Road maintenance intelligent decision-making method based on multi-modal large model and knowledge base

Abstract

The present application relates to road maintenance technology, specifically to a road maintenance intelligent decision-making method based on a multi-modal large model and a knowledge base, comprising a pavement data processing module, a pavement damage identification module, a pavement performance evaluation module, a damage prediction and decision-making module; the present application dynamically calculates the weight coefficients of various damages by combining historical data and real-time detection information, and fuses multi-dimensional indexes such as driving quality, rut depth, anti-skid performance and structural strength to construct a pavement service performance index, breaking through the static and single nature of traditional pavement damage evaluation; by quantifying the correlation strength between damage types, triggering critical time and superposition coupling effect, using historical data to train the coupling evolution equation, the evolution trend and key development stage of various damages are predicted, targeted maintenance measures are started before damage deterioration, effectively blocking the chain triggering and coupling deterioration between damages, significantly prolonging the service life of the pavement, reducing the whole cycle maintenance cost, and improving the resource utilization efficiency.

Description

Technical Field

[0001] This invention relates to road maintenance technology, specifically to an intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base. Background Technology

[0002] Road maintenance is a crucial aspect of ensuring road traffic safety, improving road network service levels, and extending pavement lifespan. Traditional road maintenance decisions rely heavily on manual inspections, periodic testing, and experience-based judgment, which presents the following significant technical challenges: Current pavement damage assessments often use fixed weighting coefficients and evaluation standards, failing to fully consider the dynamic influence relationships between different damage types and their evolutionary characteristics with time, environment, and traffic load. This leads to a disconnect between assessment results and actual conditions, making it difficult to scientifically guide differentiated maintenance decisions. Existing assessment systems focus primarily on the Pavement Damage Index (PCI), lacking a systematic integration and comprehensive analysis of multi-dimensional performance indicators such as driving quality, structural strength, and skid resistance. This results in maintenance decisions that are biased towards "treating the symptoms," making it difficult to achieve overall improvement and long-term maintenance of pavement performance. Traditional maintenance models are mostly "post-incident repairs," lacking the ability to quantitatively model and predict the formation mechanism, coupling relationship between types, and evolution trend of road damage. They cannot take intervention measures before damage occurs or in the early stages of deterioration, resulting in high maintenance costs, serious waste of resources, and difficulty in stopping the chain development of damage. To address the aforementioned technical problems, this application proposes a solution. Summary of the Invention

[0003] The purpose of this invention is to solve the problems raised in the background art and to propose an intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base.

[0004] The objective of this invention can be achieved through the following technical solutions: A road maintenance intelligent decision-making method based on a multimodal large model and knowledge base includes a pavement data processing module, a pavement damage identification module, a pavement performance evaluation module, and a damage prediction and decision-making module. The road surface data processing module performs grayscale processing on the acquired images, divides and numbers them by pixel blocks, sets the normal fluctuation range by statistically analyzing the mean and standard deviation of grayscale values, identifies outliers and filters or re-detects them, and finally determines the image with the fewest abnormal grayscale blocks as the "final analysis image". The road surface damage identification module calculates the density of various types of damage and queries damage level data. Then calculate the road surface damage index. The weighting coefficients for various types of damage are dynamically determined by combining historical data. ; The road performance evaluation module calculates, including the ride quality index. rut depth index Anti-slip performance index Structural strength index The pavement performance index is calculated by combining multiple pavement performance indicators, including those with different weights. ; combination and Based on historical data, the remaining service life of the road surface is classified, and corresponding maintenance measures and priorities are determined according to the classification. The damage prediction and decision-making module uses historical data to train model parameters and predicts quantitative indicators for various damage types within a future timeframe. Based on the evolution trend, the system determines the sequence of damage formation and mutual triggering according to the prediction results, divides the evolution stages, and initiates targeted maintenance measures before key time nodes to prevent the coupled deterioration of damage and achieve intelligent maintenance decision-making.

[0005] In a preferred embodiment of the present invention, the specific steps for determining abnormal grayscale values ​​in the road surface data processing module include: P1: Perform grayscale processing on the real-time acquired image data, and segment the grayscale image according to the size of pixel blocks, dividing it into... The image data consists of several identical grayscale blocks, numbered according to their row and column numbers on the grayscale image. The acquired image data is sorted by acquisition time, and the corresponding numbered grayscale blocks within a single image acquired at the same time are further analyzed. Average the gray values and standard deviation The calculation, and the mean obtained from the calculation. and standard deviation Range of grayscale data The settings; P2: Compare the corresponding grayscale value data with the corresponding grayscale value fluctuation range, mark the corresponding grayscale value data that is outside the fluctuation range as outliers, and record the number of outliers. ;like If the grayscale data is abnormal, the grayscale data will be detected again. This is a preset proportional coefficient; if If outliers are removed, the remaining grayscale data after outlier removal is averaged. The calculation, and the mean obtained from the calculation. As the grayscale value data detected at the corresponding time; P3: If the grayscale data is still determined to be abnormal after re-detection, then the corresponding grayscale block number is determined to be abnormal.

[0006] In a preferred embodiment of the present invention, the specific steps for determining the "final analysis image" in the road data processing module include: Q1: After determining the grayscale values ​​of all numbered grayscale blocks in a grayscale image, how many grayscale blocks are abnormal? Record and take the quantity The smallest grayscale image is the preliminary analysis image; based on the grayscale features of the standard road surface structure pre-stored in the knowledge base, the grayscale values ​​of all grayscale blocks on the preliminary analysis image are compared to determine the corresponding grayscale block number, and these grayscale blocks are named structural grayscale blocks. Q2: What is the number of grayscale block anomalies in the structural grayscale blocks of the preliminary analysis image? Perform statistics and collect quantities. The smallest grayscale image is the final analysis image; the terminal of the road damage detection vehicle identifies the road condition of the final analysis image in real time, distinguishes the damage type, and synchronously stores the final analysis image, point cloud, location and size data of the corresponding damage type.

[0007] In a preferred embodiment of the present invention, the road surface damage identification module dynamically determines the weight coefficients for various types of damage. The methods include: W1: Calculate the total area data of the corresponding damage type identified in real time, divide it by the total area of ​​the detected road segment to obtain the damage density data of the corresponding damage type, and then retrieve the damage degree data of the corresponding damage type from the knowledge base based on the real-time acquired damage density. , The index corresponds to the type of damage; therefore, the road surface damage index is... ,in To detect the total number of different types of damage found within the road section, For the first Weighting coefficients for different types of damage; W2: Obtain historical data, acquire the pavement damage degree corresponding to the corresponding damage type in the historical data, and filter out the data group with only one damage type data fluctuation under different damage degrees. By comparing the fluctuation of the corresponding damage type with the damage degree in the data group, the change in damage degree caused by the unit fluctuation of the damage type can be obtained. W3: Calculate the total change in damage severity caused by unit fluctuations for each damage type. Use the proportion of the change in damage severity caused by unit fluctuations for the corresponding damage type to the total change in damage severity caused by unit fluctuations for all damage types as the weighting coefficient for the corresponding damage type. .

[0008] In a preferred embodiment of the present invention, the pavement performance evaluation module includes a pavement service performance index. Computational and decision fusion methods include: U1: Road surface smoothness index obtained from the road damage detection vehicle To obtain the driving quality index ; Obtain the rut depth at the location of the tire track on the road surface The rut depth index was obtained. Based on the obtained lateral force coefficient The anti-skid performance index of the road surface was calculated. Based on the measured rebound deflection value Initial deflection value and ultimate deflection value The structural strength index was calculated. ; corresponding road surface performance index Equal to the corresponding road surface , , and The sum of the products of the index and the corresponding index weight coefficients, divided by the sum of the corresponding index weight coefficients; U2: Get value corresponding Value, based on the corresponding historical data The remaining service life data of road surface damage is reclassified, and the two classification areas are compared. If the classification areas are the same, the classification area is determined to be normal; if the classification areas are different, the same area that does not require maintenance is retained, and the remaining area is classified into the next level. Based on the final classification result, recommended maintenance measures and their implementation priorities are matched from the knowledge base.

[0009] In a preferred embodiment of the present invention, the method for quantifying the correlation and coupling relationship between damage types in the damage prediction and decision module includes: G1: Correlation Strength ,in The total number of historical samples. This refers to the historical sample sequence number. For the first The weights of each sample, For indicator functions, For time window, For the first Damage type in each sample At any moment Quantitative indicators For the first Damage type in each sample At any moment Quantitative indicators and This is the trigger threshold value for the corresponding damage type; G2: When Types of damage over time Quantitative indicators And the correlation strength When determining the type of damage This will trigger the damage type. Triggering critical time , This is the current comprehensive environmental load factor. and These are the fitting parameters; G3: Coupling coefficient caused by mutual influence ,in and This refers to the natural deterioration rate when each damage type, categorized by pavement type and obtained from the knowledge base, exists alone, and the combined deterioration rate when two damage types coexist. , for Types of damage over time Quantitative indicators; based on the superposition coupling coefficient, the deterioration rate when damage types coexist is the same as when they exist independently. times.

[0010] In a preferred embodiment of the present invention, the damage prediction and decision-making module further includes a prediction and blocking decision-making method based on coupled evolution equations; H1: Damage Type Coupled evolution equations , The total number of coexisting damage types. Type of damage For types of damage The triggering growth rate, It is a step function; and In relation to the above analysis and The computational logic and physical meaning are completely identical, only the triggering direction is opposite; H2: Real-time input of quantitative indicators for each type of damage Combined coefficient of environmental load And import the data obtained from the large model. , and Solve the coupled evolution equations to obtain the future time. of sequence, For the prediction period; according to achieve The order of events determines the evolutionary sequence, and the corresponding quantitative indicators for the damage type are used. achieve When this type of damage is determined to have officially formed and begun to affect other types of damage; H3: The evolutionary stages are divided into three phases: single damage type, superposition of two damage types, and coupling of multiple damage types, according to the coupling evolution equation. Predict the evolution time of the corresponding stage, and before the corresponding evolution time is reached, initiate corresponding maintenance measures according to the damage type of the evolution to block the coupled evolution.

[0011] Compared with the prior art, the beneficial effects of the present invention are: By combining historical data with real-time detection information, the weighting coefficients of various types of damage are dynamically calculated. Furthermore, by integrating multi-dimensional indicators such as driving quality, rut depth, skid resistance, and structural strength, a pavement service performance index is constructed. This breaks through the static and singular nature of traditional pavement damage assessment, achieving dynamic and multi-dimensional comprehensive evaluation. By quantifying the correlation strength, triggering critical time, and superimposed coupling effect between damage types, and using historical data to train coupling evolution equations, the evolution trend and key development stages of various types of damage are predicted. Targeted maintenance measures are initiated before damage deteriorates, effectively preventing chain triggering and coupled deterioration between damage types, significantly extending pavement service life, reducing full-cycle maintenance costs, and improving resource utilization efficiency. 1. By integrating heterogeneous data from multiple sources such as images, point clouds, and localization, and through grayscale analysis, anomaly screening, and structural feature comparison, data quality and identification accuracy are improved. Relying on industry standards, historical cases, and expert experience stored in the knowledge base, and combining multimodal large models for reasoning and recommendation, the entire process from data collection and condition assessment to maintenance decision-making is automated and intelligent, significantly improving decision-making efficiency and operability. The constructed knowledge base and evolutionary prediction model have continuous learning and updating capabilities, and can continuously optimize parameters and strategies as detection data accumulates and the environment changes. Attached Figure Description

[0012] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0013] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0014] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Example: Please refer to Figure 1As shown, the intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base includes a pavement data processing module, a pavement damage identification module, a pavement performance evaluation module, and a damage prediction and decision-making module. The road surface data processing module performs grayscale processing on the acquired images, divides and numbers them by pixel blocks, sets the normal fluctuation range by statistically analyzing the mean and standard deviation of grayscale values, identifies outliers and filters or re-detects them, and finally determines the image with the fewest abnormal grayscale blocks as the "final analysis image". The road surface damage identification module calculates the density of various types of damage and queries damage level data. Then calculate the road surface damage index. The weighting coefficients for various types of damage are dynamically determined by combining historical data. ; The road performance evaluation module calculates, including the ride quality index. rut depth index Anti-slip performance index Structural strength index The pavement performance index is calculated by combining multiple pavement performance indicators, including those with different weights. ; combination and Based on historical data, the remaining service life of the road surface is classified, and corresponding maintenance measures and priorities are determined according to the classification. The damage prediction and decision-making module uses historical data to train model parameters and predicts quantitative indicators for various damage types within a future timeframe. Based on the prediction results, the formation and mutual triggering sequence of damage are determined, the evolution stages are divided, and targeted maintenance measures are initiated before the key time nodes to prevent the coupled deterioration of damage and realize intelligent maintenance decision-making. The road damage detection vehicle uses cameras to capture images of the road surface at a fixed frame rate, while LiDAR collects 3D point clouds of the road surface. A positioning system pinpoints the precise damaged sections (station numbers). The real-time image data undergoes grayscale processing, and the grayscale images are segmented according to pixel block size. The image data consists of several identical grayscale blocks, numbered according to their row and column numbers on the grayscale image. The acquired image data is sorted by acquisition time, and the corresponding numbered grayscale blocks within a single image acquired at the same time are further analyzed. Average the gray values and standard deviation The calculation, and the mean obtained from the calculation. and standard deviation Range of grayscale data The setting compares the corresponding grayscale value data with the corresponding grayscale value fluctuation range, marks the corresponding grayscale value data that is outside the fluctuation range as an outlier, and records the number of outliers. ; like If the grayscale data is abnormal, the grayscale data will be detected again. This is a preset proportional coefficient; if If outliers are removed, the remaining grayscale data after outlier removal is averaged. The calculation, and the mean obtained from the calculation. As the grayscale value data detected at the corresponding time; If the grayscale data is still determined to be abnormal upon re-inspection, then the corresponding numbered grayscale block is determined to be abnormal; after determining the grayscale values ​​of all numbered grayscale blocks on the grayscale image, the number of abnormal grayscale blocks is calculated. Record and take the quantity The smallest grayscale image is the preliminary analysis image. Based on the corresponding standard part's grayscale value, all grayscale blocks in the preliminary analysis image are compared to determine their corresponding numbers, and these grayscale blocks are named structural grayscale blocks. The number of grayscale block anomalies in the structural grayscale blocks of the preliminary analysis image is then analyzed. Perform statistics and collect quantities. The image with the smallest grayscale value is the final image for analysis; The terminal of the road damage detection vehicle performs real-time road condition identification on the final analysis image, distinguishes damage types, and synchronously stores the final analysis image, point cloud, location, and size data for the corresponding damage type. It calculates the total area data of the corresponding damage type based on the real-time identification and divides it by the total area of ​​the inspected road segment to obtain the damage density data for the corresponding damage type. Based on the real-time acquired damage density, it retrieves the damage degree data for the corresponding damage type from the "Road Damage Degree Assessment Standard Table" in the JTG5210 standard in the knowledge base. , The index corresponds to the type of damage; therefore, the road surface damage index is... ,in To detect the total number of different types of damage found within the road section, For the first Weighting coefficients for different types of damage; Historical data is acquired, and the corresponding pavement damage severity for each damage type is obtained from the historical data. Data sets with fluctuations in only one damage type under different damage severity levels are selected. By comparing the fluctuations of the corresponding damage type with the damage severity in this data set, the change in damage severity caused by a unit fluctuation of that damage type is determined. The total change in damage severity caused by a unit fluctuation of each damage type is calculated, and the proportion of the change in damage severity caused by a unit fluctuation of the corresponding damage type to the total change in damage severity caused by a unit fluctuation of all damage types is used as the weighting coefficient for the corresponding damage type. ; Obtain historical data and match the corresponding The data on road surface damage and remaining service life are acquired, and the corresponding data are processed according to the set service life intervals. Values ​​are categorized into levels; According to the road surface smoothness index obtained by the road damage detection vehicle To obtain the driving quality index , The minimum value is zero; obtain the rut depth at the location of the tire track on the road surface. Then when At that time, the rut depth index ;when At that time, the rut depth index Based on the obtained lateral force coefficient The anti-skid performance index of the road surface was calculated. ,when hour, ;when hour, Based on the measured rebound deflection value Initial deflection value and ultimate deflection value The structural strength index was calculated. ,when hour, ;when hour, ; In summary, the corresponding road surface performance index Equal to the corresponding road surface , , and The sum of the products of the index and its corresponding weight coefficient, divided by the sum of the corresponding weight coefficients, yields the result. value corresponding Value, based on the corresponding historical data The remaining service life data of road surface damage is used to classify the area again. The two classification areas are compared. If the classification areas are the same, the classification area is considered normal; if the classification areas are different, the same area that does not require maintenance is retained, and the remaining area is classified into the next level. The classification area is then compared with the corresponding... Value comparison determines the corresponding maintenance measures for the graded areas.

[0016] There are mutual influences among different types of road surface damage, depending on the type of damage. and types of damage The strength of the association is used to determine the strength of the association. ,in The total number of historical samples. This refers to the historical sample sequence number. For the first The weights of each sample, the indicator function When the conditions within the parentheses are met ,otherwise , For time window, For the first Damage type in each sample At any moment Quantitative indicators For the first Damage type in each sample At any moment Quantitative indicators and This is the trigger threshold value for the corresponding damage type; when Types of damage over time Quantitative indicators And the correlation strength When determining the type of damage This will trigger the damage type. Triggering critical time , This is the current comprehensive environmental load factor. and For fitting parameters; the superposition coupling coefficient caused by mutual influence. ,in and This refers to the natural deterioration rate when each damage type, categorized by pavement type and obtained from the knowledge base, exists alone, and the combined deterioration rate when two damage types coexist. , for Types of damage over time Quantitative indicators; based on the superposition coupling coefficient, the deterioration rate when damage types coexist is the same as when they exist independently. times; Acquire historical data and plot it on the damage type The quantitative indicators have reached the critical value Afterwards, type of damage Quantitative indicators Curves changing over time; types of damage Triggering time The time variable after the trigger is recorded as Damage type The growth of [these] typically conforms to the exponential growth model, i.e. , This is the comprehensive environmental load coefficient; for the model Taking the natural logarithm of both sides, we get... ,make , , , The model is simplified to a linear equation: The coefficients are obtained by training with historical sample data as input. and intercept Then the fitting parameters , ; Type of damage Coupled evolution equations , The total number of coexisting damage types. Type of damage For types of damage The triggering growth rate, step function exist hour Indicates the type of damage. Damage type triggered ;exist hour This indicates that it has not been triggered. and In relation to the above analysis and The computational logic and physical meaning are completely identical, only the triggering direction is opposite; Real-time input of quantitative indicators for each type of damage Combined coefficient of environmental load And import the data obtained from the large model. , and Solve the coupled evolution equations to obtain the future time. of sequence, For the prediction period; according to achieve The order of events determines the evolutionary sequence, and the corresponding quantitative indicators for the damage type are used. achieve When the damage type is determined to have formally formed and begun to affect other damage types, the evolutionary stages are divided into three stages: single damage type, superposition of two damage types, and coupling of multiple damage types, according to the coupling evolution equation. Predict the evolution time of the corresponding stage, and before the corresponding evolution time is reached, initiate corresponding maintenance measures according to the damage type of the evolution to block the coupled evolution.

[0017] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A road maintenance intelligent decision-making method based on a multimodal large model and knowledge base, characterized in that, It includes a road surface data processing module, a road surface damage identification module, a road surface performance evaluation module, and a damage prediction and decision-making module; The road data processing module converts the acquired images to grayscale, segments and numbers them by pixel blocks, sets the normal fluctuation range by statistically analyzing the mean and standard deviation of grayscale values, identifies outliers and filters or re-detects them, and finally determines the image with the fewest abnormal grayscale blocks as the "final analysis image". The road surface damage identification module calculates the density of various types of damage and queries damage level data. Then calculate the road surface damage index. The weighting coefficients for various types of damage are dynamically determined by combining historical data. ; The road performance evaluation module calculates, including the ride quality index. rut depth index Anti-slip performance index Structural strength index The pavement performance index is calculated by combining multiple pavement performance indicators, including those with different weights. ; combination and Based on historical data, the remaining service life of the road surface is classified, and corresponding maintenance measures and priorities are determined according to the classification. The damage prediction and decision-making module uses historical data to train model parameters and predicts quantitative indicators for various damage types within a future timeframe. Based on the evolution trend, the system determines the sequence of damage formation and mutual triggering according to the prediction results, divides the evolution stages, and initiates targeted maintenance measures before key time nodes to prevent the coupled deterioration of damage and achieve intelligent maintenance decision-making.

2. The intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base as described in claim 1, characterized in that, The specific steps for determining abnormal grayscale values ​​in the road surface data processing module include: P1: Perform grayscale processing on the real-time acquired image data, and segment the grayscale image according to the size of pixel blocks, dividing it into... The image data consists of several identical grayscale blocks, numbered according to their row and column numbers on the grayscale image. The acquired image data is sorted by acquisition time, and the corresponding numbered grayscale blocks within a single image acquired at the same time are further analyzed. Average the gray values and standard deviation The calculation, and the mean obtained from the calculation. and standard deviation Range of grayscale data The settings; P2: Compare the corresponding grayscale value data with the corresponding grayscale value fluctuation range, mark the corresponding grayscale value data that is outside the fluctuation range as outliers, and record the number of outliers. ;like If the grayscale data is abnormal, the grayscale data will be detected again. This is a preset proportional coefficient; if If outliers are removed, the remaining grayscale data after outlier removal is averaged. The calculation, and the mean obtained from the calculation. As the grayscale value data detected at the corresponding time; P3: If the grayscale data is still determined to be abnormal after re-detection, then the corresponding grayscale block number is determined to be abnormal.

3. The intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base according to claim 2, characterized in that, The specific steps for determining the "final analysis image" in the road surface data processing module include: Q1: After determining the grayscale values ​​of all numbered grayscale blocks in a grayscale image, how many grayscale blocks are abnormal? Record and take the quantity The smallest grayscale image is the preliminary analysis image; based on the grayscale features of the standard road surface structure pre-stored in the knowledge base, the grayscale values ​​of all grayscale blocks on the preliminary analysis image are compared to determine the corresponding grayscale block number, and these grayscale blocks are named structural grayscale blocks. Q2: What is the number of grayscale block anomalies in the structural grayscale blocks of the preliminary analysis image? Perform statistics and collect quantities. The smallest grayscale image is the final analysis image; the terminal of the road damage detection vehicle identifies the road condition of the final analysis image in real time, distinguishes the damage type, and synchronously stores the final analysis image, point cloud, location and size data of the corresponding damage type.

4. The intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base as described in claim 1, characterized in that, In the road surface damage identification module, the weight coefficients for various types of damage are dynamically determined. The methods include: W1: Calculate the total area data of the corresponding damage type in real time, divide it by the total area of ​​the detected road segment to obtain the damage density data of the corresponding damage type. Based on the damage density of the corresponding damage type acquired in real time, retrieve the damage degree data of the corresponding damage type from the knowledge base. , The index corresponds to the type of damage; therefore, the road surface damage index is... ,in To detect the total number of different types of damage found within the road section, For the first Weighting coefficients for different types of damage; W2: Obtain historical data, acquire the pavement damage degree corresponding to the corresponding damage type in the historical data, and filter out the data group with only one damage type data fluctuation under different damage degrees. By comparing the fluctuation of the corresponding damage type with the damage degree in the data group, the change in damage degree caused by the unit fluctuation of the damage type can be obtained. W3: Calculate the total change in damage severity caused by unit fluctuations for each damage type. Use the proportion of the change in damage severity caused by unit fluctuations for the corresponding damage type to the total change in damage severity caused by unit fluctuations for all damage types as the weighting coefficient for the corresponding damage type. .

5. The intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base according to claim 1, characterized in that, In the road performance evaluation module, the road service performance index The computational and decision fusion methods include: U1: Road surface smoothness index obtained from the road damage detection vehicle To obtain the driving quality index ; Obtain the rut depth at the location of the tire track on the road surface The rut depth index was obtained. Based on the obtained lateral force coefficient The anti-skid performance index of the road surface was calculated. Based on the measured rebound deflection value Initial deflection value and ultimate deflection value The structural strength index was calculated. ; corresponding road surface performance index Equal to the corresponding road surface , , and The sum of the products of the index and the corresponding index weight coefficients, divided by the sum of the corresponding index weight coefficients; U2: Get value corresponding Value, based on historical data The remaining service life data of road surface damage is reclassified, and the two classification areas are compared. If the classification areas are the same, the classification area is determined to be normal; if the classification areas are different, the same area that does not require maintenance is retained, and the remaining area is classified into the next level. Based on the final classification result, recommended maintenance measures and their implementation priorities are matched from the knowledge base.

6. The intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base according to claim 1, characterized in that, The methods for quantifying the correlation and coupling relationships between damage types in the damage prediction and decision-making module include: G1: Correlation Strength ,in The total number of historical samples. This refers to the historical sample sequence number. For the first The weights of each sample, For indicator functions, For time window, For the first Damage type in each sample At any moment Quantitative indicators For the first Damage type in each sample At any moment Quantitative indicators and This is the trigger threshold value for the corresponding damage type; G2: When Types of damage over time Quantitative indicators And the correlation strength When determining the type of damage This will trigger the damage type. Triggering critical time , This is the current comprehensive environmental load factor. and These are the fitting parameters; G3: Coupling coefficient caused by mutual influence ,in and This refers to the natural deterioration rate when each damage type, categorized by pavement type and obtained from the knowledge base, exists alone, and the combined deterioration rate when two damage types coexist. , for Types of damage over time Quantitative indicators; based on the superposition coupling coefficient, the deterioration rate when damage types coexist is the same as when they exist independently. times.

7. The intelligent decision-making method for road maintenance based on a multimodal large model and knowledge base as described in claim 6, characterized in that, The damage prediction and decision-making module further includes a prediction and blocking decision-making method based on coupled evolution equations; H1: Damage Type Coupled evolution equations , The total number of coexisting damage types. Type of damage For types of damage The triggering growth rate, It is a step function; and In relation to the above analysis and The computational logic and physical meaning are completely identical, only the triggering direction is opposite; H2: Real-time input of quantitative indicators for each type of damage Combined coefficient of environmental load And import the data obtained from the large model. , and Solve the coupled evolution equations to obtain the future time. of sequence, For the prediction period; according to achieve The order of events determines the evolutionary sequence, and the corresponding quantitative indicators for the damage type are used. achieve When this type of damage is determined to have officially formed and begun to affect other types of damage; H3: The evolutionary stages are divided into three phases: single damage type, superposition of two damage types, and coupling of multiple damage types, according to the coupling evolution equation. Predict the evolution time of the corresponding stage, and before the corresponding evolution time is reached, initiate corresponding maintenance measures according to the damage type of the evolution to block the coupled evolution.

Images