monitoring, diagnosis, decision-making, and maintenance are scattered across different departments or systems, making information sharing difficult and hindering the development of globally optimal strategies, this invention employs a multi-agent collaborative mechanism. Each agent has a clear division of labor yet collaborates closely. The monitoring agent collects data in real time, the diagnostic agent accurately assesses health status, the decision-making agent generates optimized solutions based on multiple objectives, the resource scheduling agent efficiently allocates resources, and the coordination / communication agent oversees the overall situation. This ensures smooth information flow and seamless integration across all stages, truly achieving a shift from local optimization to global optimization. This invention comprehensively optimizes decisions from multiple dimensions, including bridge safety, cost, traffic, and lifespan, balancing long-term stable operation with short-term economic benefits and the impact on traffic operations. This makes the management and maintenance strategy more scientific and rational, comprehensively ensuring the performance and benefits of the bridge.

[0067] (2) Addressing the problem that traditional maintenance decision-making cycles are fixed and difficult to respond to sudden damage or rapid deterioration, the system of this invention can operate 24 / 7 without interruption, receiving data from the monitoring agent in real time. Once an anomaly is detected, the diagnostic agent reacts quickly, the decision-making agent adjusts the plan promptly, and the resource scheduling agent quickly allocates resources, realizing the transformation from "periodic maintenance" to "condition-based maintenance" and "preventive maintenance," effectively reducing safety risks and repair costs caused by sudden situations. Furthermore, existing technologies often respond slowly to environmental and structural changes. This invention, through an online learning mechanism, enables the system to learn new data and structural characteristics under new environments in real time, automatically updating the diagnostic model and decision-making strategy to always adapt to the actual condition of the bridge, enhancing the system's stability and reliability, and ensuring the timeliness and accuracy of maintenance decisions.

[0068] (3) To address the problem that traditional decision-making relies excessively on expert experience, is highly subjective, and difficult to quantify, this invention, based on a quantifiable model and multi-objective optimization algorithm, transforms complex maintenance problems into quantifiable mathematical models. By solving for the optimal solution through scientific algorithms, it reduces the interference of human factors, making decisions more objective and accurate. This effectively addresses complex maintenance problems involving multiple conflicting objectives, finding the optimal balance between multiple requirements such as cost, safety, traffic, and lifespan. To address the problem of opaque and difficult-to-trace decision-making processes in existing technologies, this invention uses a coordinating / communication agent to record in detail the negotiation process and decision-making basis between agents. This allows managers to clearly understand the background of each decision, facilitating supervision and evaluation of the decision-making process, improving transparency and traceability, and enhancing the scientific nature and credibility of the decisions.

[0069] (4) Addressing the problem of high manpower and time consumption in traditional maintenance work, especially the low efficiency in managing large-scale bridge groups, this invention achieves full-process automation from data collection to maintenance plan formulation through efficient collaboration between monitoring agents, diagnostic agents, decision-making agents, resource scheduling agents, and coordination / communication agents deployed on cloud servers. The collaborative operation of these agents significantly shortens the decision-making cycle, reduces manual intervention, and alleviates the burden on management personnel. It is particularly suitable for the centralized and efficient management of large-scale bridge groups, enabling rapid response to the actual needs of bridges and improving the operational efficiency of the entire transportation network. Existing technologies suffer from blind resource allocation. The resource scheduling agent of this invention accurately allocates human, material, and financial resources based on the optimal plan generated by the decision-making agent, avoiding resource waste and idleness, improving resource utilization efficiency, reducing maintenance costs, and further enhancing the overall efficiency of maintenance work. Attached Figure Description

[0070] Figure 1This is a schematic diagram of a bridge maintenance decision-making system based on multi-agent collaborative optimization according to an embodiment of the present invention;

[0071] Figure 2 This is a schematic diagram of the structure of a diagnostic agent in a bridge maintenance decision-making system based on multi-agent collaborative optimization, according to an embodiment of the present invention.

[0072] Figure 3 This is a schematic diagram of the structure of a decision-making agent in a bridge maintenance decision-making system based on multi-agent collaborative optimization, according to an embodiment of the present invention.

[0073] Figure 4 This is a schematic diagram of the structure of a resource scheduling agent in a bridge maintenance decision-making system based on multi-agent collaborative optimization, according to an embodiment of the present invention.

[0074] Figure 5 This is a schematic diagram of the structure of a coordination / communication agent in a bridge maintenance decision-making system based on multi-agent collaborative optimization, according to an embodiment of the present invention.

[0075] Figure 6 This is a flowchart illustrating a bridge maintenance decision-making method based on multi-agent collaborative optimization according to an embodiment of the present invention.

[0076] Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention.

[0077] Related glossary:

[0078] Monitoring Agent (MA);

[0079] Diagnosis Agent (DA);

[0080] Decision Agent (DeA);

[0081] Resource Scheduling Agent (RSA);

[0082] Coordination / Communication Agent (CA). Detailed Implementation

[0083] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0084] Based on the above reasons, such as Figures 1-5 As shown, one aspect of the present invention provides a bridge maintenance decision-making system based on multi-agent collaborative optimization, including a monitoring agent, a diagnostic agent, a decision-making agent, a resource scheduling agent, and a coordination / communication agent deployed on a cloud server;

[0085] The monitoring agent is connected to embedded sensors, a drone inspection system, and a manual reporting terminal, respectively, to acquire structural and environmental data of the bridge, filter out abnormal data, and send it to the coordination / communication agent.

[0086] The diagnostic agent is used to receive abnormal data forwarded by the coordination / communication agent, use a built-in knowledge base, machine learning model and physical model to assess the health status of the bridge, identify the damage type, location and severity, output the structural health index (BHI) to generate a diagnostic report, and send the diagnostic report to the coordination / communication agent.

[0087] The decision-making agent is used to receive diagnostic reports forwarded by the coordination / communication agent, and based on the diagnostic reports output by the diagnostic agent, comprehensively consider a preset maintenance objective function to generate multiple feasible candidate maintenance schemes and submit them to the coordination / communication agent; the decision-making agent, by combining the comprehensive utility of each candidate maintenance scheme and the resource matching score given by the resource scheduling agent for each candidate maintenance scheme, selects the optimal maintenance scheme;

[0088] The Resource Scheduling Agent (RSA) receives multiple feasible candidate maintenance schemes forwarded by the Coordination / Communication Agent, and, following the negotiation instructions issued by the Coordination / Communication Agent, performs a resource feasibility assessment on each candidate maintenance scheme and provides a corresponding resource matching score. Based on the optimal maintenance scheme selected by the Decision Agent, the Resource Scheduling Agent uses optimization algorithms (such as genetic algorithms and ant colony algorithms) to coordinate and plan the resources required for executing the optimal maintenance scheme, including manpower, materials, equipment, and time windows, to generate a specific maintenance construction plan, which is then submitted to the Coordination / Communication Agent.

[0089] The coordination / communication agent is communicatively connected to the monitoring agent, diagnostic agent, decision-making agent, and resource scheduling agent, respectively. It defines the communication protocols and negotiation mechanisms between these agents, ensuring efficient collaboration and avoiding information silos and conflicts. It integrates the diagnostic report, the optimal maintenance plan, and the maintenance construction schedule, presenting them to bridge managers through a human-computer interface. After maintenance is implemented, the coordination / communication agent reassesses the bridge's structural health index based on new data collected by the monitoring agent. Using the new monitoring data and the reassessed structural health index, it updates and optimizes the diagnostic model of the diagnostic agent online, and controls the decision-making agent to update and optimize its decision-making strategy model online based on the new structural health index and the new diagnostic report, forming a "decision-execution-feedback" closed loop. Through continuous feedback and self-learning, it achieves collaborative optimization of bridge maintenance decisions.

[0090] Furthermore, the monitoring agent is a key component of the bridge maintenance decision-making system responsible for data acquisition and preliminary processing. It collects structural and environmental data of the bridge from multiple sources in real time or periodically, and performs data cleaning, preprocessing, and preliminary anomaly detection.

[0091] The monitoring intelligent agent includes a data acquisition module, a data cleaning module, a data preprocessing module, and a preliminary anomaly detection module;

[0092] The data acquisition module is used to collect bridge-related data from multiple data sources; these data sources include:

[0093] Embedded sensors: Sensors installed inside or on the surface of bridge structures to monitor physical quantities such as strain, displacement, temperature, and humidity of the bridge in real time.

[0094] Unmanned aerial vehicle (UAV) inspection system: Using UAVs equipped with high-definition cameras, thermal imagers, lidar and other equipment, the system inspects the appearance, thermal imaging and three-dimensional shape of bridges.

[0095] Manual reporting terminal: Inspectors use terminal equipment on-site to record and report bridge defects such as cracks, peeling, and corrosion;

[0096] The data cleaning module is used to preprocess the collected raw data, such as removing noise, filling in missing values, and correcting erroneous data.

[0097] The data preprocessing module is used to further process the cleaned data, including data format unification, data normalization, and feature extraction.

[0098] The preliminary anomaly detection module is used to perform preliminary analysis on the preprocessed data, identify and mark abnormal data. Specifically, by setting a threshold, an alarm is triggered when sensor data exceeds the normal range; or image recognition technology is used to preliminarily determine whether there are defects such as cracks in the image.

[0099] Furthermore, the diagnostic agent is a key component in the bridge maintenance decision-making system responsible for assessing the health status of bridges. It receives preliminary abnormal data forwarded by the coordination / communication agent, identifies the type, location, and severity of bridge damage through a series of complex data processing and analysis procedures, and outputs a structural health index (BHI); the structural health index is a value between 0 and 1.

[0100] like Figure 2 As shown, the diagnostic intelligent agent includes a data receiving module, a feature extraction module, a knowledge base module, a machine learning and physical model fusion module, and a health status assessment module;

[0101] The data receiving module is responsible for receiving preliminary abnormal data forwarded by the coordinating / communication agent;

[0102] The feature extraction module is used to extract features related to the bridge's health status from the data received by the data receiving module. These features include stress concentration areas in strain data, deformation trends in displacement data, and crack length and width in image data. For sensor data, feature extraction involves signal processing techniques, such as Fourier transform for frequency feature analysis. For image data, image processing algorithms such as edge detection and texture analysis are used to extract features of defects such as cracks. For example, edge detection algorithms can clearly outline the contours of cracks, providing a basis for subsequent crack width and length measurements.

[0103] The feature extraction described above can transform raw data into information meaningful for diagnosis, reduce data dimensionality, and improve diagnostic efficiency and accuracy. Accurately extracted features are crucial for diagnostic algorithms to correctly identify the type and severity of injuries.

[0104] The knowledge base module stores various knowledge related to bridge health diagnosis, including bridge design specifications, historical case data, material performance parameters, and damage modes. This knowledge provides theoretical support and empirical reference for the diagnostic process. For example, when the diagnostic agent identifies a specific crack mode, it can refer to historical cases in the knowledge base to understand the possible causes and severity of damage corresponding to that crack mode; or, based on bridge design specifications, it can determine whether the currently measured strain value exceeds the allowable range. The data in the knowledge base can be structured text information or multimedia data such as images and models.

[0105] The knowledge base module serves as the "brain" of the diagnostic agent, enabling it to combine professional knowledge and historical experience in the diagnostic process, thereby improving the accuracy and reliability of diagnoses. A rich knowledge base can compensate for the shortcomings of a single data-driven approach, especially in the face of complex or rare injuries, providing more comprehensive diagnostic evidence.

[0106] The machine learning and physical model fusion module is used for a comprehensive assessment of the bridge's health status. The machine learning model includes a convolutional neural network (CNN) and a long short-term memory network (LSTM). The CNN is used for image recognition and automatically learns patterns of defects such as cracks. The LSTM is used to process time-series data, such as strain and displacement data monitored by sensors, to identify potential damage development trends. Simultaneously, the physical model includes a finite element model (FEM) to simulate the stress and deformation of the bridge under different working conditions based on its structural parameters and physical characteristics, providing theoretical support for diagnosis. Specifically, when identifying cracks, the machine learning model quickly locates the crack position and initially determines the crack type, while the physical model further assesses the impact of the crack on the structure based on the stress state at the crack location.

[0107] This invention overcomes the limitations of single methods by fusing machine learning with physical models. Machine learning can process large amounts of complex data and automatically discover patterns within it, while physical models provide explanations and predictions based on physical principles. This fusion makes diagnostic results more accurate and reliable, and better adaptable to different types of bridges and damage conditions.

[0108] The health status assessment module is used to comprehensively assess the overall health status of the bridge based on the features extracted by the feature extraction module, the knowledge in the knowledge base module, and the analysis results of the machine learning and physical model fusion module. It integrates information on various damage types, locations, and severity to calculate the Structure Health Index (BHI). The BHI is a quantitative indicator that intuitively reflects the current health status of the bridge. The BHI value is weighted according to factors such as the severity of damage, the importance of its location, and its quantity. A lower BHI value indicates that the bridge may have serious health problems, requiring timely maintenance measures. The health status assessment module is the output link of the diagnostic agent; it transforms the complex diagnostic process into a concise and clear quantitative indicator, providing direct decision-making basis for the subsequent decision-making agent.

[0109] Furthermore, the structural health index is calculated using equations (1) to (3):

[0110]

[0111] Where i represents the damage type; N represents the total number of damages; C i The contribution of the i-th injury to overall health status; w i Let D be the weight of the i-th injury, based on the severity D of the i-th injury. i and positional importance L i Confirmed; w total D is the sum of all damage weights; i This represents the severity of the i-th damage (e.g., crack width, strain value, etc.); 1-D i To represent the health contribution of the injury, let's assume D. i The value is between [0,1], where 0 represents no damage and 1 represents complete damage;

[0112] Assume weight w i With positional importance L i Proportional; directly using positional importance as the weight, i.e., w i =L i Then the total weight w total This can be expressed by equation (4):

[0113]

[0114] When calculating the Structural Health Index (BHI), the weight w for each injury is... i After normalization, the sum of the weights of all damages is 1, ensuring that the calculated BHI value is between [0,1]. A BHI value close to 1 indicates that the bridge structure is in good health and the degree of damage is low; a BHI value close to 0 indicates that the bridge structure is in poor health and the degree of damage is high.

[0115] Furthermore, the decision-making agent is the core component of the bridge maintenance decision-making system. It is responsible for generating multiple feasible candidate maintenance schemes based on the bridge health status information provided by the diagnostic agent, taking into account the preset maintenance objective function, and selecting the optimal maintenance scheme by combining the comprehensive utility of each candidate maintenance scheme and the resource matching score given by the resource scheduling agent for each candidate maintenance scheme.

[0116] like Figure 3 As shown, the decision-making intelligent agent includes a data receiving and parsing module and a decision-making module;

[0117] The data receiving and parsing module is responsible for receiving health status information output by the diagnostic agent, including the Structural Health Index (BHI), lesion type, location, and severity data. Upon receipt, this data is parsed and converted into a format that the decision-making module can understand and process. For example, BHI values ​​are converted into specific health levels (e.g., good, moderate, poor), and lesion locations and types are classified and labeled for subsequent processing by other modules.

[0118] The decision-making module is used to perform comprehensive analysis and processing on the health status information parsed by the data receiving and parsing module, and finally generate a maintenance plan.

[0119] The decision-making module includes a multi-objective maintenance utility function definition module, a scheme generation module, a comprehensive utility evaluation module, and an optimal maintenance scheme selection module.

[0120] The multi-objective maintenance utility function definition module is used to define and manage the multi-objective maintenance utility function (U_total). The multi-objective maintenance utility function includes multiple objectives, including cost minimization, safety maximization, traffic impact minimization, and service life maximization; each objective has a corresponding weight, which is adjusted according to the specific conditions of the bridge and maintenance requirements. The multi-objective maintenance utility function definition module is the "compass" of the decision-making agent, determining the optimization direction of the maintenance scheme. By reasonably defining and adjusting the objective function, it can be ensured that the generated maintenance scheme meets actual needs, achieving a balance and optimization of multiple objectives; the multi-objective maintenance utility function aims to maximize comprehensive utility, expressed by equation (5):

[0121] max(U total )=w1·f s (S)+w2·f c (S)·w3·f t (S)+w4·f l (S) (5)

[0122] Among them, U total The comprehensive objective function value represents the overall performance of the maintenance scheme; max(U total ) represents the maximized overall utility; S represents a specific maintenance plan; f s (S) represents the expected safety utility function after the implementation of scheme S, which is positively correlated with the improvement of the structural health index (BHI); f c (S) is the cost-utility function of option S, usually negative, and negatively correlated with maintenance costs; f t (S) is the utility function of the traffic impact of plan S, which is usually negative and negatively correlated with the duration and extent of traffic disruption; f l(S) is the utility function of the impact of scheme S on the remaining life of the bridge; w1, w2, w3, and w4 are respectively the utility function of f s (S), f c (S), f t (S), f l (S) The weight coefficients of each utility function are set by the manager based on factors such as the importance of the bridge and the budget, and w1+w2+w3+w4=1;

[0123] The scheme generation module is used to generate multiple feasible candidate maintenance schemes based on the parsed health status information and a preset multi-objective maintenance utility function, using algorithms (such as genetic algorithms, simulated annealing algorithms, etc.). These schemes include no intervention, minor repairs, reinforcement, load limiting, and component replacement.

[0124] The comprehensive utility evaluation module performs a detailed evaluation of each candidate maintenance plan generated by the plan generation module, calculating the specific performance of each plan on various objective utility functions to obtain the comprehensive utility of each candidate maintenance plan. For example, it evaluates the cost, traffic impact, improvement in bridge safety, and extension of service life of each plan. The evaluation results are presented in the form of quantitative indicators, providing a decision-making basis for the optimal maintenance plan selection module.

[0125] The optimal maintenance scheme selection module integrates the overall utility of each candidate maintenance scheme and the resource matching score given by the resource scheduling agent for each candidate maintenance scheme. Through a multi-objective optimization algorithm (such as Pareto optimization algorithm), it finds the optimal balance point among multiple objectives and selects the optimal maintenance scheme.

[0126] Furthermore, such as Figure 4 As shown, the resource scheduling intelligent agent includes a resource demand analysis module, a resource inventory and availability management module, a resource matching degree scoring module, and a construction plan generation module.

[0127] The resource requirement analysis module is used to perform resource requirement analysis on the candidate maintenance plans generated by the decision-making module, determining the human, material, financial, and time resources required to implement the optimal maintenance plan. Specifically, this includes calculating the required number of workers, types and quantities of materials, types and quantities of equipment, and the estimated construction time. This information will provide the resource scheduling agent with a detailed resource requirement list to facilitate resource scheduling and construction plan development.

[0128] The resource inventory and availability management module manages currently available resource inventory, including human, material, and financial resources. By updating resource availability status in real time, it ensures that resources can be rationally allocated based on the current situation.

[0129] The resource matching degree scoring module is used to follow the negotiation instructions issued by the coordination / communication agent, and based on the results of the resource demand analysis module and the real-time data of the resource inventory and availability module, to evaluate the resource feasibility of each candidate maintenance scheme and give a corresponding resource matching degree score.

[0130] The construction plan generation module, based on the optimal maintenance scheme selected by the decision-making agent, uses optimization algorithms (such as genetic algorithms and ant colony algorithms) to coordinate and plan the resources such as manpower, materials, equipment, and time windows required to execute the optimal maintenance scheme, generates a specific maintenance construction plan, and submits the construction plan to the coordination / communication agent;

[0131] The maintenance and construction plan includes detailed steps for each task, start and end times, and specific information on required resources.

[0132] Furthermore, the coordination / communication agent is the central component of the bridge maintenance decision-making system, responsible for information exchange, task distribution, conflict coordination, and negotiation among all agents. It ensures efficient collaboration among agents, avoids information silos and conflicts, and achieves overall system optimization.

[0133] like Figure 5 As shown, the coordination / communication agent includes a data receiving and forwarding module, a task distribution module, a conflict coordination module, a communication protocol and negotiation mechanism definition module between agents, an information integration and presentation module, and a feedback and self-learning module;

[0134] The data receiving and forwarding module is responsible for receiving data from the monitoring agent, diagnostic agent, decision-making agent, and resource scheduling agent, and forwarding this data to the corresponding execution agent. Specifically, it receives data from each agent, including monitoring data, diagnostic reports, candidate solution sets, resource matching scores, etc.; parses the received data to ensure that the data format is correct and the content is complete; and forwards the parsed data to the corresponding agent.

[0135] The task distribution module is used to generate specific tasks according to the overall needs of the system, assign the tasks to the corresponding intelligent agents, and ensure the execution of the tasks.

[0136] The conflict coordination module is used to resolve conflicts between agents and ensure the collaborative operation of the system. Specifically, it includes conflict detection, which detects conflicts between agents in real time, such as resource allocation conflicts and task priority conflicts.

[0137] Conflict resolution involves resolving conflicts according to pre-defined rules and strategies; for example, by adjusting task priorities or reallocating resources.

[0138] The negotiation mechanism initiates a negotiation process, allowing relevant intelligent agents to negotiate. For example, in the solution negotiation and optimization phase, the negotiation process is initiated, allowing the decision-making intelligent agent and the resource scheduling intelligent agent to negotiate. The resource scheduling intelligent agent evaluates the resource feasibility of each solution and provides a resource matching score. Finally, the decision-making intelligent agent selects the solution with the highest comprehensive utility and resource matching score as the optimal maintenance solution.

[0139] The module defining communication protocols and negotiation mechanisms between agents is used to define the communication protocols and negotiation mechanisms between agents, ensuring that information exchange and negotiation processes between agents are efficient and orderly; specifically, it includes:

[0140] Define the communication protocol between the agents, including data format, transmission method, communication frequency, etc.

[0141] Define a negotiation mechanism, including the triggering conditions, the negotiation process, and the handling of the negotiation results. For example, use the Contract Net Protocol for negotiation to ensure that the negotiation process is efficient and orderly.

[0142] Manage the updating and maintenance of communication protocols and negotiation mechanisms to ensure their adaptability and effectiveness;

[0143] The information integration and presentation module is used to integrate the information provided by various intelligent agents and present it to the bridge manager through a human-computer interaction interface; specifically, it includes:

[0144] Information integration involves combining monitoring data, diagnostic reports, candidate solution sets, work plans, and other information to form a complete system status view;

[0145] Information presentation involves using a human-computer interaction interface to present the integrated information to bridge managers in a clear and intuitive way; for example, generating detailed reports and visual charts.

[0146] User interaction supports interaction between users and the system. For example, users can view detailed information, adjust task priorities, and approve plans.

[0147] The feedback and self-learning module is used to reassess the structural health index of the bridge based on the new data collected by the monitoring agent; using the new monitoring data and the reassessed structural health index, the diagnostic model of the diagnostic agent is updated and optimized online; and the control decision agent updates and optimizes its decision-making strategy model online based on the new structural health index and the new diagnostic report, forming a closed loop of "decision-execution-feedback".

[0148] Through a closed loop of "decision-execution-feedback", the bridge maintenance decision system based on multi-agent collaborative optimization of the present invention can continuously learn and adapt to the actual condition and changes of the bridge. Through continuous feedback and self-learning, it continuously improves its performance, realizes the collaborative optimization of bridge maintenance decisions, and thus improves the scientificity and accuracy of bridge maintenance decisions.

[0149] It should be noted that the bridge maintenance decision system based on multi-agent collaborative optimization provided in this embodiment can be a computer program (including program code) running on a computer device. For example, the bridge maintenance decision system based on multi-agent collaborative optimization is an application software. The bridge maintenance decision system based on multi-agent collaborative optimization can be used to execute the corresponding steps in the above-described method provided in the embodiments of this application.

[0150] In some feasible implementations, the bridge maintenance decision-making system based on multi-agent cooperative optimization provided in this embodiment can be implemented in a combination of hardware and software. As an example, the bridge maintenance decision-making system based on multi-agent cooperative optimization provided in this application embodiment can be a processor in the form of a hardware decoding processor, which is programmed to execute the bridge maintenance decision-making method based on multi-agent cooperative optimization provided in this application embodiment. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0151] In some feasible implementations, the bridge maintenance decision-making system based on multi-agent collaborative optimization provided in this embodiment can be implemented in software. It can be software in the form of programs and plug-ins, and includes a series of modules to implement the bridge maintenance decision-making method based on multi-agent collaborative optimization provided in this embodiment of the invention.

[0152] like Figure 6 As shown, a second aspect of the present invention provides a bridge maintenance decision-making method based on multi-agent cooperative optimization, implemented using the aforementioned bridge maintenance decision-making system based on multi-agent cooperative optimization, comprising the following steps:

[0153] S1: Data Acquisition and Preprocessing. The monitoring agent continuously collects structural and environmental data of the bridge, performs noise reduction and normalization on the collected data, marks abnormal data, and sends it to the coordination / communication agent.

[0154] Specifically, the structural and environmental data of the bridge includes sensor data such as strain, displacement, temperature, and humidity, as well as data from drone inspections and manual reports. Embedded sensors, drone inspection systems, and manual reporting terminals are used to collect data. Signal processing techniques (such as Fourier transform) and image processing techniques (such as edge detection) are applied for data preprocessing to obtain high-quality data that has undergone noise reduction and normalization, and outliers in the data are detected and marked.

[0155] S2: The coordination / communication agent forwards the abnormal data preprocessed in step S1 to the diagnostic agent. The diagnostic agent uses its built-in knowledge base, machine learning model, and physical model to assess the health status of the bridge, identify the damage type, location, and severity, output a structural health index, generate a diagnostic report, and send the diagnostic report to the coordination / communication agent. The diagnostic report contains detailed information such as the structural health index, damage type, location, and severity.

[0156] S3: The coordination / communication agent distributes the diagnostic report to the decision agent, which generates one or more candidate maintenance scheme sets according to a preset multi-objective maintenance utility function and submits them to the coordination / communication agent;

[0157] S4: The coordination / communication agent distributes the candidate maintenance scheme set to the resource scheduling agent, and initiates a negotiation process between the resource scheduling agent and the decision-making agent according to a predetermined negotiation mechanism. The resource scheduling agent evaluates the resource feasibility of each candidate maintenance scheme according to the negotiation instructions issued by the coordination / communication agent, and gives a corresponding resource matching score. The decision-making agent calculates the comprehensive utility of each candidate maintenance scheme, and combines the comprehensive utility of each candidate maintenance scheme with the resource matching score, and uses an optimization algorithm to select the optimal maintenance scheme.

[0158] S5: The resource scheduling agent coordinates and plans the resources such as manpower, materials, equipment and time windows required to execute the optimal maintenance plan according to the optimal maintenance plan selected by the decision-making agent, and generates a specific maintenance construction plan, and submits the construction plan to the coordination / communication agent.

[0159] S6: The coordination / communication agent integrates the diagnostic report, the optimal maintenance plan, and the maintenance construction plan, and presents them to the bridge manager through a human-computer interaction interface; the bridge manager can view detailed information, adjust task priorities, and approve maintenance plans through the human-computer interaction interface.

[0160] S7: After maintenance is implemented, the coordination / communication agent reassesses the structural health index of the bridge based on the new data collected by the monitoring agent. Using the new monitoring data and the reassessed structural health index, the diagnostic model of the diagnostic agent is updated and optimized online. The control decision agent updates and optimizes its decision-making strategy model online based on the new structural health index and the new diagnostic report, forming a "decision-execution-feedback" closed loop. Through continuous feedback and self-learning, the collaborative optimization of bridge maintenance decisions is achieved.

[0161] Furthermore, the structural health index described in step S2 is calculated using equations (1) to (3):

[0162]

[0163]

[0164] Where i represents the damage type; N represents the total number of damages; C i The contribution of the i-th injury to overall health status; w i Let D be the weight of the i-th injury, based on the severity D of the i-th injury. i and positional importance L i Confirmed; w total D is the sum of all damage weights; i This represents the severity of the i-th damage (e.g., crack width, strain value, etc.); 1-D i To represent the health contribution of the injury, let's assume D. i The value is between [0,1], where 0 represents no damage and 1 represents complete damage;

[0165] Assume weight w i With positional importance L i Proportional; directly using positional importance as the weight, i.e., w i =L i Then the total weight w total This can be expressed by equation (4):

[0166]

[0167] Furthermore, the multi-objective maintenance utility function described in step S3 is expressed by equation (5):

[0168] max(U total )=w1·f s(S)+w2·f c (S)+w3·f t (S)+w4·f l (S) (5)

[0169] Among them, U total The comprehensive objective function value represents the overall performance of the maintenance scheme; max(U total ) represents the maximized overall utility; S represents a specific maintenance plan; f s (S) represents the expected safety utility function after the implementation of scheme S, which is positively correlated with the improvement of the structural health index (BHI); f c (S) is the cost-utility function of option S, usually negative, and negatively correlated with maintenance costs; f t (S) is the utility function of the traffic impact of plan S, which is usually negative and negatively correlated with the duration and extent of traffic disruption; f l (S) is the utility function of the impact of scheme S on the remaining life of the bridge; w1, w2, w3, and w4 are respectively the utility function of f s (S), f c (S), f t (S), f l (S) The weight coefficients of each utility function are set by the manager based on factors such as the importance of the bridge and the budget, and w1+w2+w3+w4=1.

[0170] The following example, "Maintenance Decision-Making for Crack Development in an Elevated Bridge in a Certain City," illustrates the steps of the bridge maintenance decision-making method based on multi-agent collaborative optimization of this invention. Specifically, the steps include:

[0171] S100, System Initialization and Agent Deployment

[0172] Strain and displacement sensors were deployed at key sections of an elevated bridge in a certain city, and edge computing nodes were configured to run a monitoring agent on the nodes.

[0173] Deploy a coordination / communication agent, a diagnostic agent, a decision-making agent, and a resource scheduling agent on a cloud server; the diagnostic agent is pre-loaded with a CNN model trained based on historical data for identifying cracks and a finite element model corresponding to the bridge; the decision-making agent is configured with utility function weights set by the bridge management: safety weight w1 = 0.5, cost weight w2 = 0.3, and traffic impact weight w3 = 0.2.

[0174] S200: The monitoring agent continuously receives sensor data and filters and normalizes the data periodically (e.g., hourly). If the monitoring agent detects an abnormal pattern of continuous increase in the strain data of a certain beam segment over the past 24 hours, it immediately sends the data packets and abnormal tags for that time period to the coordination / communication agent.

[0175] After receiving the data, the S300 and the coordination / communication agent assign the task to the diagnostic agent. The diagnostic agent first uses a CNN model to analyze the associated camera inspection images, identifies a main crack, and preliminarily determines its width. Then, the diagnostic agent uses the crack information as input to update the finite element model, calculates that the load-bearing capacity of the beam segment has decreased by 8%, and evaluates the current structural health index (BHI) to 0.75 (out of 1). It then generates a diagnostic report containing crack images, location, size, BHI, and risk warnings, and sends it back to the coordination / communication agent.

[0176] If the S400 coordination / communication agent determines that the structural health index BHI (0.75) is below the preset intervention threshold (0.8), it immediately forwards the diagnostic report to the decision-making agent. The decision-making agent triggers the decision-making process, generates three candidate solutions based on built-in rules and a multi-objective maintenance utility function, and performs a utility evaluation. The three candidate solutions are as follows:

[0177] Option 1: Crack sealing treatment (minor repair); expected BHI to increase to 0.9, cost 50,000 yuan, requires closing one lane at night for 8 hours;

[0178] Option 2: Reinforcement with carbon fiber cloth (mid-section repair); expected BHI to increase to 0.98, cost 300,000 yuan, requires closing two lanes at night for 48 hours;

[0179] Option 3: Vehicle load limit (management measures), no maintenance required for now; BHI remains unchanged, cost is 0, but road traffic efficiency will be reduced;

[0180] S500 and the decision-making agent submit the three solutions along with their utility evaluation values ​​to the coordination / communication agent. The coordination / communication agent initiates a negotiation based on the contract network protocol. The coordination / communication agent, as the "bidder," broadcasts the resource requirements of the three solutions to the resource scheduling agent. The resource scheduling agent, as the "bidder," evaluates the matching degree of its own resource pool (available construction teams, material inventory, equipment status) with the three solutions and provides the following "bid documents": Solution 1 resource matching degree 95% (ready to execute at any time), Solution 2 resource matching degree 60% (requires external purchase of carbon cloth, waiting for one week), Solution 3 resource matching degree 100%.

[0181] S600 and the decision-making agent calculate Scheme 1 as the current optimal choice based on the comprehensive utility evaluation value and resource matching degree and send it to the coordination / communication agent; the coordination / communication agent sends Scheme 1 to the resource scheduling agent, instructing it to formulate a specific maintenance and construction plan; the resource scheduling agent then generates a detailed night construction plan, including personnel shifts, material requisition forms, traffic control application periods, etc.; the coordination / communication agent integrates the diagnostic report, the scheme comparison process, the final scheme and the construction plan, and presents them to the engineers of the bridge management center through a large visualization screen;

[0182] After review by engineers at the S700 and Bridge Management Center, the system automatically sends the construction plan to the maintenance team's mobile terminal. After the maintenance work is completed, the data collected by the monitoring agent shows that the bridge strain has returned to normal, and the diagnostic agent reassesses the structural health index (BHI) as 0.92.

[0183] Finally, the data from the "problem discovery-decision-execution-effect evaluation" process is stored in a knowledge base for future reinforcement learning training of the decision-making agent's decision model, enabling it to make more accurate decisions when faced with similar situations in the future.

[0184] A third aspect of the present invention also provides an electronic device, Figure 7 This is a schematic diagram of the electronic device in this embodiment, as shown below. Figure 7 As shown, the electronic device 1000 in this embodiment may include: a processor 1001, a network interface 1004, and a memory 1005. Furthermore, the electronic device 1000 may also include: a user interface 1003, and at least one communication bus 1002. The communication bus 1002 is used to implement communication between these components. The user interface 1003 may include a display screen and a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory, such as at least one disk storage device. Optionally, the memory 1005 may also be at least one storage device located remotely from the aforementioned processor 1001. Figure 7 As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a device control application.

[0185] like Figure 7In the electronic device 1000 shown, the network interface 1004 provides network communication functions; the user interface 1003 is mainly used to provide an input interface for users; and the processor 1001 can be used to call the device control application stored in the memory 1005 to implement each step of the bridge maintenance decision-making method based on multi-agent cooperative optimization.

[0186] It should be understood that in some feasible implementations, the processor 1001 described above may be a central processing unit (CPU), which may also be other general-purpose processors, DSPs, ASICs, FPGAs, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

[0187] In specific implementation, the aforementioned electronic device 1000 can perform the above-described actions through its built-in functional modules. Figure 2 The implementation methods provided for each step are detailed in the above-mentioned implementation methods, and will not be repeated here.

[0188] This application also provides a computer-readable storage medium storing a computer program that is executed by a processor to implement... Figure 2 The methods provided in each step are detailed in the implementation methods provided in the above steps, and will not be repeated here.

[0189] Any references to memory, storage, database, or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0190] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A bridge maintenance decision-making system based on multi-agent collaborative optimization, characterized in that: This includes monitoring agents, diagnostic agents, decision-making agents, resource scheduling agents, and coordination / communication agents deployed on cloud servers; The monitoring agent is used to acquire structural and environmental data of the bridge, filter out abnormal data, and send it to the coordination / communication agent. The diagnostic agent is used to receive abnormal data forwarded by the coordination / communication agent, and to assess the health status of the bridge using a built-in knowledge base, machine learning model and physical model, identify the type, location and severity of damage, and output a structural health index to generate a diagnostic report. The decision-making agent is used to generate multiple feasible candidate maintenance schemes based on the diagnostic report output by the diagnostic agent, comprehensively consider the preset multi-objective maintenance utility function, calculate the comprehensive utility of each candidate maintenance scheme, and select the optimal maintenance scheme by combining the comprehensive utility of each candidate maintenance scheme and the resource matching degree score given by the resource scheduling agent for each candidate maintenance scheme. The resource scheduling agent is used to follow the negotiation instructions issued by the coordination / communication agent, perform resource feasibility assessment on each candidate maintenance scheme, and give a corresponding resource matching score; based on the optimal maintenance scheme selected by the decision agent, it uses an optimization algorithm to generate a specific maintenance construction plan and submits the construction plan to the coordination / communication agent. The coordination / communication agent ensures efficient collaboration among its various agents through communication protocols and negotiation mechanisms; it integrates the diagnostic report, the optimal maintenance plan, and the maintenance construction plan, presenting them to bridge managers through a human-computer interaction interface; and through continuous feedback and self-learning, it achieves collaborative optimization of bridge maintenance decisions.

2. The bridge maintenance decision-making system based on multi-agent collaborative optimization according to claim 1, characterized in that: The monitoring intelligent agent includes a data acquisition module, a data cleaning module, a data preprocessing module, and a preliminary anomaly detection module; The data acquisition module is used to collect bridge-related data from multiple data sources; The data sources include: embedded sensors for real-time monitoring of bridge strain, displacement, temperature, and humidity; a drone inspection system that uses high-definition cameras, thermal imagers, and lidar mounted on drones to inspect the bridge's appearance, thermal imaging, and three-dimensional shape; and a manual reporting terminal for inspection personnel to record and report bridge defects on-site. The data cleaning module is used to preprocess the collected raw data; The data preprocessing module is used to further process the cleaned data, including data format unification, data normalization, and feature extraction. The preliminary anomaly detection module is used to perform preliminary analysis on the preprocessed data, identify and mark abnormal data.

3. A bridge maintenance decision-making system based on multi-agent collaborative optimization according to claim 2, characterized in that: The diagnostic intelligent agent includes a data receiving module, a feature extraction module, a knowledge base module, a machine learning and physical model fusion module, and a health status assessment module; The data receiving module is responsible for receiving preliminary abnormal data forwarded by the coordinating / communication agent; The feature extraction module is used to extract features related to the bridge's health status from the preliminary anomaly data; the features related to the bridge's health status include stress concentration areas in strain data, deformation trends in displacement data, and crack length and width in image data; The knowledge base module is used to store various knowledge related to bridge health diagnosis, including bridge design specifications, historical case data, material performance parameters, and damage modes. The machine learning and physical model fusion module is used to comprehensively assess the health status of the bridge; the machine learning model includes a convolutional neural network (CNN) and a long short-term memory network (LSTM); the physical model includes a finite element model (FEM) to simulate the stress and deformation of the bridge under different working conditions based on the bridge's structural parameters and physical characteristics. The health status assessment module is used to comprehensively assess the overall health status of the bridge based on the features extracted by the feature extraction module, the knowledge in the knowledge base module, and the analysis results of the machine learning and physical model fusion module; it integrates information on various damage types, locations, and severity to calculate the structural health index.

4. A bridge maintenance decision-making system based on multi-agent collaborative optimization according to any one of claims 1-3, characterized in that: The decision-making intelligent agent includes a data receiving and parsing module and a decision-making module; The data receiving and parsing module is responsible for receiving the health status information output by the diagnostic agent, including structural health index, damage type, location and severity data; and parsing this data to convert it into a format that the decision module can understand and process. The decision-making module is used to comprehensively analyze and process the health status information parsed by the data receiving and parsing module, and finally generate a maintenance plan; the decision-making module includes a multi-objective maintenance utility function definition module, a plan generation module, a comprehensive utility evaluation module, and an optimal maintenance plan selection module; The multi-objective maintenance utility function definition module is used to define and manage multi-objective maintenance utility functions. The multi-objective maintenance utility function includes multiple objectives, including cost minimization, safety maximization, traffic impact minimization, and service life maximization. Each objective has a corresponding weight, which is adjusted according to the specific conditions of the bridge and maintenance requirements. The scheme generation module is used to generate multiple feasible candidate maintenance schemes based on the parsed health status information and the preset multi-objective maintenance utility function through an algorithm. The comprehensive utility evaluation module performs a detailed evaluation of each candidate maintenance plan generated by the plan generation module, calculates the specific performance of each plan on each objective utility function, and obtains the comprehensive utility of each candidate maintenance plan. The optimal maintenance scheme selection module integrates the overall utility of each candidate maintenance scheme and the resource matching score given by the resource scheduling agent for each candidate maintenance scheme. Through a multi-objective optimization algorithm, it finds the optimal balance point among multiple objectives and selects the optimal maintenance scheme.

5. A bridge maintenance decision-making system based on multi-agent collaborative optimization according to any one of claims 1-3, characterized in that: The resource scheduling intelligent agent includes a resource demand analysis module, a resource inventory and availability management module, a resource matching degree scoring module, and a construction plan generation module; The resource demand analysis module is used to perform resource demand analysis on the candidate maintenance plans generated by the decision-making module, and determine the human, material, financial and time resources required to implement the optimal maintenance plan; The resource inventory and availability management module is used to manage the currently available resource inventory, including human, material and financial resources. By updating the availability status of resources in real time, it ensures that reasonable scheduling can be carried out according to the current resource situation. The resource matching degree scoring module is used to follow the negotiation instructions issued by the coordination / communication agent, and based on the results of the resource demand analysis module and the real-time data of the resource inventory and availability module, to evaluate the resource feasibility of each candidate maintenance plan and give a corresponding resource matching degree score. The construction plan generation module generates a specific maintenance construction plan based on the optimal maintenance scheme selected by the decision-making agent, using optimization algorithms to coordinate and plan the manpower, materials, equipment, and time window resources required to execute the optimal maintenance scheme.

6. A bridge maintenance decision-making system based on multi-agent cooperative optimization according to any one of claims 1-3, characterized in that: The coordination / communication agent includes a data receiving and forwarding module, a task distribution module, a conflict coordination module, a communication protocol and negotiation mechanism definition module between agents, an information integration and presentation module, and a feedback and self-learning module. The data receiving and forwarding module is responsible for receiving data from the monitoring agent, the diagnostic agent, the decision-making agent, and the resource scheduling agent, and forwarding this data to the corresponding execution agents. The task distribution module is used to generate specific tasks according to the overall needs of the system, assign the tasks to the corresponding intelligent agents, and ensure the execution of the tasks. The conflict coordination module is used to resolve conflicts between the agents and ensure the coordinated operation of the system. This includes conflict detection, which detects conflicts between agents in real time; and conflict resolution, which resolves conflicts according to preset rules and strategies. A consultation mechanism is used to initiate a consultation process, allowing relevant intelligent agents to negotiate. The module defining communication protocols and negotiation mechanisms between agents is used to define the communication protocols and negotiation mechanisms between agents, ensuring that the information exchange and negotiation process between agents is efficient and orderly. The information integration and presentation module is used to integrate the information provided by various intelligent agents and present it to the bridge manager through a human-computer interaction interface. The feedback and self-learning module is used to reassess the structural health index of the bridge based on the new data collected by the monitoring agent. Using the new monitoring data and the reassessed structural health index, the diagnostic model of the diagnostic agent is updated and optimized online. The control decision agent updates and optimizes its decision strategy model online based on the new structural health index and the new diagnostic report, forming a "decision-execution-feedback" closed loop to achieve collaborative optimization of bridge maintenance decisions.

7. A bridge maintenance decision-making system based on multi-agent cooperative optimization according to any one of claims 1-3, characterized in that: The structural health index is calculated using equations (1) to (3): Where i represents the damage type; N represents the total number of damages; C i The contribution of the i-th injury to overall health status; w i Let D be the weight of the i-th injury, based on the severity D of the i-th injury. i and positional importance L i Confirmed; w total D is the sum of all damage weights; i This represents the severity of the i-th damage (e.g., crack width, strain value, etc.); 1-D i To represent the health contribution of the injury, let's assume D. i The value is between [0,1], where 0 represents no damage and 1 represents complete damage; Assume weight w i With positional importance L i Proportional; directly using positional importance as the weight, i.e., w i =L i Then the total weight w total This can be expressed by equation (4):

8. A bridge maintenance decision-making system based on multi-agent cooperative optimization according to any one of claims 1-3, characterized in that: The multi-objective maintenance utility function is expressed by equation (5): max(U total )=w1·f s (S)+w2·f c (S)+w3·f t (S)+w4·f l (S) (5) Among them, Ut otal The comprehensive objective function value represents the overall performance of the maintenance scheme; max(U total ) represents the maximized overall utility; S represents a specific maintenance plan; f s (S) represents the expected safety utility function after the implementation of scheme S, which is positively correlated with the improvement of the structural health index (BHI); f c (S) is the cost-utility function of option S, usually negative, and negatively correlated with maintenance costs; f t (S) is the utility function of the traffic impact of scheme S, which is usually negative and negatively correlated with the traffic interruption time and range; f l (S) is the utility function of the impact of scheme S on the remaining life of the bridge; w1, w2, w3, and w4 are respectively the utility function of f s (S), f c (S), f t (S), f l (S) The weight coefficients of each utility function, and w1+w2+w3+w4=1.

9. A bridge maintenance decision-making system based on multi-agent collaborative optimization, characterized in that, The bridge maintenance decision-making system based on multi-agent cooperative optimization as described in any one of claims 1-8 is implemented, comprising: S1: Data Acquisition and Preprocessing. The monitoring agent continuously collects structural and environmental data of the bridge, performs noise reduction and normalization on the collected data, marks abnormal data, and sends it to the coordination / communication agent. S2: The coordination / communication agent forwards the abnormal data preprocessed in step S1 to the diagnostic agent. The diagnostic agent uses its built-in knowledge base, machine learning model and physical model to assess the health status of the bridge, identify the damage type, location and severity, output the structural health index, generate a diagnostic report, and send the diagnostic report to the coordination / communication agent. S3: The coordination / communication agent distributes the diagnostic report to the decision agent, which generates one or more candidate maintenance scheme sets according to a preset multi-objective maintenance utility function and submits them to the coordination / communication agent; S4: The coordination / communication agent distributes the candidate maintenance scheme set to the resource scheduling agent, and initiates a negotiation process between the resource scheduling agent and the decision-making agent according to a predetermined negotiation mechanism. The resource scheduling agent evaluates the resource feasibility of each candidate maintenance scheme according to the negotiation instructions issued by the coordination / communication agent, and gives a corresponding resource matching score. The decision-making agent calculates the comprehensive utility of each candidate maintenance scheme, and combines the comprehensive utility of each candidate maintenance scheme with the resource matching score, and uses an optimization algorithm to select the optimal maintenance scheme. S5: The resource scheduling agent, based on the optimal maintenance scheme selected by the decision-making agent, uses optimization algorithms to coordinate and plan the human, material, equipment, and time window resources required to execute the optimal maintenance scheme, generates a specific maintenance construction plan, and submits the construction plan to the coordination / communication agent; S6: The coordination / communication agent integrates the diagnostic report, the optimal maintenance plan, and the maintenance construction plan, and presents them to the bridge manager through a human-computer interaction interface; the bridge manager can view detailed information, adjust task priorities, and approve maintenance plans through the human-computer interaction interface. S7: After maintenance is implemented, the coordination / communication agent reassesses the structural health index of the bridge based on the new data collected by the monitoring agent. Using the new monitoring data and the reassessed structural health index, the diagnostic model of the diagnostic agent is updated and optimized online. The control decision agent updates and optimizes its decision-making strategy model online based on the new structural health index and the new diagnostic report, forming a "decision-execution-feedback" closed loop. Through continuous feedback and self-learning, the collaborative optimization of bridge maintenance decisions is achieved.

10. An electronic device, characterized in that, It includes a processor and a memory, which are interconnected; The memory is used to store computer programs; The processor is configured to execute, when the computer program is invoked, the bridge maintenance decision-making method based on multi-agent cooperative optimization as described in any one of claims 1 to 8.

Images