Intelligent bridge monitoring method and system based on Multi-Agent architecture, medium and processor

Abstract

The invention discloses an intelligent bridge monitoring method and system based on Multi-Agent architecture, a medium and a processor, and relates to the technical field of intelligent bridge monitoring. According to the method, a bridge state and environment monitoring Agent collects multi-dimensional environment data (such as water flow velocity, temperature, air temperature and humidity) in real time, a traffic flow Agent counts traffic load data by using an image processing technology, a diagnosis interaction Agent carries out fusion analysis on the data, an environment influence coefficient gamma is calculated, and a real-time monitoring result is obtained. And predicting the residual life of the key component in combination with material fatigue life data. The system adopts a distributed task scheduling architecture, realizes multi-Agent efficient cooperation, and avoids decision conflicts. Through data fusion and an intelligent early warning mechanism, the problem of data islanding of a traditional monitoring system is solved, the real-time performance, the accuracy and the intelligent level of bridge monitoring are remarkably improved, and reliable support is provided for bridge safety management.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent bridge monitoring, and in particular to a method, system, medium and processor for intelligent bridge monitoring based on a Multi-Agent architecture. Background Art

[0002] Due to the aging of transportation infrastructure and the frequent occurrence of extreme climate events, the shortcomings of traditional bridge monitoring methods have become increasingly prominent. They are unable to meet the needs of real-time, accurate, and intelligent management, and are in urgent need of technological innovation. Foreign countries started early in the research and early application of cutting-edge algorithms for intelligent monitoring systems, and have a certain amount of technological accumulation and leading advantages. However, the current domestic intelligent bridge monitoring system does not have in-depth research on the collaborative optimization mechanism between multiple agents (intelligent bodies). In complex bridge monitoring tasks, multiple agents may encounter problems such as collaborative conflicts, making it difficult to achieve efficient linkage. Although AI algorithms are constantly innovating, in the field of bridge monitoring, the interpretability of the model and its integration with bridge professional knowledge are still insufficient. Therefore, the domestic bridge monitoring system is in urgent need of technological upgrades.

[0003] In the research of multi-agent interactive intelligent bridge monitoring systems, how to enable multiple agents to achieve efficient collaboration, avoid conflicts, and achieve optimal resource allocation through innovative collaborative mechanisms in complex monitoring tasks is a key research direction for the future. How to make the prediction and diagnostic results of AI models more reliable and explainable remains a difficult problem that requires further exploration. Therefore, in the field of bridge monitoring, the deep integration of multi-agent and AI technologies is an inevitable trend in future development.

[0004] In view of this, a bridge intelligent monitoring method, system, medium and processor based on Multi-Agent architecture are needed. Summary of the Invention

[0005] In response to the problems of multiple agents potentially encountering conflicts in collaboration and difficulty in achieving efficient linkage in existing technologies, the present invention provides a bridge intelligent monitoring method, system, medium, and processor based on a multi-agent architecture. This method can achieve efficient collaboration among multiple agents through data fusion and bridge monitoring calculations, thus enabling efficient and intelligent monitoring of bridges. The specific technical solution is as follows:

[0006] A bridge intelligent monitoring method based on a Multi-Agent architecture includes:

[0007] S1: Bridge status and environment monitoring agent collects multi-dimensional environmental data of the bridge in real time;

[0008] S2: Traffic flow agent monitors and collects statistics on bridge traffic flow in real time to obtain traffic load data;

[0009] S3: Write multi-dimensional environmental data and traffic load data into the MySQL database for access by the diagnostic interaction agent;

[0010] S4: The diagnostic interaction agent performs a fusion analysis on multi-dimensional environmental data and traffic load data to obtain the environmental impact coefficient γ;

[0011] S5: Obtain the cumulative fatigue damage data of key bridge components based on traffic load data and material fatigue life data corresponding to the load;

[0012] S6: Predict the remaining life of key components based on their cumulative fatigue damage data and environmental impact coefficient γ, and immediately issue an alarm for abnormal data.

[0013] Furthermore, the multi-dimensional environmental data includes water flow velocity, water flow temperature, air temperature, and air humidity.

[0014] Furthermore, in step S2, the traffic flow agent performs real-time monitoring and statistics on the bridge traffic flow to obtain traffic load data, including the following steps:

[0015] S21: Traffic flow agent collects video images through cameras installed at appropriate locations on the bridge;

[0016] S22: Processing the collected video image to extract the vehicle target;

[0017] S23: Based on the extracted vehicle target, perform vehicle type identification to obtain the load type;

[0018] S24: Count the number of times each load type passes through the bridge within a certain period of time to obtain the number of times it acts on the bridge.

[0019] Furthermore, in step S23, the vehicle type identification is performed based on the extracted vehicle target to obtain the load type, including the following steps:

[0020] Normalize the extracted vehicle target images and adjust them to a fixed size;

[0021] Use a pre-trained CNN model (such as ResNet50) to extract the feature vector f of the vehicle image;

[0022] f=[f1,f2,…,f d ];

[0023] Among them, f is the eigenvector; f i is the eigenvalue of the i-th dimension;

[0024] The Softmax classifier is used to classify the feature vector, and the probability distribution function P(y=c|f) is obtained as follows:

[0025]

[0026] Where C is the total number of vehicle types (such as trucks, cars, etc.), w c and b c is the weight and bias of class c;

[0027] Using the cross entropy loss function Optimize model parameters:

[0028]

[0029] Among them, y c One-hot encoding of the true label

[0030] Update model parameters using stochastic gradient descent or Adam optimizer;

[0031] According to the probability distribution of the classifier output, select the category y with the highest probability pred As the final recognition result:

[0032]

[0033] Furthermore, in step S4, the diagnostic interaction agent performs a fusion analysis on the multi-dimensional environmental data and traffic load data to obtain the environmental impact coefficient γ, which includes the following steps:

[0034] S41: Use linear interpolation to align the data of the same sensor. The formula is as follows:

[0035]

[0036] x(t) represents the data value after alignment at time point t; t is the time point of target alignment; t i and t i+1 : represent two original data time points adjacent to t, and satisfy t i ≤t≤t i+1 ;x(t i ) and x(t i+1 ) represent the time point t i and t i+1 The original data value collected at

[0037] S42: Fusion calculation of environmental impact coefficient γ, the formula is as follows:

[0038]

[0039] In the above formula, E env is the environmental factor; x i is the i-th environmental data; x max ,x min is the maximum and minimum value in the historical data of environmental data; n is the total number of environmental data dimensions; L traffic is the traffic load factor; n j is the number of times the j-th type of vehicle or load acts; k j is the vehicle type weight; N max is the maximum theoretical number of passes within the statistical period; w1 and w2 are the weights of environmental and traffic loads; α and β are the scaling coefficients and offsets.

[0040] Furthermore, in step S5, the cumulative fatigue damage calculation formula is as follows:

[0041]

[0042] In the above formula, D is the cumulative fatigue damage; n i is the actual number of actions of the i-th type load; M i is the material fatigue life corresponding to the i-th type of load.

[0043] Furthermore, in step S6, the formula for predicting the remaining life of the key components is as follows:

[0044] T remaining =T design ×(1-D)×γ;

[0045] In the above formula, T remaining Remaining life of component; T design is the design life of the component; D is the cumulative fatigue damage; γ is the environmental impact coefficient.

[0046] A bridge intelligent monitoring system based on a Multi-Agent architecture is applied to the above-mentioned bridge intelligent monitoring method based on a Multi-Agent architecture, comprising:

[0047] Bridge status and environment monitoring agent, which is used to collect multi-dimensional environmental data of bridges in real time;

[0048] Traffic flow agent, which is used to monitor and count bridge traffic flow in real time and obtain traffic load data;

[0049] The transmission module is used to write multi-dimensional environmental data and traffic load data into the MySQL database for access by the diagnostic interaction agent;

[0050] The diagnostic interaction agent is used to integrate and analyze multi-dimensional environmental data and traffic load data to obtain the environmental impact coefficient γ; the cumulative fatigue damage data of key bridge components is obtained based on the traffic load data and the material fatigue life data corresponding to the load; the remaining life of key components is predicted based on the cumulative fatigue damage data of key components and the environmental impact coefficient γ, and an alarm will be issued immediately for abnormal data.

[0051] A computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned bridge intelligent monitoring method based on the Multi-Agent architecture.

[0052] A processor is used to run a program, wherein the program executes the above-mentioned bridge intelligent monitoring method based on the Multi-Agent architecture when running.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] 1. Enable multiple agents to collaborate through data fusion and conduct efficient and intelligent monitoring of bridges.

[0055] 2. This application solution adopts a distributed task scheduling architecture and a two-layer structure of "central diagnosis agent + distributed field agent". That is, the bridge status and environment monitoring agent and the traffic flow agent are responsible for information monitoring in different fields respectively. The decision-making diagnosis function is only completed by the diagnosis interaction agent, avoiding decision conflicts and resource competition conflicts.

[0056] 3. The Diagnostic Interaction Agent coordinates the real-time data collected by the Environmental Monitoring Agent and the Traffic Flow Agent, using data analysis algorithms and threshold judgment mechanisms to analyze the collected data in real time. If abnormal data is detected, an alarm is immediately sent to bridge maintenance personnel. Based on a pre-set risk classification table, the agent automatically assesses the risk level and generates recommendations and measures, providing timely and effective support for bridge safety management.

[0057] 4. This solution adopts a collaborative "hardware + communication + data interaction" architecture, using Arduino as the core control board to connect various sensors, and using a data acquisition card to accurately acquire data at the hardware level. Through serial communication and the FastAPI framework, an efficient data transmission channel is constructed. Function calling technology is introduced to achieve intelligent interaction between the agent and the MySQL database, optimize the data processing process, and improve system operation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly describes the drawings required for the specific embodiments or the description of the prior art. Similar elements or parts are generally identified by similar reference numerals throughout the drawings. Elements or parts in the drawings are not necessarily drawn to scale.

[0059] Figure 1 This is a flow chart of a bridge intelligent monitoring method based on Multi-Agent architecture;

[0060] Figure 2 This is a structural diagram of a bridge intelligent monitoring system based on Multi-Agent architecture. DETAILED DESCRIPTION

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0062] It should be understood that when used in this application, the terms "include" and "comprising" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0063] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0064] It should be further understood that the term "and / or" used in this application refers to and includes any and all possible combinations of one or more of the associated listed items.

[0065] Example 1

[0066] Traditional bridge monitoring is often limited to the structural data of the bridge itself, but this project extends the monitoring scope to include the surrounding environment and traffic flow factors.

[0067] like Figure 1The figure shows a flow chart of an intelligent bridge monitoring method based on a multi-agent system architecture. This multi-agent system utilizes a distributed computing model and features multiple agents with independent computing and decision-making capabilities, including a bridge status and environmental monitoring agent, a traffic flow agent, and a diagnostic interaction agent. These agents can collaborate to achieve intelligent decision-making through information sharing and interaction. This architecture enhances the system's fault tolerance and reliability. If an agent fails, other agents can continue to operate, ensuring overall system stability. The method specifically includes the following steps:

[0068] S1: The bridge status and environment monitoring agent deploys high-precision temperature sensors and temperature and humidity composite sensors to collect multi-dimensional environmental data such as water flow velocity, water flow temperature, air temperature, and air humidity under the bridge in real time.

[0069] S2: Traffic flow agent uses OpenCV image processing technology to monitor and count bridge traffic flow in real time and obtain traffic load data.

[0070] Furthermore, the traffic load data includes load type and the corresponding number of times the load acts on the bridge.

[0071] S21: Traffic flow agent collects video images through cameras installed at appropriate locations on the bridge.

[0072] Select a suitable camera installation location to ensure that vehicles crossing the bridge can be clearly captured.

[0073] The acquired video images will serve as the basis for subsequent image processing and analysis.

[0074] S22: Use OpenCV image processing technology, combined with the Haar cascade detector technology principle, to process the collected video images and extract vehicle targets.

[0075] The Haar cascade detector uses Haar features to quickly extract the grayscale differences in the target area, and uses the cascade structure to filter non-target areas layer by layer, significantly improving detection efficiency and accuracy.

[0076] The integral image technology reduces the grayscale and computational complexity of any rectangular area to a constant level, making real-time calculation of Haar features possible.

[0077] The AdaBoost algorithm trains cascade classifiers, combining multiple weak classifiers according to weights to form a high-precision strong classifier, achieving "high speed + high recall".

[0078] Furthermore, when processing the collected video images, Gaussian filtering, noise reduction, target tracking, and counting logic algorithms were used to enhance the stability and accuracy of traffic flow monitoring.

[0079] 1. Gaussian filtering denoising can effectively remove noise from images, improve image quality, and thus enhance the accuracy of subsequent target detection.

[0080]

[0081] In the above formula, G(x,y) is the weight of the Gaussian filter at the pixel point (x,y), where x and y are the coordinate offsets of the current pixel relative to the filter center. σ is the standard deviation, which controls the smoothness of the filter. A larger value blurs the image (stronger denoising, but with greater loss of detail). The core purpose of calculating the filter weight at a pixel point (i.e., the Gaussian kernel function value G(x,y)) is to achieve image smoothing and denoising. By weighted averaging the values ​​of neighboring pixels, image noise is suppressed, improving the accuracy of subsequent object detection.

[0082] 2. Kalman filter prediction is used for target tracking. It estimates and predicts the position and velocity of the vehicle through state equations and observation equations, which can better handle the motion changes of the target and improve the stability and accuracy of tracking.

[0083] Equation of state:

[0084] In the above formula, is the state prediction value (vehicle position and speed) at the kth moment; x k is the state vector, defined as [p x ,p y ,v x ,v y ] T , where: p x ,p y is the two-dimensional coordinate of the vehicle in the image; v x ,v y is the vehicle's speed in the x and y directions; A is the state transfer matrix, describing the vehicle's motion model (such as uniform motion); B,u k is the control input matrix and control amount (may not be used in this scenario, set to 0); w k is the process noise and represents the prediction error.

[0085] Observation equation: z k =Hx k +v k , where x k =[p x ,p y ,v x ,vy ] T is the vehicle position and speed;

[0086] z k Observation value in (such as the actual detected vehicle position); H is the observation matrix, which maps the state to the observation space; v k is the observation noise, which represents the sensor error.

[0087] Through the prediction-correction mechanism, the vehicle trajectory is stabilized and the target loss or jitter is reduced.

[0088] 3. Counting logic algorithms are used to accurately count the number of vehicles and ensure the accuracy of traffic flow data.

[0089]

[0090] is the vertical coordinate of the vehicle in the t-th frame image (such as the position of the bottom of the vehicle); y0 is the preset virtual detection line (such as the horizontal line drawn in the middle of the image); Count is the cumulative number of vehicles passing.

[0091] When a vehicle moves from one side of the detection line to the other (such as when the y coordinate crosses y0), counting is triggered to avoid duplication or missed detection.

[0092] S23: Based on the extracted vehicle target, the vehicle type is identified to obtain the load type.

[0093] Use machine learning or deep learning methods to classify the extracted vehicle targets and identify different types of vehicles, such as trucks, cars, etc.

[0094] Vehicle type recognition can be performed based on features such as the vehicle's shape, size, and color.

[0095] 1. Normalize the extracted vehicle target image and adjust it to a fixed size (e.g., 224x224 pixels).

[0096] The image is normalized using the following formula:

[0097]

[0098] Where I is the original image pixel value, μ is the mean, and σ is the standard deviation; I norm is the normalized image pixel value.

[0099] 2. Use a pre-trained CNN model (such as ResNet50) to extract feature vectors of vehicle images.

[0100] The dimension of the feature vector is d, which is expressed as:

[0101] f=[f1,f2,…,f d ];

[0102] Among them, f is the eigenvector, f i is the eigenvalue of the i-th dimension.

[0103] 3. Use the Softmax classifier to classify the feature vector. The probability distribution function P(y=c|f) is as follows:

[0104]

[0105] Where C is the total number of vehicle types (such as trucks, cars, etc.), w c and b c are the weights and biases for class c.

[0106] 4. Use cross entropy loss function Optimize model parameters:

[0107]

[0108] Among them, y c is the one-hot encoding of the true label.

[0109] 5. Training and Optimization

[0110] Stochastic gradient descent (SGD) or Adam optimizer is used to update model parameters.

[0111] Introducing data enhancement techniques (such as rotation and flipping) to improve the generalization ability of the model.

[0112] 6. Vehicle type determination

[0113] According to the probability distribution output by the classifier, the category with the highest probability is selected as the final recognition result:

[0114]

[0115] S24: Count the number of times each load type passes through the bridge within a certain period of time to obtain the number of times it acts on the bridge.

[0116] For each identified vehicle type, count the number of times it passes the bridge within a certain period of time.

[0117] The number of times traffic impacts a bridge can be used as part of traffic load data, providing reliable data support for bridge management and maintenance.

[0118] S3: The multi-dimensional environmental data and traffic load data collected above are transmitted to the terminal via wireless transmission (such as LoRa, 4G / NB-IoT) and written into the MySQL database through Function calling technology. FastAPI is used to build an API for access by the diagnostic interaction agent.

[0119] Furthermore, the data collected by this solution can meet the needs of industrial-grade long-distance and anti-interference communication through wireless transmission (such as LoRa, 4G / NB-IoT).

[0120] S4: Perform fusion analysis on multi-dimensional environmental data and traffic load data to obtain the environmental impact coefficient γ, so as to improve the timeliness and reliability of the results in subsequent further analysis and avoid data consistency conflicts.

[0121] The diagnostic interaction agent uses function calling technology to access the API to obtain multi-dimensional data in the MySQL database, and accurately writes the collected multi-dimensional environmental data and traffic load data.

[0122] Furthermore, when fusion analysis of multi-dimensional environmental data and traffic load data is performed to obtain the environmental impact coefficient γ, multiple technologies including data spatiotemporal alignment and feature-level fusion are used to ensure the feasibility of multi-dimensional data fusion analysis and avoid data consistency conflicts. Specifically, the following steps are included:

[0123] S41: Use linear interpolation to align the data of the same sensor. The formula is as follows:

[0124]

[0125] x(t) represents the data value after alignment at time point t (i.e., interpolation result); t is the time point of target alignment; t i and t i+1 : represent two original data time points adjacent to t, and satisfy t i ≤t≤t i+1 ;x(t i ) and x(t i+1 ) represent the time point t i and t i+1 The original data value collected at .

[0126] S42: Fusion calculation of environmental impact coefficient γ, the formula is as follows:

[0127]

[0128] In the above formula, E envis the environmental factor, which is obtained by normalizing the environmental data such as air temperature and humidity, water flow velocity, and water flow temperature; i is the i-th environmental data (such as air temperature, air humidity, water flow velocity, water flow temperature); x max ,x min is the maximum and minimum value in the historical data of environmental data; n is the total number of environmental data dimensions (such as air temperature, air humidity, water flow speed, water flow temperature); L traffic is the traffic load factor, calculated based on vehicle type and number of actions; n j k is the number of times the jth type of vehicle or load (such as trucks, cars) acts; j is the vehicle type weight (truck>sedan); N max is the maximum theoretical number of passes within the statistical period; w1, w2 are the weights of the environment and traffic loads (the default is 0.5 each, which can be adjusted); α, β are the scaling coefficients and offsets (such as α = 0.4, β = 0.8), ensuring that γ ranges from 0.8 to 1.2, which meets the subsequent remaining life prediction requirements.

[0129] S5: Obtain the cumulative fatigue damage data of key bridge components based on traffic load data and material fatigue life data corresponding to the load.

[0130] Furthermore, the material fatigue life corresponding to the load refers to the life of the bridge material corresponding to the traffic load type, such as the life corresponding to a truck or the life corresponding to a bridge vehicle, etc.

[0131] Furthermore, the cumulative fatigue damage (Miner linear cumulative damage law) is calculated as follows:

[0132]

[0133] Parameter description: D is the cumulative fatigue damage (when D>1, it is determined that the bridge structure has reached the fatigue failure threshold); n i is the actual number of times the i-th type of load (such as trucks, cars, etc.) acts, which is obtained by the traffic flow agent through OpenCV image recognition statistics; M i is the material fatigue life (times) corresponding to the i-th type of load, which is stored in the bridge knowledge base of the diagnostic interaction agent and is related to the material type and environmental parameters (temperature, humidity, corrosion factor).

[0134] S6: Predict the remaining life of key components based on their cumulative fatigue damage data and environmental impact coefficient γ, and immediately issue an alarm for abnormal data.

[0135] Furthermore, the formula for predicting the remaining life of the key components is as follows:

[0136] T remaining=T design ×(1-D)×γ;

[0137] In the above formula, T remaining Remaining life of component (years); T design is the component design life (years), taken from the bridge design document (built-in knowledge base); D is the cumulative fatigue damage; γ is the environmental impact coefficient (range 0.8-1.2), which is dynamically adjusted according to the temperature, humidity, water erosion and other data collected by the environmental monitoring agent.

[0138] Example 2

[0139] In this project, we chose the Qwen3 language model as the core foundation for building a multi-agent collaborative system. There are other large language models with similar performance, or even better performance in specific areas, such as the GPT series of models and the LLaMA series of models.

[0140] When building the intelligent environmental monitoring agent, we used high-precision temperature sensors and combined temperature and humidity sensors, such as the DS18B20 and DHT11. If you encounter other sensors with better performance (such as those with higher measurement accuracy, stronger anti-interference capabilities, or better suited to complex environments such as bridges), you can consider using them instead. This will further improve the reliability and stability of data collection.

[0141] Currently, the system primarily uses wireless transmission (LoRa, 4G / NB-IoT) combined with FastAPI to transmit data. However, in certain scenarios, a hybrid wired and wireless transmission method is employed, using wired transmission at key nodes while maintaining wireless transmission elsewhere, balancing stability and flexibility. This allows the transmission method to be adjusted based on actual needs, ensuring both reliable and efficient data transmission.

[0142] Example 3

[0143] like Figure 2 As shown, a bridge intelligent monitoring system based on a Multi-Agent architecture is applied to the above-mentioned bridge intelligent monitoring method based on a Multi-Agent architecture, including:

[0144] Bridge status and environment monitoring agent, which is used to collect multi-dimensional environmental data of bridges in real time;

[0145] Traffic flow agent, which is used to monitor and count bridge traffic flow in real time and obtain traffic load data;

[0146] The transmission module is used to write multi-dimensional environmental data and traffic load data into the MySQL database for access by the diagnostic interaction agent.

[0147] The diagnostic interaction agent is used to integrate and analyze multi-dimensional environmental data and traffic load data to obtain the environmental impact coefficient γ; the cumulative fatigue damage data of key bridge components is obtained based on the traffic load data and the material fatigue life data corresponding to the load; the remaining life of key components is predicted based on the cumulative fatigue damage data of key components and the environmental impact coefficient γ, and an alarm will be issued immediately for abnormal data.

[0148] Collaboration and task priority of each agent:

[0149] (1) Task Scheduling

[0150] Task allocation process = publish tasks → Agents provide data → Coordinator selects tasks based on task keywords

[0151] Steps: The diagnostic interaction agent acts as a coordinator and issues tasks; the environmental monitoring agent and traffic flow agent each submit a response plan; the coordinator selects the execution agent based on response time and task matching to avoid parallel conflicts.

[0152] (2) Resource access: Task priority

[0153] Resource access order = critical tasks (priority P1) > real-time data tasks (P2) > historical query tasks (P3)

[0154] P1: The operator issues tasks, triggers key tasks (abnormal alarms), etc.; P2: Real-time data collection and other tasks; P3: Daily data report generation and other tasks.

[0155] Example 4

[0156] A computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned bridge intelligent monitoring method based on the Multi-Agent architecture.

[0157] Example 5

[0158] A processor is used to run a program, wherein the program executes the above-mentioned bridge intelligent monitoring method based on the Multi-Agent architecture when running.

[0159] Beneficial effects:

[0160] 1. This application solution adopts a distributed task scheduling architecture and a two-layer structure of "central diagnosis agent + distributed field agent". That is, the bridge status and environment monitoring agent and the traffic flow agent are responsible for information monitoring in different fields respectively. The decision-making and diagnosis function is only completed by the diagnosis interaction agent, avoiding decision conflicts and resource competition conflicts.

[0161] 2. The Diagnostic Interaction Agent coordinates the real-time data collected by the Environmental Monitoring Agent and the Traffic Flow Agent, analyzing it in real time using data analysis algorithms and threshold judgment mechanisms. If abnormal data is detected, an alert is immediately sent to bridge maintenance personnel. Based on a pre-set risk classification table, the agent automatically assesses the risk level and generates recommendations and measures, providing timely and effective support for bridge safety management.

[0162] 3. This solution adopts a collaborative "hardware + communication + data interaction" architecture, using Arduino as the core control board to connect various sensors, and using a data acquisition card to accurately acquire data at the hardware level. Through serial communication and the FastAPI framework, an efficient data transmission channel is constructed. Function calling technology is introduced to achieve intelligent interaction between the agent and the MySQL database, optimize the data processing process, and improve system operation efficiency.

[0163] 4. This application leverages the Qwen3 large-scale model to construct a bridge knowledge base and question-and-answer engine. When non-professional users query using natural language, the knowledge base retrieval function is invoked to provide users with professional bridge status information and maintenance recommendations. Furthermore, FastAPI is used to build an API. Leveraging ASGI's asynchronous processing capabilities, agents can retrieve data via HTTP requests, enabling real-time multi-agent interaction. For data storage, agents utilize function calling technology to access a MySQL database and accurately write collected data. This approach lowers the barrier to information acquisition, making it easy for even non-professional users to understand bridge status.

[0164] 5. The present invention provides an intelligent system for bridge monitoring. Through distributed agent intelligent collaboration, it can establish a multi-dimensional perception network across the bridge's hydrological environment (water temperature), atmospheric environment (temperature and humidity), and traffic load, quantify the cumulative fatigue damage of the bridge, and predict the remaining life of key components, thereby solving the "data islanding" problem of traditional systems.

[0165] 6. Data fusion and high-frequency data collection mechanism to improve reliability: Extending the monitoring scope to include surrounding environmental factors and traffic flow, simultaneously collecting environmental information such as water temperature, air temperature and humidity under the bridge, and traffic flow information, to achieve multi-dimensional data fusion analysis and improve the timeliness and reliability of the results.

[0166] 7. The technical solution of this application has made innovations in data fusion, architecture design, early warning mechanism, technical collaboration, traffic flow monitoring and intelligent diagnosis. Compared with traditional solutions, it has significant advantages and can effectively improve the intelligence level of bridge monitoring.

[0167] 8. The Multi-Agent architecture enhances the system's fault tolerance: Using a distributed computing model, each agent possesses independent computing and decision-making capabilities. When faced with complex bridge monitoring data, multiple agents can work collaboratively, sharing information and interacting to achieve intelligent decision-making.

[0168] 9. Intelligent early warning and emergency response to improve equipment stability and safety: When monitoring data shows anomalies, the system automatically triggers the early warning mechanism. Through collaborative work between agents, it automatically assesses the risk level, generates suggestions and measures, and provides timely and effective support for bridge safety management.

[0169] The present application discloses a bridge intelligent monitoring method, system, medium and processor based on Multi-Agent architecture, which relates to the field of bridge intelligent monitoring technology. The method collects multi-dimensional environmental data (such as water velocity, temperature, air temperature and humidity, etc.) in real time through the bridge status and environment monitoring agent, the traffic flow agent uses image processing technology to statistically calculate traffic load data, and the diagnostic interaction agent performs fusion analysis on the data, calculates the environmental impact coefficient γ, and combines the material fatigue life data to predict the remaining life of key components. The system adopts a distributed task scheduling architecture to achieve efficient collaboration among multiple agents and avoid decision conflicts. The present invention solves the "data islanding" problem of traditional monitoring systems through data fusion and intelligent early warning mechanisms, significantly improves the real-time, accuracy and intelligence level of bridge monitoring, and provides reliable support for bridge safety management.

[0170] Those skilled in the art will appreciate that the units of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0171] In the embodiments provided by the present invention, it should be understood that the division of units is merely a logical function division, and there may be other division methods in actual implementation, for example, multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored, etc.

[0172] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0173] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nly Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc., various media that can store program code.

[0174] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of this application.

Claims

1. A bridge intelligent monitoring method based on Multi-Agent architecture, characterized in that: include: S1: Bridge status and environment monitoring agent collects multi-dimensional environmental data of the bridge in real time; S2: Traffic flow agent monitors and collects statistics on bridge traffic flow in real time to obtain traffic load data; S3: Write multi-dimensional environmental data and traffic load data into the MySQL database for access by the diagnostic interaction agent; S4: The diagnostic interaction agent performs a fusion analysis on multi-dimensional environmental data and traffic load data to obtain the environmental impact coefficient γ; S5: Obtain the cumulative fatigue damage data of key bridge components based on traffic load data and material fatigue life data corresponding to the load; S6: Predict the remaining life of key components based on their cumulative fatigue damage data and environmental impact coefficient γ, and immediately issue an alarm for abnormal data.

2. The bridge intelligent monitoring method based on Multi-Agent architecture according to claim 1 is characterized in that: The multi-dimensional environmental data includes water flow velocity, water flow temperature, air temperature, and air humidity.

3. The bridge intelligent monitoring method based on Multi-Agent architecture according to claim 2 is characterized in that: In step S2, the traffic flow agent monitors and collects statistics on the bridge traffic flow in real time to obtain traffic load data, which includes the following steps: S21: Traffic flow agent collects video images through cameras installed at appropriate locations on the bridge; S22: Processing the collected video image to extract the vehicle target; S23: Based on the extracted vehicle target, perform vehicle type identification to obtain the load type; S24: Count the number of times each load type passes through the bridge within a certain period of time to obtain the number of times it acts on the bridge.

4. The bridge intelligent monitoring method based on Multi-Agent architecture according to claim 3 is characterized in that: In step S23, the vehicle type is identified based on the extracted vehicle target to obtain the load type, including the following steps: Normalize the extracted vehicle target images and adjust them to a fixed size; Use the pre-trained CNN model to extract the feature vector f of the vehicle image; f=[f1,f2,…,f d ]; Among them, f is the eigenvector; f i is the eigenvalue of the i-th dimension; The Softmax classifier is used to classify the feature vector, and the probability distribution function P(y=c|f) is obtained as follows: Where C is the total number of vehicle types; w c and b c is the weight and bias of class c; Using the cross entropy loss function Optimize model parameters: Among them, y c One-hot encoding of the true label Update model parameters using stochastic gradient descent or Adam optimizer; According to the probability distribution of the classifier output, select the category y with the highest probability pred As the final recognition result:

5. The bridge intelligent monitoring method based on Multi-Agent architecture according to claim 1 is characterized in that: In step S4, the diagnostic interaction agent performs a fusion analysis on the multi-dimensional environmental data and traffic load data to obtain the environmental impact coefficient γ, which includes the following steps: S41: Use linear interpolation to align the data of the same sensor. The formula is as follows: x(t) represents the data value after alignment at time point t; t is the time point of target alignment; t i and t i+1 : represent two original data time points adjacent to t, and satisfy t i ≤t≤t i+1 ;x(t i ) and x(t i+1 ) represent the time point t i and t i+1 The original data value collected at S42: Fusion calculation of environmental impact coefficient γ, the formula is as follows: In the above formula, E env is the environmental factor; x i is the i-th environmental data; x max ,x min is the maximum and minimum value in the historical data of environmental data; n is the total number of environmental data dimensions; L traffic is the traffic load factor; n j is the number of times the j-th type of vehicle or load acts; k j is the vehicle type weight; N max is the maximum theoretical number of passes within the statistical period; w1 and w2 are the weights of environmental and traffic loads; α and β are the scaling coefficients and offsets.

6. The bridge intelligent monitoring method based on Multi-Agent architecture according to claim 1 is characterized in that: In step S5, the cumulative fatigue damage calculation formula is as follows: In the above formula, D is the cumulative fatigue damage; n i is the actual number of actions of the i-th type of load; M i is the material fatigue life corresponding to the i-th type of load.

7. The bridge intelligent monitoring method based on Multi-Agent architecture according to claim 1 is characterized in that: In step S6, the formula for predicting the remaining life of the key components is as follows: T remaining =T design ×(1-D)×γ; In the above formula, T remaining Remaining life of component; T design is the design life of the component; D is the cumulative fatigue damage; γ is the environmental impact coefficient.

8. A bridge intelligent monitoring system based on Multi-Agent architecture, characterized by: The bridge intelligent monitoring method based on the Multi-Agent architecture as claimed in any one of claims 1 to 7 comprises: Bridge status and environment monitoring agent, which is used to collect multi-dimensional environmental data of bridges in real time; Traffic flow agent, which is used to monitor and count bridge traffic flow in real time and obtain traffic load data; The transmission module is used to write multi-dimensional environmental data and traffic load data into the MySQL database for access by the diagnostic interaction agent; The diagnostic interaction agent is used to integrate and analyze multi-dimensional environmental data and traffic load data to obtain the environmental impact coefficient γ; the cumulative fatigue damage data of key bridge components is obtained based on the traffic load data and the material fatigue life data corresponding to the load; the remaining life of key components is predicted based on the cumulative fatigue damage data of key components and the environmental impact coefficient γ, and an alarm will be issued immediately for abnormal data.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the bridge intelligent monitoring method based on the Multi-Agent architecture according to any one of claims 1 to 7.

10. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes the bridge intelligent monitoring method based on the Multi-Agent architecture according to any one of claims 1 to 7.

Images