Digital integrated application system based on intelligent highway management and maintenance

Abstract

The invention discloses a digital integrated application system based on intelligent highway management and maintenance, and the system comprises a data collection and sensing module, a data transmission and communication module, a digital integrated application module, an intelligent highway management and maintenance module, a data storage and management module, and a user monitoring and decision module. The data transmission and communication module is used for data transmission and communication. The digital integration application module is used for data fusion processing and analysis. The road intelligent management and maintenance module is used for road state detection and preventive maintenance. The data storage and management module is used for data storage and management. The user monitoring and decision-making module is used for providing a user interface and decision-making support. The invention provides a digital integrated application system based on intelligent road management and maintenance, provides a road information extraction algorithm based on multi-modal data fusion for multi-modal data fusion and information extraction, and provides a road maintenance strategy optimization algorithm based on a genetic algorithm for making and optimizing a road maintenance strategy.

Description

Technical Field

[0001] The present invention relates to the fields of multimodal data fusion, intelligent information extraction and genetic algorithms, and specifically to a digital integrated application system based on intelligent highway maintenance. Background Art

[0002] Multimodal data fusion and intelligent information extraction technology is a multimodal data processing and information extraction method designed to solve the problems of multimodal data processing and information extraction in sensor networks. Multimodal data fusion technology fuses multimodal data from different sensors, and intelligent information extraction technology can intelligently extract the most valuable information for highway status monitoring and maintenance. Through the effective combination of the two technologies, the system can more comprehensively and accurately grasp the real-time status of the highway, providing a solid data foundation for intelligent highway management and maintenance.

[0003] Genetic algorithm is an optimization algorithm based on evolutionary theory, which aims to solve the problems of highway maintenance demand prediction and maintenance strategy optimization. By analyzing historical data and current status indicators, genetic algorithm can predict future highway maintenance needs and formulate and optimize highway maintenance strategies. The use of genetic algorithm can significantly improve maintenance effects and resource utilization efficiency, ensure that highways operate in the optimal state, extend their service life, and reduce maintenance costs.

[0004] However, the existing digital integrated application system based on intelligent highway maintenance lacks effective methods for multimodal data processing and information extraction, resulting in low data utilization efficiency and incomplete information extraction. Secondly, the system lacks advanced algorithm support for highway maintenance demand prediction and maintenance strategy optimization, and is unable to accurately predict future maintenance needs, resulting in unscientific maintenance strategies and low resource utilization efficiency. Summary of the Invention

[0005] The purpose of the present invention is to provide a digital integrated application system based on intelligent highway management and maintenance, so as to solve the problems existing in the existing digital integrated application system based on intelligent highway management and maintenance proposed in the above background technology, namely, the lack of effective methods in multimodal data processing and information extraction, resulting in low data utilization efficiency and incomplete information extraction, and the system lacks advanced algorithm support in highway maintenance demand prediction and maintenance strategy optimization, resulting in the inability to accurately predict future maintenance needs, and further resulting in the problem that the maintenance strategy is not scientific enough and the resource utilization efficiency is low.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a digital integrated application system based on intelligent highway management and maintenance, including a data acquisition and sensing module, a data transmission and communication module, a digital integrated application module, an intelligent highway management and maintenance module, a data storage and management module, and a user monitoring and decision-making module, characterized in that: the data acquisition and sensing module is used to collect highway status data in real time through a sensor network, including temperature, humidity, pressure and vibration information; the data transmission and communication module is used to transmit the collected data through a wireless network and 5G technology and communicate with a central server; the digital integrated application module includes a data fusion processing unit and an intelligent analysis and application unit; the data fusion processing unit proposes a highway information extraction algorithm based on multimodal data fusion for performing data fusion processing on multimodal data from the sensor network and intelligently extracting effective information; the intelligent The analysis and application unit is used to perform intelligent analysis on the effective information extracted from the fused data and apply the analysis results to highway status prediction, maintenance strategy optimization and decision support. The highway intelligent management and maintenance module includes a highway status monitoring unit and a maintenance strategy optimization unit. The highway status monitoring unit is used to monitor the specific status indicators of the highway in real time, including road surface temperature, humidity, crack width, settlement height and traffic flow. The maintenance strategy optimization unit proposes a highway maintenance strategy optimization algorithm based on genetic algorithm to formulate and optimize highway maintenance strategies to improve maintenance effects and resource utilization efficiency. The data storage and management module is used to centrally store and manage the collected data and processed analysis data to ensure data security and availability. The user monitoring and decision-making module is used to provide users with a real-time monitoring interface and decision support functions, including data visualization, early warning notification and decision control functions.

[0007] Preferably, the data acquisition and sensing module collects multimodal data of the highway in real time through a sensor network constructed on the highway, including temperature sensors, humidity sensors, pressure sensors and vibration sensors, to ensure that the status information of the highway can be obtained in real time, providing a basis for subsequent data processing and analysis.

[0008] Preferably, the data transmission and communication module transmits the collected data to the central server in a timely manner by utilizing wireless networks and 5G communication technology, ensuring the real-time and accuracy of the data to support immediate decision-making and response.

[0009] Preferably, the digital integrated application module includes a data fusion processing unit, which proposes a highway information extraction algorithm based on multimodal data fusion, integrates and processes multimodal data from different sensors, ensures the consistency and integrity of the data, and extracts effective information and utilization value of the multimodal data.

[0010] Specifically, the highway information extraction algorithm based on multimodal data fusion is as follows: First, assume the organizational structure of a data unit. The data unit constitutes the basic unit of the overall multimodal data fusion algorithm. Each data unit contains an information core, a strategy set, and a data pipeline set corresponding to the strategy. The specific formula of the organizational structure of the data unit is expressed as follows:

[0011] U={C,S,O}

[0012] Among them, U represents the data unit, C represents the information core, S represents the strategy set, and O represents the data pipeline set. The information core C is the core component of the data unit and carries the key information for data intelligence implementation. The strategy set S includes input and output strategies, data standardization strategies, and mathematical logic operation strategies. The data pipeline O is the channel for information transmission and communication, which corresponds one-to-one with the strategy. A new information core is generated through the association of data pipelines to realize the hierarchical structure of the data unit. Then, the initial data unit is constructed to be directly associated with the database, including the construction of the initial information core, the construction of the initial strategy, and the construction of the initial data pipeline. The initial information core is the field name and technical name extracted from the database and is the common attribute of the data. The initial strategy set includes basic strategies and advanced strategies. The basic strategy involves data input and output, standardization, and basic mathematical logic operations. The initial data pipeline corresponds one-to-one with the strategy. The data pipeline is defined by the strategy for data transmission and communication. The specific formula is expressed as follows:

[0013] U i ={C i ,S(C i ),O(C i ,S(C i ))}

[0014] Among them, U i Represented as the i-th initial data unit, C i Represented as the i-th information core, the key information extracted from the database, S(C i ) is expressed as information core C i The policy function, O(C i ,S(C i )) is represented as a data pipeline function, which generates a data pipeline based on the information core and strategy. i represents the index of the data unit. The specific formula of the data normalization strategy is expressed as:

[0015]

[0016] Among them, S norm (C i ) is applied to the information core C i The data standardization strategy function, norm represents the standardization operation, Represented as information core C i The mean of Represented as information core C i Secondly, randomly associate the data pipelines and continuously generate information cores of high-level data units containing new information to construct high-level data units. Through all the data pipelines of the initial data unit, a data pipeline set of the initial data unit is formed. On this basis, randomly associate the data pipelines to generate information cores of binary combination data units. The specific formula is expressed as follows:

[0017]

[0018] Among them, R represents the random association result set of the data pipeline, r(O p ,O q ) is represented as a data pipeline O p and O q The correlation results of O p Represented as the pth element in the data pipeline, O q Represents the qth element in the data pipeline, p and q represent the index of the element in the data pipeline, It is expressed as a full quantitative operation. Then, by building a reward and punishment model and training the model by simulating the manual scoring method of information cores, the intelligent scoring and storage of the fused multimodal data are realized. By automatically screening out information cores with practical significance, high-quality data information is provided for the digital integrated application system based on smart highway management and maintenance. The specific formula of the reward and punishment model is expressed as:

[0019]

[0020] Among them, Loss(Θ) represents the loss function of the reward and punishment model, Θ represents the parameter set of the reward and punishment model, x represents the data sample of the combined data unit, X represents the set of training samples, r Θ (x) represents the prediction of the reward and punishment model for the data sample x based on the parameter set Θ, ∈ represents the regularization parameter, Denotes the actual label of the data sample x, σ denotes the Sigmoid function, and log denotes the logarithmic function. Finally, a task-driven intelligent information extraction method is constructed. The actual task request is used as information input. The topic model of natural language understanding is used to decompose the task request into a set of topic titles for retrieving information cores. Then, the random association of the data pipeline is used to generate an information core set of combined data units related to the task request. Through the training of the constructed reward and punishment model, the information cores are intelligently sorted and the optimal matching information core is stored for a given task request. The specific formula is expressed as follows:

[0021]

[0022] Among them, Loss Request (Θ) represents the loss function of the reward and punishment model in the actual task requirement Request, and Request represents the actual task requirement. It is expressed as an expected value operation, which means that all possible task requests and information checks are expected to be performed. β Represented as the βth information core, C φ It is represented as the information core of φ, β and φ are the indexes of the information core, D is the set of training samples, and each sample is a triple (Request, c β ,c φ ), from the pairing of task request and information core, Rank Θ (Request,c β ) and Rank Θ (Request,c φ ) are respectively represented as the information core c under a given task request Request β and c φ The ranking in the sorted list is achieved by minimizing the loss function Loss Request (Θ) is used to optimize the model parameters Θ, so that the model can realize the intelligent and autonomous extraction of the fused multimodal data information.

[0023] Preferably, the digital integrated application module includes an intelligent analysis and application unit, which performs in-depth analysis on the processed data through the effective information of the multimodal data extracted by the data fusion processing unit to ensure the generation of valuable application solutions and provide intelligent support for highway management.

[0024] Preferably, the smart highway maintenance module includes a highway status monitoring unit, which ensures that potential problems and risks can be discovered in a timely manner by monitoring various highway status data in real time, including road surface temperature, humidity, crack width, settlement height and traffic flow.

[0025] Preferably, the intelligent highway maintenance module includes a maintenance strategy optimization unit, which proposes a highway maintenance strategy optimization algorithm based on a genetic algorithm. By analyzing historical data and current status indicators, it predicts future highway maintenance needs, ensures that reasonable maintenance strategies can be formulated in advance, and improves the resource utilization efficiency of the highway.

[0026] Specifically, the highway maintenance strategy optimization algorithm based on genetic algorithm is as follows: First, a mathematical model of objective function is constructed to quantify the decision problem of highway maintenance, and the decision variable y is defined as iIndicates whether the i-th road is selected as the maintenance object. When the value is 1, it means maintenance is selected, and when the value is 0, it means no maintenance. The specific formula of the objective function mathematical model is expressed as:

[0027]

[0028] Where F is the objective function, which aims to minimize the total maintenance cost while maximizing the maintenance quality, min is the minimization function, and c i It is expressed as the maintenance cost of the ith road, which means the economic investment required to maintain the ith road. i is the road index, N is the total number of roads, st is the conditional function, Quality is the sum of the maintenance quality, which is used to reflect the overall quality effect of the maintenance strategy, max is the maximization function, and b i It is expressed as a measure of highway maintenance quality. The constructed objective function mathematical model converts the actual maintenance demand into a mathematical expression solved by the algorithm, ensuring that the model can effectively reflect the economy and efficiency of the highway maintenance strategy. Furthermore, considering that maintenance costs are also affected by time and space, the impact of time mainly includes the real-time fluctuations of labor costs and material costs, and the impact of space mainly includes the coordinated maintenance needs of adjacent road sections. The time attenuation factor and spatial correlation term are introduced into the original objective function F. The updated objective function is F(t), which is expressed as follows:

[0029]

[0030] Where t is time, c i (t) is the time-varying maintenance cost, α is the recent maintenance penalty coefficient, which is used to characterize the economic penalty intensity for repeated maintenance of the same road section in a short period of time, and β is the time decay rate, which is used to control the penalty term with the time interval (tt last ) decay rate, (tt last ) is the time interval, φ ij is the spatial coupling coefficient, E is the road network topology edge set, y i (t) represents whether the i-th road is selected as the maintenance object at time t, y j (t) represents whether the jth road is selected as the maintenance object at time t. The composite Poisson-Wiener process is introduced to enhance the random interference term, which is expressed as follows:

[0031]

[0032] Where κ is the pavement natural recovery coefficient, which reflects the self-repair ability of the pavement when there is no traffic load; ε is the traffic flow fluctuation intensity coefficient, which quantifies the randomness of the accumulation of small pavement damage caused by normal traffic flow; W tThe standard Brownian motion of the Wiener process is used to simulate the continuous high-frequency small shocks generated by randomly distributed light vehicles in the traffic flow. t is a burst discharge counting process, which is used to express the expected number of heavy vehicles passing by and causing significant damage per unit time, P t Obeying the Poisson distribution, ξ k is the single impact damage intensity, which obeys the Pareto distribution, τ k is the moment of impact, δ(t-τ k ) is expressed as a Dirac impulse function. Then, it provides a starting point for the iterative search process of the genetic algorithm, generates an initial solution set with sufficient diversity, and ensures that the genetic algorithm can cover a wide area of ​​the solution space. The specific formula is expressed as follows:

[0033] S(0)=y1(0),y2(0),…,y P (0) = {y i (0)|i=1,2…,P}

[0034] Among them, S(0) represents the initial population, which includes the P solutions at the beginning of the genetic algorithm, y1(0), y2(0),…, y P (0) represents the independent solution in the P initial population, y i (0) represents the i-th solution in the initial population, i represents the index of the population number, ranging from 1 to P, P represents the population size, representing the total number of solutions in the initial population. By initializing the population, a set of initial solutions based on heuristic rules are generated. The initial solutions constitute the first generation population of the genetic algorithm. The diversity of the population directly affects the search ability of the algorithm and the quality of the final solution. The well-constructed initial population helps the algorithm to effectively explore the solution space and avoid falling into the local optimal solution. Secondly, the fitness of each solution is quantitatively evaluated to determine the quality of the solution and screen out the solution that contributes more to the optimization goal as a candidate for subsequent iterations. The specific formula is expressed as follows:

[0035] f(y)=w1·Cost(y)+w2·QUL(y)

[0036] Among them, f(y) is represented as the fitness function, which is used to evaluate the overall performance of solution y, y is represented as an independent solution in the population, w1 is represented as the cost weight coefficient, which is used to adjust the influence of maintenance cost in the fitness function, Cost(y) is represented as the maintenance cost function, which represents the total maintenance cost calculated according to solution y, w2 is represented as the quality weight coefficient, which is used to adjust the influence of maintenance quality in the fitness function, QUL(y) is represented as the maintenance quality function, which represents the maintenance quality calculated according to solution y. Through the fitness function f(y), the genetic algorithm can find a balance point so that the total cost is minimized while the maintenance quality reaches a level that meets the standards. Then, through the selection mechanism, excellent individual solutions are screened from the current population according to fitness. Excellent individual solutions will serve as the parents of the next generation of the genetic algorithm. The specific formula is expressed as follows:

[0037]

[0038] Among them, Probability(y j ) is expressed as the individual solution y j The probability of being selected represents a relative measure based on individual fitness and is used in the roulette wheel selection mechanism, y j It represents the jth individual solution in the current population, j represents the index of the individual solution in the current population, f(y j ) is represented by individual y j The fitness function of the algorithm is based on the principle of fitness selection, which makes the individual solutions with higher fitness have a high probability of being selected, so that the algorithm can select excellent solutions in the iterative process while maintaining the diversity of the overall algorithm. Finally, new genetic characteristics are introduced into the population through crossover and mutation operations to increase the diversity of the population and improve the probability of finding the global optimal solution. The specific formula is expressed as follows:

[0039] Offspring=Crossover(y a ,y b )

[0040] y Offspring,变异a =Mutate(y a )

[0041] y Offspring,变异b =Mutate(y b )

[0042] Among them, Offspring represents the newly generated offspring individual, which represents the result of the crossover operation. Crossover(y a ,y b ) is represented as a crossover operation, which is used to solve y from two parent individuals. a and y bExchange genetic information to produce new offspring, y a with y b It represents the parent individual solution, a and b represent the parent individual solution index, y Offspring,变异1 with y Offspring,变异2 It is represented as the offspring individual solution after the mutation operation, Mutate(y a ) and Mutate(y b ) is expressed as a mutation operation function. New genetic features are introduced through mutation operation to prevent premature convergence of individual solutions. Crossover operation creates offspring by combining the features of two parents. Mutation operation introduces new genetic diversity by randomly changing the features of offspring. By combining crossover operation and mutation operation, the algorithm can better search the solution space and find the optimal solution to the problem.

[0043] Preferably, the data storage and management module ensures the security and availability of data by centrally storing and managing the collected multimodal data, while supporting the storage and retrieval of big data, providing query and analysis functions for historical data, and providing data support for the long-term management and planning of highways.

[0044] Preferably, the user monitoring and decision-making module helps managers make correct decisions by providing an intuitive user interface and decision support functions, including data visualization tools and report generation functions, while ensuring that highway managers can conveniently use the system to query data, view analysis results and receive early warning information.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. Data fusion processing unit A highway information extraction algorithm based on multimodal data fusion is proposed. This algorithm realizes the fusion and intelligent extraction of multimodal data by constructing a data unit containing an information core, a strategy set, and a data pipeline set. The organizational structure of each data unit includes an information core carrying key information, a strategy set including input and output strategies, data standardization strategies, and mathematical logic operation strategies, and a data pipeline set for information transmission and communication, thereby ensuring efficient data processing and utilization. First, the initial data unit is directly associated with the database. The initial information core is constructed by extracting the field names and technology names in the database. In combination with basic and advanced strategies, the data pipeline is defined. The consistency and integrity of the data are guaranteed. Based on these initial data units, the system can effectively standardize and process the data, making the data comparable and operable between different sources. Through the data standardization strategy, the system can eliminate the deviation between different data sources and ensure the accuracy and reliability of the data. Next, through the random association data pipeline, the system generates a high-level data unit information core to achieve deep fusion of multi-source data. This method not only improves the efficiency of data utilization, but also enables the system to extract valuable information cores from large amounts of data, thereby supporting more intelligent decision-making and predictions. The randomly associated data pipeline can construct a multi-combination of data units to further enrich It enriches the data hierarchy of the system and provides strong data support for smart highway maintenance. In addition, based on the task-driven intelligent information extraction method, the actual task request is decomposed into a set of subject titles for retrieval information cores through the topic model of natural language understanding, and a set of combined data unit information cores related to the task request is generated through random association of the data pipeline. By constructing a reward and punishment model and simulating the manual scoring method, the system can realize intelligent scoring and sorting of the fused multimodal data to ensure that the extracted information cores are of practical significance and high quality. The intelligent information extraction method can improve the response speed of the system and significantly enhance the flexibility and efficiency of the system in dealing with complex task requests. Accuracy. Through the objective function, the economic decay effect in the time dimension and spatial correlation constraints are incorporated into the cost calculation, so that maintenance decisions not only consider the current input-output ratio, but also intelligently avoid repeated investment and waste in the same road section in the short term. By introducing the α and β parameters, the system can automatically identify economically inefficient behaviors in historical maintenance records. The composite Poisson-Wiener process is introduced to accurately characterize the differentiated damage mechanism of traffic loads on the road surface. The continuous Brownian motion term captures the uniform wear caused by daily traffic flow, while the Poisson jump process simulates sudden impact events such as overloaded vehicles. This two-level stochastic modeling realizes the differentiated warning of "chronic damage" and "acute trauma" in the maintenance strategy.In summary, the highway information extraction algorithm based on multi-modal data fusion provides strong technical support for highway intelligent management and maintenance by constructing data units, implementing data fusion and intelligent extraction. This algorithm significantly improves the intelligent level of the system while ensuring efficient processing and utilization of data, meeting the demand for high-quality data and intelligent decision-making in highway management and maintenance. Through the highway information extraction algorithm based on multi-modal data fusion, the highway intelligent management and maintenance system can better achieve real-time monitoring, predictive maintenance, and resource optimization, thereby improving the management efficiency and maintenance effect of highways.

[0047] 2. The maintenance strategy optimization unit proposes a highway maintenance strategy optimization algorithm based on genetic algorithm. First, the algorithm quantifies the decision-making problem of highway maintenance by constructing a mathematical model of the objective function, ensuring that the model effectively reflects the economy and efficiency of the maintenance strategy. The objective function aims to minimize the total maintenance cost while maximizing the maintenance quality. By converting the actual maintenance needs into a mathematical expression for algorithm solving, the decision-making is more scientific and reasonable. Second, the initial population generation step of the genetic algorithm ensures extensive coverage of the solution space. By generating a sufficient diversity of initial solution sets, the algorithm can explore different maintenance strategies from multiple angles, avoiding local optimal solutions and improving the global search ability and quality of the final solution. Third, the fitness evaluation mechanism quantifies the performance of each solution, selecting solutions that contribute more to the optimization target. The fitness function considers both maintenance cost and maintenance quality, allowing the algorithm to balance these two key factors during the iteration process, ensuring that costs are controlled while maintaining quality meets standards. In terms of selection mechanism, the selection principle based on fitness gives individuals with high fitness a higher probability of being selected. This mechanism not only selects excellent solutions but also maintains the diversity of the population, preventing premature convergence and improving the probability of discovering the global optimal solution. Finally, through crossover and mutation operations, new genetic characteristics are introduced, further increasing the diversity of the population. Crossover operation creates offspring by combining the characteristics of two parent individuals, while mutation operation introduces new genetic diversity by randomly changing the characteristics of offspring. The combination of these two operations allows the algorithm to better explore the solution space and find the optimal maintenance strategy. In summary, the highway maintenance strategy optimization algorithm based on genetic algorithm provides efficient and intelligent decision support for highway maintenance in a digital integrated application system based on highway intelligent management and maintenance. This algorithm not only improves the scientificity and rationality of decision-making, reducing maintenance costs, but also enhances maintenance quality, ultimately achieving intelligent management of highway maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0048] The invention is further illustrated by the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the invention. A person skilled in the art can obtain other drawings based on the following drawings without making any creative effort.

[0049] Figure 1 It is a structural schematic diagram of the present invention. DETAILED DESCRIPTION

[0050] 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 the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0051] See also Figure 1 The present invention provides a digital integrated application system based on smart highway management and maintenance, including a data acquisition and sensing module, a data transmission and communication module, a digital integrated application module, a smart highway management and maintenance module, a data storage and management module, and a user monitoring and decision-making module. The system is characterized in that: the data acquisition and sensing module is used to collect highway status data in real time through a sensor network, including temperature, humidity, pressure and vibration information; the data transmission and communication module is used to transmit the collected data through a wireless network and 5G technology and communicate with a central server; the digital integrated application module includes a data fusion processing unit and an intelligent analysis and application unit; the data fusion processing unit proposes a highway information extraction algorithm based on multimodal data fusion for performing data fusion processing on multimodal data from the sensor network and intelligently extracting effective information; the intelligent analysis and application unit It is used to intelligently analyze the effective information extracted from the fused data and apply the analysis results to highway status prediction, maintenance strategy optimization and decision support. The highway intelligent management and maintenance module includes a highway status monitoring unit and a maintenance strategy optimization unit. The highway status monitoring unit is used to monitor the specific status indicators of the highway in real time, including road surface temperature, humidity, crack width, settlement height and traffic flow. The maintenance strategy optimization unit proposes a highway maintenance strategy optimization algorithm based on genetic algorithm to formulate and optimize highway maintenance strategies to improve maintenance effects and resource utilization efficiency. The data storage and management module is used to centrally store and manage collected data and processed analysis data to ensure data security and availability. The user monitoring and decision-making module is used to provide users with a real-time monitoring interface and decision support functions, including data visualization, early warning notifications and decision control functions.

[0052] See Figure 1Furthermore, the data acquisition and sensing module collects multimodal data of the highway in real time through a sensor network constructed on the highway, including temperature sensors, humidity sensors, pressure sensors and vibration sensors, ensuring that the highway status information can be obtained in real time, providing a basis for subsequent data processing and analysis.

[0053] See Figure 1 Furthermore, the data transmission and communication module transmits the collected data to the central server in a timely manner by utilizing wireless networks and 5G communication technology, ensuring the real-time and accuracy of the data to support immediate decision-making and response.

[0054] See Figure 1 Furthermore, the digital integrated application module includes a data fusion processing unit, which proposes a highway information extraction algorithm based on multimodal data fusion. By integrating and processing multimodal data from different sensors, it ensures the consistency and integrity of the data and extracts the effective information and utilization value of the multimodal data.

[0055] See Figure 1 Furthermore, the highway information extraction algorithm based on multimodal data fusion is specifically described as follows: First, assume the organizational structure of a data unit. The data unit constitutes the basic unit of the overall multimodal data fusion algorithm. Each data unit contains an information core, a strategy set, and a data pipeline set corresponding to the strategy. The specific formula of the organizational structure of the data unit is expressed as follows:

[0056] U={C,S,O}

[0057] Among them, U represents the data unit, C represents the information core, S represents the strategy set, and O represents the data pipeline set. The information core C is the core component of the data unit and carries the key information for data intelligence implementation. The strategy set S includes input and output strategies, data standardization strategies, and mathematical logic operation strategies. The data pipeline O is the channel for information transmission and communication, which corresponds one-to-one with the strategy. A new information core is generated through the association of data pipelines to realize the hierarchical structure of the data unit. Then, the initial data unit is constructed to be directly associated with the database, including the construction of the initial information core, the construction of the initial strategy, and the construction of the initial data pipeline. The initial information core is the field name and technical name extracted from the database and is the common attribute of the data. The initial strategy set includes basic strategies and advanced strategies. The basic strategy involves data input and output, standardization, and basic mathematical logic operations. The initial data pipeline corresponds one-to-one with the strategy. The data pipeline is defined by the strategy for data transmission and communication. The specific formula is expressed as follows:

[0058] U i ={C i ,S(C i),O(C i ,S(C i ))}

[0059] Among them, U i Represented as the i-th initial data unit, C i Represented as the i-th information core, the key information extracted from the database, S(C i ) is expressed as information core C i The policy function, O(C i ,S(C i )) is represented as a data pipeline function, which generates a data pipeline based on the information core and strategy. i represents the index of the data unit. The specific formula of the data normalization strategy is expressed as:

[0060]

[0061] Among them, S norm (C i ) is applied to the information core C i The data standardization strategy function, norm represents the standardization operation, Represented as information core C i The mean of Represented as information core C i Secondly, randomly associate the data pipelines and continuously generate information cores of high-level data units containing new information to construct high-level data units. Through all the data pipelines of the initial data unit, a data pipeline set of the initial data unit is formed. On this basis, randomly associate the data pipelines to generate information cores of binary combination data units. The specific formula is expressed as follows:

[0062]

[0063] Among them, R represents the random association result set of the data pipeline, r(O p ,O q ) is represented as a data pipeline O p and O q The correlation results of O p Represented as the pth element in the data pipeline, O q Represents the qth element in the data pipeline, p and q represent the index of the element in the data pipeline, It is expressed as a full quantitative operation. Then, by building a reward and punishment model and training the model by simulating the manual scoring method of information cores, the intelligent scoring and storage of the fused multimodal data are realized. By automatically screening out information cores with practical significance, high-quality data information is provided for the digital integrated application system based on smart highway management and maintenance. The specific formula of the reward and punishment model is expressed as:

[0064]

[0065] Among them, Loss(Θ) represents the loss function of the reward and punishment model, Θ represents the parameter set of the reward and punishment model, x represents the data sample of the combined data unit, X represents the set of training samples, r Θ (x) represents the prediction of the reward and punishment model for the data sample x based on the parameter set Θ, ∈ represents the regularization parameter, Denotes the actual label of the data sample x, σ denotes the Sigmoid function, and log denotes the logarithmic function. Finally, a task-driven intelligent information extraction method is constructed. The actual task request is used as information input. The topic model of natural language understanding is used to decompose the task request into a set of topic titles for retrieving information cores. Then, the random association of the data pipeline is used to generate an information core set of combined data units related to the task request. Through the training of the constructed reward and punishment model, the information cores are intelligently sorted and the optimal matching information core is stored for a given task request. The specific formula is expressed as follows:

[0066]

[0067] Among them, Loss Request (Θ) represents the loss function of the reward and punishment model in the actual task requirement Request, and Request represents the actual task requirement. It is expressed as an expected value operation, which means that all possible task requests and information checks are expected to be performed. β Represented as the βth information core, C φ It is represented as the information core of φ, β and φ are the indexes of the information core, D is the set of training samples, and each sample is a triple (Request, c β ,c φ ), from the pairing of task request and information core, Rank Θ (Request,c β ) and Rank Θ (Request,c φ ) are respectively represented as the information core c under a given task request Request β and c φ The ranking in the sorted list is achieved by minimizing the loss function Loss Request (Θ) is used to optimize the model parameters Θ, so that the model can realize the intelligent and autonomous extraction of the fused multimodal data information.

[0068] See Figure 1Further, the digital integrated application module includes an intelligent analysis and application unit that performs in-depth analysis on the processed data based on the multi-modal data effective information extracted by the data fusion processing unit, ensuring the generation of valuable application schemes and providing intelligent support for highway management.

[0069] Referring to Figure 1 Further, the highway intelligent management and maintenance module includes a highway state monitoring unit that monitors various state data of the highway in real time, including pavement temperature, humidity, crack width, settlement height, and traffic flow, ensuring that potential problems and risks can be discovered in a timely manner.

[0070] Referring to Figure 1 Further, the highway intelligent management and maintenance module includes a maintenance strategy optimization unit that proposes a genetic algorithm-based highway maintenance strategy optimization algorithm, which analyzes historical data and current state indicators to predict future highway maintenance needs, ensuring that reasonable maintenance strategies can be developed in advance and improving the resource utilization efficiency of highways.

[0071] Referring to Figure 1 Further, the genetic algorithm-based highway maintenance strategy optimization algorithm is as follows: first, a target function mathematical model is constructed to quantify the decision-making problem of highway maintenance, and decision variables y i are defined, where y i represents whether the i-th highway is selected as a maintenance object, and takes a value of 1 when maintenance is selected and a value of 0 when maintenance is not selected, and the specific formula of the target function mathematical model is:

[0072]

[0073] where F represents the target function, aiming to minimize the total maintenance cost while maximizing the maintenance quality, min represents the minimization function, c i represents the maintenance cost of the i-th highway, represents the economic input required for the maintenance of the i-th highway, i represents the highway number index, N represents the total number of highways, s.t. represents the conditional function, Quality represents the total sum of maintenance quality, reflecting the overall quality effect of the maintenance strategy, max represents the maximization function, b iIt is expressed as a measure of highway maintenance quality. The constructed objective function mathematical model converts the actual maintenance demand into a mathematical expression solved by the algorithm, ensuring that the model can effectively reflect the economy and efficiency of the highway maintenance strategy. Furthermore, considering that maintenance costs are also affected by time and space, the impact of time mainly includes the real-time fluctuations of labor costs and material costs, and the impact of space mainly includes the coordinated maintenance needs of adjacent road sections. The time attenuation factor and spatial correlation term are introduced into the original objective function F. The updated objective function is F(t), which is expressed as follows:

[0074]

[0075] Where t is time, c i (t) is the time-varying maintenance cost, α is the recent maintenance penalty coefficient, which is used to characterize the economic penalty intensity for repeated maintenance of the same road section in a short period of time, and β is the time decay rate, which is used to control the penalty term with the time interval (tt last ) decay rate, (tt last ) is the time interval, φ ij is the spatial coupling coefficient, E is the road network topology edge set, y i (t) represents whether the i-th road is selected as the maintenance object at time t, y j (t) represents whether the jth road is selected as the maintenance object at time t. The composite Poisson-Wiener process is introduced to enhance the random interference term, which is expressed as follows:

[0076]

[0077] Where κ is the pavement natural recovery coefficient, which reflects the self-repair ability of the pavement when there is no traffic load; ε is the traffic flow fluctuation intensity coefficient, which quantifies the randomness of the accumulation of small pavement damage caused by normal traffic flow; W t The standard Brownian motion of the Wiener process is used to simulate the continuous high-frequency small shocks generated by randomly distributed light vehicles in the traffic flow. t is a burst discharge counting process, which is used to express the expected number of heavy vehicles passing by and causing significant damage per unit time, P t Obeying the Poisson distribution, ξ k is the single impact damage intensity, which obeys the Pareto distribution, τ k is the moment of impact, δ(t-τ k ) is expressed as a Dirac impulse function. Then, it provides a starting point for the iterative search process of the genetic algorithm, generates an initial solution set with sufficient diversity, and ensures that the genetic algorithm can cover a wide area of ​​the solution space. The specific formula is expressed as follows:

[0078] S(0)=y1(0),y2(0),…,y P(0) = {y i (0)|i=1,2…,P}

[0079] Among them, S(0) represents the initial population, which includes the P solutions at the beginning of the genetic algorithm, y1(0), y2(0),…, y P (0) represents the independent solution in the P initial population, y i (0) represents the i-th solution in the initial population, i represents the index of the population number, ranging from 1 to P, P represents the population size, representing the total number of solutions in the initial population. By initializing the population, a set of initial solutions based on heuristic rules are generated. The initial solutions constitute the first generation population of the genetic algorithm. The diversity of the population directly affects the search ability of the algorithm and the quality of the final solution. The well-constructed initial population helps the algorithm to effectively explore the solution space and avoid falling into the local optimal solution. Secondly, the fitness of each solution is quantitatively evaluated to determine the quality of the solution and screen out the solution that contributes more to the optimization goal as a candidate for subsequent iterations. The specific formula is expressed as follows:

[0080] f(y)=w1·Cost(y)+w2·QUL(y)

[0081] Among them, f(y) is represented as the fitness function, which is used to evaluate the overall performance of solution y, y is represented as an independent solution in the population, w1 is represented as the cost weight coefficient, which is used to adjust the influence of maintenance cost in the fitness function, Cost(y) is represented as the maintenance cost function, which represents the total maintenance cost calculated according to solution y, w2 is represented as the quality weight coefficient, which is used to adjust the influence of maintenance quality in the fitness function, QUL(y) is represented as the maintenance quality function, which represents the maintenance quality calculated according to solution y. Through the fitness function f(y), the genetic algorithm can find a balance point so that the total cost is minimized while the maintenance quality reaches a level that meets the standards. Then, through the selection mechanism, excellent individual solutions are screened from the current population according to fitness. Excellent individual solutions will serve as the parents of the next generation of the genetic algorithm. The specific formula is expressed as follows:

[0082]

[0083] Among them, Probability(y j ) is expressed as the individual solution y j The probability of being selected represents a relative measure based on individual fitness and is used in the roulette wheel selection mechanism, y j It represents the jth individual solution in the current population, j represents the index of the individual solution in the current population, f(y j ) is represented by individual y jThe fitness function of the algorithm is based on the principle of fitness selection, which makes the individual solutions with higher fitness have a high probability of being selected, so that the algorithm can select excellent solutions in the iterative process while maintaining the diversity of the overall algorithm. Finally, new genetic characteristics are introduced into the population through crossover and mutation operations to increase the diversity of the population and improve the probability of finding the global optimal solution. The specific formula is expressed as follows:

[0084] Offspring=Crossover(y a ,y b )

[0085] y Offspring,变异a =Mutate(y a )

[0086] y Offspring,变异b =Mutate(y b )

[0087] Among them, Offspring represents the newly generated offspring individual, which represents the result of the crossover operation. Crossover(y a ,y b ) is represented as a crossover operation, which is used to solve y from two parent individuals. a and y b Exchange genetic information to produce new offspring, y a with y b It represents the parent individual solution, a and b represent the parent individual solution index, y Offspring,变异1 with y Offspring,变异2 It is represented as the offspring individual solution after the mutation operation, Mutate(y a ) and Mutate(y b ) is expressed as a mutation operation function. New genetic features are introduced through mutation operation to prevent premature convergence of individual solutions. Crossover operation creates offspring by combining the features of two parents. Mutation operation introduces new genetic diversity by randomly changing the features of offspring. By combining crossover operation and mutation operation, the algorithm can better search the solution space and find the optimal solution to the problem.

[0088] See Figure 1 Furthermore, the data storage and management module ensures the security and availability of data by centrally storing and managing the collected multimodal data, while supporting the storage and retrieval of big data, providing query and analysis functions for historical data, and providing data support for the long-term management and planning of highways.

[0089] See Figure 1Furthermore, the user monitoring and decision-making module helps managers make correct decisions by providing an intuitive user interface and decision support functions, including data visualization tools and report generation functions, while ensuring that highway managers can conveniently use the system to query data, view analysis results and receive warning information.

[0090] In specific use, first, the data acquisition and sensing module is used to collect the status data of the highway in real time through the sensor network, including temperature, humidity, pressure and vibration information. The data transmission and communication module is used to transmit the collected data through wireless networks and 5G technology and communicate with the central server. Then, the digital integrated application module includes a data fusion processing unit and an intelligent analysis and application unit. The data fusion processing unit proposes a highway information extraction algorithm based on multimodal data fusion to perform data fusion processing on the multimodal data from the sensor network and intelligently extract effective information. The intelligent analysis and application unit is used to perform intelligent analysis on the effective information extracted from the fused data and apply the analysis results to highway status prediction and maintenance strategies. Strategy optimization and decision support. Secondly, the highway intelligent management and maintenance module includes a highway status monitoring unit and a maintenance strategy optimization unit. The highway status monitoring unit is used to monitor the specific status indicators of the highway in real time, including road surface temperature, humidity, crack width, settlement height and traffic flow. The maintenance strategy optimization unit proposes a highway maintenance strategy optimization algorithm based on genetic algorithm to formulate and optimize highway maintenance strategies to improve maintenance effects and resource utilization efficiency. Finally, the data storage and management module is used to centrally store and manage collected data and processed analysis data to ensure data security and availability. The user monitoring and decision-making module is used to provide users with a real-time monitoring interface and decision support functions, including data visualization, early warning notifications and decision control functions.

[0091] Compared with the prior art, the present invention has the following beneficial effects:

[0092] 3. Data fusion processing unit proposes a highway information extraction algorithm based on multimodal data fusion. This algorithm realizes the fusion and intelligent extraction of multimodal data by constructing a data unit containing an information core, a strategy set, and a data pipeline set. The organizational structure of each data unit includes an information core carrying key information, a strategy set including input and output strategies, data standardization strategies, and mathematical logic operation strategies, and a data pipeline set for information transmission and communication, thereby ensuring efficient processing and utilization of data. First, the initial data unit is directly associated with the database. The initial information core is constructed by extracting the field names and technology names in the database, and the data pipeline is defined by combining basic and advanced strategies. The consistency and integrity of the data are guaranteed. Based on these initial data units, the system can effectively standardize and process the data, making the data comparable and operable between different sources. Through the data standardization strategy, the system can eliminate the deviation between different data sources and ensure the accuracy and reliability of the data. Next, through the random association data pipeline, the system generates a high-level data unit information core to achieve deep fusion of multi-source data. This method not only improves the efficiency of data utilization, but also enables the system to extract valuable information cores from large amounts of data, thereby supporting more intelligent decision-making and predictions. The randomly associated data pipeline can construct a multi-combination of data units to further enrich It enriches the data hierarchy of the system and provides strong data support for smart highway maintenance. In addition, based on the task-driven intelligent information extraction method, the actual task request is decomposed into a set of subject titles for retrieval information cores through the topic model of natural language understanding, and a set of combined data unit information cores related to the task request is generated through random association of the data pipeline. By constructing a reward and punishment model and simulating the manual scoring method, the system can realize intelligent scoring and sorting of the fused multimodal data to ensure that the extracted information cores are of practical significance and high quality. The intelligent information extraction method can improve the response speed of the system and significantly enhance the flexibility and efficiency of the system in dealing with complex task requests. Accuracy. Through the objective function, the economic decay effect in the time dimension and spatial correlation constraints are incorporated into the cost calculation, so that maintenance decisions not only consider the current input-output ratio, but also intelligently avoid repeated investment and waste in the same road section in the short term. By introducing the α and β parameters, the system can automatically identify economically inefficient behaviors in historical maintenance records. The composite Poisson-Wiener process is introduced to accurately characterize the differentiated damage mechanism of traffic loads on the road surface. The continuous Brownian motion term captures the uniform wear caused by daily traffic flow, while the Poisson jump process simulates sudden impact events such as overloaded vehicles. This two-level stochastic modeling realizes the differentiated warning of "chronic damage" and "acute trauma" in the maintenance strategy.In summary, the highway information extraction algorithm based on multimodal data fusion provides strong technical support for smart highway management and maintenance by constructing data units and realizing data fusion and intelligent extraction. While ensuring efficient data processing and utilization, the algorithm significantly improves the intelligence level of the system and meets the needs of high-quality data and intelligent decision-making in highway management and maintenance. Through the highway information extraction algorithm based on multimodal data fusion, the smart highway management and maintenance system can better realize real-time monitoring, predictive maintenance and optimal resource allocation, thereby improving highway management efficiency and maintenance effects.

[0093] 4. The maintenance strategy optimization unit proposes a highway maintenance strategy optimization algorithm based on genetic algorithm. First, the algorithm quantifies the decision-making problem of highway maintenance by constructing a mathematical model of the objective function to ensure that the model can effectively reflect the economy and efficiency of the maintenance strategy. The objective function aims to minimize the total maintenance cost while maximizing the maintenance quality. By converting the actual maintenance needs into mathematical expressions solved by the algorithm, the decision is ensured to be more scientific and reasonable. Secondly, the initial population generation step of the genetic algorithm ensures a wide coverage of the understanding space. By generating an initial solution set with sufficient diversity, the algorithm can explore different maintenance strategies from multiple perspectives to avoid falling into local optimal solutions, thereby improving the global search ability of the algorithm and the quality of the final solution. Then, the fitness evaluation mechanism quantifies the performance of each solution and selects solutions that contribute more to the optimization goal. The fitness function comprehensively considers the maintenance cost and maintenance quality, so that the algorithm can balance these two key factors during the iteration process, ensuring that both costs can be controlled and maintenance quality can be guaranteed. In terms of the selection mechanism, the fitness-based selection principle makes individual solutions with high fitness have a higher probability of being selected. This mechanism can not only select excellent solutions, but also maintain the diversity of the population, prevent the algorithm from converging prematurely, and thus increase the probability of finding the global optimal solution. Finally, through crossover and mutation operations, new genetic characteristics are introduced, which further increases the diversity of the population. The crossover operation creates offspring by combining the characteristics of two parents, and the mutation operation introduces new genetic diversity by randomly changing the characteristics of the offspring. The combination of these two operations enables the algorithm to better explore the solution space and find the optimal maintenance strategy. In summary, the highway maintenance strategy optimization algorithm based on genetic algorithm provides efficient and intelligent highway maintenance decision support through scientific modeling and optimization methods in a digital integrated application system based on highway intelligent management and maintenance, which not only improves the scientificity and rationality of decision-making, reduces maintenance costs, but also improves maintenance quality, and ultimately realizes intelligent management of highway maintenance.

[0094] Although the present application has been described in detail with reference to the foregoing embodiments, the technical solutions recorded in the foregoing embodiments can be modified, or some of the technical features can be replaced by equivalent features, by those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A digital integrated application system based on smart highway management and maintenance, comprising a data acquisition and sensing module, a data transmission and communication module, a digital integrated application module, a smart highway management and maintenance module, a data storage and management module, and a user monitoring and decision-making module, characterized by: The data acquisition and sensing module is used to collect highway status data in real time through the sensor network, including temperature, humidity, pressure and vibration information. The data transmission and communication module is used to transmit the collected data and communicate with the central server through wireless networks and 5G technology. The digital integrated application module includes a data fusion processing unit and an intelligent analysis and application unit. The data fusion processing unit proposes a highway information extraction algorithm based on multimodal data fusion for data fusion processing of multimodal data from the sensor network and intelligent extraction of effective information. The intelligent analysis and application unit is used to perform intelligent analysis on the effective information extracted from the fused data and apply the analysis results to highway status prediction and maintenance strategy optimization. and decision support, the highway intelligent management and maintenance module includes a highway status monitoring unit and a maintenance strategy optimization unit, the highway status monitoring unit is used to monitor the specific status indicators of the highway in real time, including road surface temperature, humidity, crack width, settlement height and traffic flow, the maintenance strategy optimization unit proposes a highway maintenance strategy optimization algorithm based on genetic algorithm to formulate and optimize highway maintenance strategies to improve maintenance effects and resource utilization efficiency, the data storage and management module is used to centrally store and manage collected data and processed analysis data to ensure data security and availability, the user monitoring and decision-making module is used to provide users with a real-time monitoring interface and decision support functions, including data visualization and early warning notification and decision control functions.

2. The digital integrated application system based on intelligent highway management and maintenance according to claim 1 is characterized by: The data acquisition and sensing module collects multimodal data of the highway in real time through a sensor network built on the highway, including temperature sensors, humidity sensors, pressure sensors and vibration sensors, ensuring that the highway status information can be obtained in real time, providing a basis for subsequent data processing and analysis.

3. The digital integrated application system based on intelligent highway management and maintenance according to claim 1 is characterized by: The data transmission and communication module uses wireless networks and 5G communication technology to transmit the collected data to the central server in a timely manner, ensuring the real-time and accuracy of the data to support immediate decision-making and response.

4. A digital integrated application system based on intelligent highway management and maintenance according to claim 1, characterized in that: The digital integrated application module includes a data fusion processing unit and an intelligent analysis and application unit. The data fusion processing unit proposes a highway information extraction algorithm based on multimodal data fusion. By integrating and processing multimodal data from different sensors, it ensures the consistency and integrity of the data and extracts the effective information and utilization value of the multimodal data. The intelligent analysis and application unit performs in-depth analysis on the processed data using the effective information of the multimodal data extracted by the data fusion processing unit, ensuring the generation of valuable application solutions and providing intelligent support for highway management.

5. The digital integrated application system based on intelligent highway management and maintenance according to claim 4 is characterized by: First, assume the organizational structure of a data unit. The data unit constitutes the basic unit of the overall multimodal data fusion algorithm. Each data unit contains an information core, a set of policies, and a set of data pipelines corresponding to the policies. The specific formula for the organizational structure of the data unit is expressed as: U={C,S,O} Among them, U represents the data unit, C represents the information core, S represents the strategy set, and O represents the data pipeline set. The information core C is the core component of the data unit and carries the key information for data intelligence implementation. The strategy set S includes input and output strategies, data standardization strategies, and mathematical logic operation strategies. The data pipeline O is the channel for information transmission and communication, which corresponds one-to-one with the strategy. A new information core is generated through the association of data pipelines to realize the hierarchical structure of the data unit. Then, the initial data unit is constructed to be directly associated with the database, including the construction of the initial information core, the construction of the initial strategy, and the construction of the initial data pipeline. The initial information core is the field name and technical name extracted from the database and is the common attribute of the data. The initial strategy set includes basic strategies and advanced strategies. The basic strategy involves data input and output, standardization, and basic mathematical logic operations. The initial data pipeline corresponds one-to-one with the strategy. The data pipeline is defined by the strategy for data transmission and communication. The specific formula is expressed as follows: U i ={C i ,S(C i ),O(C i ,S(C i ))} Among them, U i Represented as the i-th initial data unit, C i Represented as the i-th information core, the key information extracted from the database, S(C i ) is expressed as information core C i The policy function, O(C i ,S(C i )) is represented as a data pipeline function, which generates a data pipeline based on the information core and strategy. i represents the index of the data unit. The specific formula of the data normalization strategy is expressed as: Among them, S norm (C i ) is applied to the information core C i The data standardization strategy function, norm represents the standardization operation, Represented as information core C i The mean of Represented as information core C i Secondly, randomly associate the data pipelines and continuously generate information cores of high-level data units containing new information to construct high-level data units. Through all the data pipelines of the initial data unit, a data pipeline set of the initial data unit is formed. On this basis, randomly associate the data pipelines to generate information cores of binary combination data units. The specific formula is expressed as follows: Among them, R represents the random association result set of the data pipeline, r(O p ,O q ) is represented as a data pipeline O p and O q The correlation results of O p Represented as the pth element in the data pipeline, O q Represents the qth element in the data pipeline, p and q represent the index of the element in the data pipeline, It is expressed as a full quantitative operation. Then, by building a reward and punishment model and training the model by simulating the manual scoring method of information cores, the intelligent scoring and storage of the fused multimodal data are realized. By automatically screening out information cores with practical significance, high-quality data information is provided for the digital integrated application system based on smart highway management and maintenance. The specific formula of the reward and punishment model is expressed as: Among them, Loss(Θ) represents the loss function of the reward and punishment model, Θ represents the parameter set of the reward and punishment model, x represents the data sample of the combined data unit, X represents the set of training samples, r Θ (x) represents the prediction of the reward and punishment model for the data sample x based on the parameter set Θ, ∈ represents the regularization parameter, Denotes the actual label of the data sample x, σ denotes the Sigmoid function, and log denotes the logarithmic function. Finally, a task-driven intelligent information extraction method is constructed. The actual task request is used as information input. The topic model of natural language understanding is used to decompose the task request into a set of topic titles for retrieving information cores. Then, the random association of the data pipeline is used to generate an information core set of combined data units related to the task request. Through the training of the constructed reward and punishment model, the information cores are intelligently sorted and the optimal matching information core is stored for a given task request. The specific formula is expressed as follows: Among them, Loss Request (Θ) represents the loss function of the reward and punishment model in the actual task requirement Request, and Request represents the actual task requirement. It is expressed as an expected value operation, which means that all possible task requests and information checks are expected to be performed. β Represented as the βth information core, C φ It is represented as the information core of φ, β and φ are the indexes of the information core, D is the set of training samples, and each sample is a triple (Request, c β ,c φ ), from the pairing of task request and information core, Rank Θ (Request,c β ) and Rank Θ (Request,c φ ) are respectively represented as the information core c under a given task request Request β and c φ The ranking in the sorted list is achieved by minimizing the loss function Loss Request (Θ) is used to optimize the model parameters Θ, so that the model can realize the intelligent and autonomous extraction of the fused multimodal data information.

6. A digital integrated application system based on intelligent highway management and maintenance according to claim 1, characterized in that: The intelligent highway maintenance module includes a highway status monitoring unit and a maintenance strategy optimization unit. The highway status monitoring unit ensures that potential problems and risks can be discovered in a timely manner by monitoring various highway status data in real time, including road surface temperature, humidity, crack width, settlement height and traffic flow; the maintenance strategy optimization unit proposes a highway maintenance strategy optimization algorithm based on genetic algorithm, which predicts future highway maintenance needs by analyzing historical data and current status indicators, ensuring that reasonable maintenance strategies can be formulated in advance and improving the resource utilization efficiency of highways.

7. The digital integrated application system based on intelligent highway management and maintenance according to claim 6 is characterized by: First, a mathematical model of objective function is constructed to quantify the decision problem of highway maintenance, and the decision variable y is defined as i Indicates whether the i-th road is selected as the maintenance object. When the value is 1, it means maintenance is selected, and when the value is 0, it means no maintenance. The specific formula of the objective function mathematical model is expressed as: Where F is the objective function, which aims to minimize the total maintenance cost while maximizing the maintenance quality, min is the minimization function, and c i It is expressed as the maintenance cost of the ith road, which means the economic investment required to maintain the ith road. i is the road index, N is the total number of roads, st is the conditional function, Quality is the sum of the maintenance quality, which is used to reflect the overall quality effect of the maintenance strategy, max is the maximization function, and b i It is expressed as a measure of highway maintenance quality. The constructed objective function mathematical model converts the actual maintenance demand into a mathematical expression solved by the algorithm, ensuring that the model can effectively reflect the economy and efficiency of the highway maintenance strategy. Furthermore, considering that maintenance costs are also affected by time and space, the impact of time mainly includes the real-time fluctuations of labor costs and material costs, and the impact of space mainly includes the coordinated maintenance needs of adjacent road sections. The time attenuation factor and spatial correlation term are introduced into the original objective function F. The updated objective function is F(t), which is expressed as follows: Where t is time, c i (t) is the time-varying maintenance cost, α is the recent maintenance penalty coefficient, which is used to characterize the economic penalty intensity for repeated maintenance of the same road section in a short period of time, and β is the time decay rate, which is used to control the penalty term with the time interval (tt last ) decay rate, (tt last ) is the time interval, φ ij is the spatial coupling coefficient, E is the road network topology edge set, y i (t) represents whether the i-th road is selected as the maintenance object at time t, y j (t) represents whether the jth road is selected as the maintenance object at time t. The composite Poisson-Wiener process is introduced to enhance the random interference term, which is expressed as follows: Where κ is the pavement natural recovery coefficient, which reflects the self-repair ability of the pavement when there is no traffic load; ε is the traffic flow fluctuation intensity coefficient, which quantifies the randomness of the accumulation of small pavement damage caused by normal traffic flow; W t The standard Brownian motion of the Wiener process is used to simulate the continuous high-frequency small shocks generated by randomly distributed light vehicles in the traffic flow. t is a burst discharge counting process, which is used to express the expected number of heavy vehicles passing by and causing significant damage per unit time, P t Obeying the Poisson distribution, ξ k is the single impact damage intensity, which obeys the Pareto distribution, τ k is the moment of impact, δ(t-τ k ) is expressed as a Dirac impulse function. Then, it provides a starting point for the iterative search process of the genetic algorithm, generates an initial solution set with sufficient diversity, and ensures that the genetic algorithm can cover a wide area of ​​the solution space. The specific formula is expressed as follows: S(0)=y1(0),y2(0),…,y P (0)={y i (0)|i=1,2…,P} Among them, S(0) represents the initial population, which includes the P solutions at the beginning of the genetic algorithm, y1(0), y2(0),…, y P (0) represents the independent solution in the P initial population, y i (0) represents the i-th solution in the initial population, i represents the index of the population number, ranging from 1 to P, P represents the population size, representing the total number of solutions in the initial population. By initializing the population, a set of initial solutions based on heuristic rules are generated. The initial solutions constitute the first generation population of the genetic algorithm. The diversity of the population directly affects the search ability of the algorithm and the quality of the final solution. The well-constructed initial population helps the algorithm to effectively explore the solution space and avoid falling into the local optimal solution. Secondly, the fitness of each solution is quantitatively evaluated to determine the quality of the solution and screen out the solution that contributes more to the optimization goal as a candidate for subsequent iterations. The specific formula is expressed as follows: f(y)=w1·Cost(y)+w2·QUL(y) Among them, f(y) is represented as the fitness function, which is used to evaluate the overall performance of solution y, y is represented as an independent solution in the population, w1 is represented as the cost weight coefficient, which is used to adjust the influence of maintenance cost in the fitness function, Cost(y) is represented as the maintenance cost function, which represents the total maintenance cost calculated according to solution y, w2 is represented as the quality weight coefficient, which is used to adjust the influence of maintenance quality in the fitness function, QUL(y) is represented as the maintenance quality function, which represents the maintenance quality calculated according to solution y. Through the fitness function f(y), the genetic algorithm can find a balance point so that the total cost is minimized while the maintenance quality reaches a level that meets the standards. Then, through the selection mechanism, excellent individual solutions are screened from the current population according to fitness. Excellent individual solutions will serve as the parents of the next generation of the genetic algorithm. The specific formula is expressed as follows: Among them, Probability(y j ) is expressed as the individual solution y j The probability of being selected represents a relative measure based on individual fitness and is used in the roulette wheel selection mechanism, y j It represents the jth individual solution in the current population, j represents the index of the individual solution in the current population, f(y j ) is represented by individual y j The fitness function of the algorithm is based on the principle of fitness selection, which makes the individual solutions with higher fitness have a high probability of being selected, so that the algorithm can select excellent solutions in the iterative process while maintaining the diversity of the overall algorithm. Finally, new genetic characteristics are introduced into the population through crossover and mutation operations to increase the diversity of the population and improve the probability of finding the global optimal solution. The specific formula is expressed as follows: Offspring=Crossover(y a ,y b ) and Offspring,变异a =Mutate(and a ) and Offspring,变异b =Mutate(and b ) Among them, Offspring represents the newly generated offspring individual, which represents the result of the crossover operation. Crossover(y a ,y b ) is represented as a crossover operation, which is used to solve y from two parent individuals. a and y b Exchange genetic information to produce new offspring, y a with y b It represents the parent individual solution, a and b represent the parent individual solution index, y Offspring,变异1 with y Offspring,变异2 It is represented as the offspring individual solution after the mutation operation, Mutate(y a ) and Mutate(y b ) is expressed as a mutation operation function. New genetic features are introduced through mutation operation to prevent premature convergence of individual solutions. Crossover operation creates offspring by combining the features of two parents. Mutation operation introduces new genetic diversity by randomly changing the features of offspring. By combining crossover operation and mutation operation, the algorithm can better search the solution space and find the optimal solution to the problem.

8. The digital integrated application system based on intelligent highway management and maintenance according to claim 1 is characterized in that: The data storage and management module ensures data security and availability by centrally storing and managing collected multimodal data, while supporting the storage and retrieval of big data, providing query and analysis functions for historical data, and providing data support for long-term management and planning of highways.

9. The digital integrated application system based on intelligent highway management and maintenance according to claim 1 is characterized in that: The user monitoring and decision-making module helps managers make correct decisions by providing an intuitive user interface and decision support functions, including data visualization tools and report generation functions, while ensuring that highway managers can easily use the system to query data, view analysis results and receive warning information.

Images