Highway network facility maintenance demand prediction and resource optimization allocation method and system

Abstract

This invention relates to the field of highway maintenance management technology, specifically a method and system for predicting highway network facility maintenance needs and optimizing resource allocation. The method involves acquiring data on the morphology, initial environment, and condition of target road sections; determining the initial equivalent extreme load using an inversion model; outputting initial predicted data using a prediction model and calculating the temporal residual between the predicted data and the measured data; if the residual exceeds a threshold, generating a decay compensation coefficient to correct the initial environmental characteristics and triggering iterations until the residual accuracy requirements are met, outputting the final environmental characteristics and extreme load equivalent; outputting high-precision predicted data based on the final parameters and identifying road sections requiring maintenance; and finally, performing resource optimization allocation according to maintenance priorities. The method utilizes real observation data to dynamically calibrate the implicit variables of environmental and material resistance coupling attenuation, avoiding prediction distortion caused by statically fixed boundary conditions, and improving the accuracy and economy of road network maintenance decisions.

Description

Technical Field

[0001] This invention relates to the field of highway maintenance management technology, specifically to a method and system for predicting highway network facility maintenance needs and optimizing resource allocation. Background Technology

[0002] As highway networks age, forecasting pavement maintenance needs and allocating resources are crucial for maintaining the overall service level of the network. Existing technologies typically acquire periodic monitoring data on the macroscopic technical condition of the pavement and combine this data with traditional machine learning or statistical regression models to establish a forward mapping relationship between pavement performance indicators and known traffic loads and typical climatic conditions. This allows for the prediction of apparent performance degradation trends on conventional road sections with comprehensive traffic monitoring data. Based on this prediction, planning can be developed, and preliminary maintenance plans and funding allocation schemes for shallow repair projects can be output within given budget constraints.

[0003] However, existing technologies are limited by the underlying architecture of unidirectional forward drive and open-loop static inference, and heavily rely on complete and explicit prior traffic load data as feature inputs, making it difficult to cope with complex road sections lacking dynamic weighing and monitoring equipment. Existing technologies cannot penetrate apparent indicators to trace the hidden extreme loads that cause disasters from the microscopic morphology of road surface defects, nor can they use prediction bias to perform closed-loop dynamic correction of physical parameters for long-term nonlinear dark decay of materials. The lack of reverse tracing and time-varying feedback mechanisms makes it easy for model predictions to generate serious cumulative errors over long service cycles, causing the backend resource allocation system to be completely detached from the actual underlying mechanical fatigue state, and thus unable to achieve scientific prediction of highway network facility maintenance needs and accurate allocation of resources. Summary of the Invention

[0004] To address the problems in related technologies, this invention provides a method and system for predicting highway network facility maintenance needs and optimizing resource allocation, thereby overcoming the aforementioned technical problems in existing related technologies.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a method for predicting highway network facility maintenance needs and optimizing resource allocation, comprising the following steps:

[0006] Acquire the disease morphology characteristics, initial environmental characteristics, and initial condition data of the target road section;

[0007] Based on the disease morphology data and initial environmental data, combined with the preset disease source tracing and inversion model, the initial equivalent extreme load data characterizing the posterior distribution characteristics of the extreme load is determined.

[0008] The initial equivalent extreme load equivalent data, initial environmental characteristic data, and initial condition data are input into a preset road surface decay prediction model for prediction, and the initial prediction data of the current node is output.

[0009] Obtain the measured status data of the target road segment at the current node, and calculate the temporal residual data between the initial prediction data and the measured status data;

[0010] Determine whether the time-series residual data is greater than a preset residual threshold;

[0011] If so, then a decay compensation coefficient is generated based on the time-series residual data to correct the initial environmental characteristic data. The inversion and prediction steps are re-executed based on the corrected environmental characteristic data until the time-series residual data is less than or equal to the residual threshold. Finally, the final environmental characteristic data and the final equivalent extreme load equivalent data are output.

[0012] The final environmental characteristic data, the final equivalent extreme load equivalent data, and the measured condition data are input into the pavement decay prediction model for prediction, and the final prediction data of the current node is output.

[0013] Determine whether each time-series predicted value in the final predicted data is greater than the preset maintenance threshold. If not, mark the target road segment as a road segment to be maintained.

[0014] After traversing all road segments within the target area, the maintenance priority sequence is determined based on the final predicted data corresponding to each road segment to be maintained, and resource optimization allocation is performed.

[0015] Preferably, determining the initial equivalent extreme load data characterizing the posterior distribution features of the extreme load includes the following steps:

[0016] The disease morphology feature data and the initial environmental feature data are subjected to feature fusion processing to generate a conditional semantic vector representing known physical boundary constraints;

[0017] Random sampling is performed in a pre-defined standard normal latent variable space to obtain a latent variable feature sequence for simulating unknown dynamic load variables;

[0018] Using the conditional semantic vector as the mapping guide condition, the latent variable feature sequence is input into the disease source tracing inversion model to perform inverse bijective transformation, and the implicit extreme load probability distribution data representing the posterior probability density of extreme load is output.

[0019] The integral expectation data of the implicit extreme load probability distribution data within a preset high confidence interval is calculated to obtain the initial equivalent extreme load equivalent data.

[0020] A source inversion model based on the morphological characteristics of road damage is proposed. By randomly sampling in a pre-defined standard normal variable space and combining semantic vector mapping, the posterior probability distribution of extreme loads is generated. This model can infer the unobserved historical real heavy loads from the road damage results that have already occurred, effectively making up for the shortcomings of traditional monitoring methods that are difficult to continuously record traffic extreme loads, and providing more realistic and reliable load benchmark data for subsequent decay prediction.

[0021] Preferably, calculating the time-series residual data between the initial predicted data and the measured condition data includes the following steps:

[0022] The initial equivalent extreme load equivalent data, initial environmental characteristic data, and initial condition data are input into a preset pavement decay prediction model for prediction, and the initial prediction data of the current node is output; wherein, the initial prediction data represents the multi-dimensional pavement performance index of the current node generated based on the physical decay law simulation.

[0023] Obtain measured condition data of the target road segment at the current node; the measured condition data represents the actual multi-dimensional pavement performance indicators of the current node obtained through on-site monitoring;

[0024] Calculate the normalized relative deviations of the initial prediction data and the measured data on each corresponding feature dimension;

[0025] The normalized relative deviation is weighted and aggregated according to the preset sensitivity weights of each feature dimension to obtain the time-series residual data between the initial prediction data and the measured data.

[0026] Preferably, the output of the final environmental characteristic data and the final equivalent extreme load equivalent data includes the following steps:

[0027] Determine whether the time-series residual data is greater than a preset residual threshold;

[0028] If not, it means that the current prediction result meets the preset residual accuracy requirements, and the initial environmental feature data and the initial equivalent extreme load equivalent data are respectively used as the final environmental feature data and the final equivalent extreme load equivalent data;

[0029] If so, a corresponding decay compensation coefficient is generated based on the deviation of the time-series residual data, and the parameters of the initial environmental feature data are updated using the decay compensation coefficient to obtain the corrected environmental feature data;

[0030] Using the corrected environmental feature data as a new input benchmark, the steps of determining the initial equivalent extreme load equivalent data, outputting the initial prediction data, and calculating the time series residual data are returned and executed repeatedly until the output time series residual data is less than or equal to the preset residual threshold. The environmental feature data and equivalent extreme load equivalent data corresponding to the current round are respectively used as the final environmental feature data and the final equivalent extreme load equivalent data.

[0031] The obtained corrected environmental feature data includes the following steps:

[0032] Calculate the difference between the time-series residual data and the preset residual threshold to obtain the out-of-bounds amplitude variable that characterizes the severity of the error;

[0033] The super-boundary amplitude variable is input into a preset nonlinear mapping function with boundary constraints to calculate the decay compensation coefficient that characterizes the compensation adjustment direction and step size.

[0034] Based on the decay compensation coefficient, the local components corresponding to each environmental feature dimension in the initial environmental feature data are corrected to obtain the corrected environmental feature data.

[0035] By calculating the multidimensional normalized relative deviation between the initial predicted value and the measured value, and introducing a nonlinear mapping function with saturation constraint terms to generate decay compensation coefficients, dynamic dimensionality reduction calculations are performed on environmental characteristic parameters such as macroscopic nominal resistance using posterior data from real observations. This effectively avoids prediction distortion caused by the static solidification of initial boundary conditions in the decay model, quantifies latent variables such as microscopic fatigue damage that are difficult to obtain directly, and makes the prediction results more consistent with the actual deterioration level of the road surface.

[0036] Preferably, the step of determining the maintenance priority sequence based on the final predicted data corresponding to each road segment to be maintained, and performing resource optimization allocation operations, includes the following steps:

[0037] The final environmental characteristic data, the final equivalent extreme load equivalent data, and the measured condition data are input into the road surface decay prediction model for prediction, and the final prediction data of the current node is output.

[0038] If all time-series predicted values ​​in the final predicted data are greater than the preset maintenance threshold, then the target road segment is determined to be a steady-state service road segment that does not require maintenance intervention, and the current monitoring strategy is maintained.

[0039] If not, the target road segment will be marked as a road segment requiring maintenance;

[0040] Select other road segments within the target area as target road segments and repeat the above operations until all road segments within the target area have been traversed. Then, extract the target time-series nodes in the final prediction data corresponding to each road segment to be maintained that are first less than or equal to the preset maintenance threshold.

[0041] Calculate the time span between the target time node and the current node to obtain the expected failure period of the road section to be maintained;

[0042] Based on the expected failure period and the decay gradient of the final predicted data, the maintenance priority sequence of each road segment to be maintained in the target area is determined; wherein, the road segment with the shorter expected failure period and the larger decay gradient has a higher maintenance priority.

[0043] Based on the maintenance priority time sequence, and combined with the final prediction data corresponding to each road segment to be maintained, resource optimization and allocation operations are performed.

[0044] The process of performing resource optimization allocation based on the maintenance priority time sequence and combined with the final prediction data corresponding to each road segment to be maintained includes the following steps:

[0045] Set the current iteration number to And the maximum number of iterations is Define a road segment maintenance resource optimization space, and randomly generate resources within that space. Each road segment maintenance resource optimization data point corresponds to a road segment maintenance plan data point, resulting in a road segment maintenance resource optimization dataset.

[0046] An objective function is constructed based on the maintenance priority time series and the final prediction data, and the fitness value of each road segment maintenance resource optimization data in the road segment maintenance resource optimization dataset is calculated based on the objective function.

[0047] The road segment maintenance resource optimization data in the aforementioned road segment maintenance resource optimization dataset is evenly divided into... For each road segment maintenance resource optimization data subset, the road segment maintenance resource optimization data with the highest fitness value in each road segment maintenance resource optimization data subset is selected as the current optimal solution in that road segment maintenance resource optimization data subset;

[0048] The maintenance resource optimization data of each road segment in the respective road segment maintenance resource optimization data subset is updated around the position of the corresponding current optimal solution;

[0049] The current optimal solution in each road segment maintenance resource optimization data subset is updated based on the center position of the corresponding road segment maintenance resource optimization data subset;

[0050] Remove the maintenance resource optimization data of the road segment with the lowest fitness value from each road segment maintenance resource optimization data subset, and use a random perturbation strategy to update the position of the removed road segment maintenance resource optimization data;

[0051] The fitness value of each road segment maintenance resource optimization data after location update is calculated according to the objective function. The road segment maintenance resource optimization data with the highest fitness value in each road segment maintenance resource optimization data subset is selected as the current optimal solution in that road segment maintenance resource optimization data subset. At the same time, if the fitness value of the road segment maintenance resource optimization data after location update is greater than the original fitness value, the new position is used to replace the original position; otherwise, the original position is retained.

[0052] Determine the Is it greater than or equal to the stated If the above Greater than or equal to the If the optimal road segment maintenance plan data is found to be the one with the highest fitness value, then the road segment maintenance plan data corresponding to the optimized road segment maintenance resource data will be output as the optimal road segment maintenance plan data for the target area; otherwise, the iteration will continue until the optimal plan data for the target area is found to be the one with the highest fitness value. Greater than or equal to the .

[0053] By constructing a fitness function that includes expected failure time, decay gradient, and systematic weights, and combining it with an intelligent optimization algorithm, a maintenance plan that fits the current situation is matched from the maintenance strategy database. During the solution process, a nonlinear penalty mechanism is introduced for solutions that exceed the resource limit, ensuring that under budget constraints, the most effective and economical combination of maintenance plans can be automatically matched from the database. This achieves the optimal allocation of maintenance funds in time and space dimensions and improves the scientific allocation capability of highway network facility maintenance resources in complex decision-making environments.

[0054] The present invention also includes a system for predicting the maintenance needs of highway network facilities and optimizing resource allocation, comprising a data acquisition module, an initial equivalent extreme load equivalent measurement module, a time-series residual calculation module, an environmental characteristic data correction module, and a resource optimization allocation module;

[0055] The data acquisition module is used to acquire the disease morphology characteristics data, initial environmental characteristics data, and initial condition data of the target road section;

[0056] The initial equivalent extreme load equivalent measurement module is used to determine the initial equivalent extreme load equivalent data that characterizes the posterior distribution characteristics of the extreme load based on the disease morphology feature data and the initial environmental feature data combined with the preset disease source tracing inversion model.

[0057] The temporal residual calculation module is used to input the initial equivalent extreme load equivalent data, initial environmental characteristic data and initial condition data into a preset pavement decay prediction model for prediction, and output the initial prediction data of the current node; obtain the measured condition data of the target road segment at the current node, and calculate the temporal residual data between the initial prediction data and the measured condition data.

[0058] An environmental feature data correction module is used to determine whether the time-series residual data is greater than a preset residual threshold. If so, a decay compensation coefficient is generated based on the time-series residual data to correct the initial environmental feature data. The inversion and prediction steps are re-executed based on the corrected environmental feature data until the time-series residual data is less than or equal to the residual threshold. Finally, the final environmental feature data and the final equivalent extreme load equivalent data are output.

[0059] The resource optimization and allocation module is used to input the final environmental characteristic data, the final equivalent extreme load equivalent data, and the measured condition data into the pavement decay prediction model for prediction, and output the final prediction data of the current node; determine whether each time-series prediction value in the final prediction data is greater than the preset maintenance threshold; if not, mark the target road segment as a road segment to be maintained; after traversing all road segments in the target area, determine the maintenance priority time sequence according to the final prediction data corresponding to each road segment to be maintained, and perform resource optimization and allocation operations.

[0060] By employing the above technical solution, the present invention provides a method and system for predicting highway network facility maintenance needs and optimizing resource allocation, which has at least the following beneficial effects:

[0061] 1. This invention integrates disease source tracing and inversion, digital twin dynamic calibration, and multi-objective resource optimization to form a complete full-cycle technology system. First, it inverts the implicit load from the terminal disease morphology, then uses the temporal residuals of measured data and predicted values ​​to drive the iterative correction of environmental characteristic parameters, and finally, based on the calibrated high-precision prediction results, it performs intelligent optimization configuration in combination with a preset strategy database. It breaks through the limitations of traditional maintenance relying on static models or single experience, and realizes closed-loop control of the entire process from physical entity state to digital precision simulation, and then to efficient resource output.

[0062] 2. This invention proposes a source inversion model based on the morphological characteristics of road damage. By randomly sampling in a pre-defined standard normal variable space and combining semantic vector mapping to generate the posterior probability distribution of extreme loads, it can infer the unobserved historical real heavy loads from the road damage results that have occurred. This effectively makes up for the shortcomings of traditional monitoring methods that are difficult to continuously record traffic extreme loads, and provides more realistic and reliable load benchmark data for subsequent decay prediction.

[0063] 3. This invention generates decay compensation coefficients by calculating the multidimensional normalized relative deviation between the initial predicted value and the measured value, and by introducing a nonlinear mapping function with saturation constraint terms; it uses posterior data from real observations to dynamically reduce the dimensionality of environmental characteristic parameters such as macroscopic nominal resistance; it effectively avoids prediction distortion caused by the static solidification of initial boundary conditions in the decay model, quantifies implicit variables such as microscopic fatigue damage that are difficult to obtain directly, and makes the prediction results more consistent with the actual deterioration level of the road surface.

[0064] 4. This invention constructs a fitness function that includes expected failure time, decay gradient, and systematic weights, and combines it with an intelligent optimization algorithm to match maintenance schemes that meet the current situation from a maintenance strategy database. During the solution process, a nonlinear penalty mechanism is introduced for solutions that exceed the resource limit, ensuring that under budget constraints, the most effective and economical combination of maintenance schemes can be automatically matched from the database. This achieves the optimal allocation of maintenance funds in time and space dimensions and improves the scientific allocation capability of highway network facility maintenance resources in complex decision-making environments. Attached Figure Description

[0065] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0066] Figure 1 A flowchart of the method for predicting highway network facility maintenance needs and optimizing resource allocation provided by the present invention;

[0067] Figure 2 A schematic diagram of the modules of the highway network facility maintenance demand prediction and resource optimization allocation system provided by the present invention. Detailed Implementation

[0068] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0069] Exemplary method:

[0070] To address the limitations of existing technologies, which fail to penetrate surface indicators to trace the hidden extreme loads causing road surface defects from microscopic morphology, and also lack the ability to utilize prediction biases for closed-loop dynamic correction of physical parameters based on long-term nonlinear dark decay of materials, the absence of reverse tracing and time-varying feedback mechanisms leads to severe cumulative errors in model predictions over long service periods. This causes the backend resource allocation system to become completely detached from the actual underlying mechanical fatigue state, thus hindering the scientific prediction of highway network facility maintenance needs and the accurate allocation of optimized resources. This embodiment proposes a method for predicting highway network facility maintenance needs and optimizing resource allocation. Figure 1 As shown, the method includes the following steps:

[0071] Acquire the disease morphology characteristics, initial environmental characteristics, and initial condition data of the target road section;

[0072] Based on the disease morphology data and initial environmental data, combined with the preset disease source tracing and inversion model, the initial equivalent extreme load data characterizing the posterior distribution characteristics of the extreme load is determined.

[0073] The initial equivalent extreme load equivalent data, initial environmental characteristic data, and initial condition data are input into a preset road surface decay prediction model for prediction, and the initial prediction data of the current node is output.

[0074] Obtain the measured status data of the target road segment at the current node, and calculate the temporal residual data between the initial prediction data and the measured status data;

[0075] Determine whether the time-series residual data is greater than a preset residual threshold;

[0076] If so, then a decay compensation coefficient is generated based on the time-series residual data to correct the initial environmental characteristic data. The inversion and prediction steps are re-executed based on the corrected environmental characteristic data until the time-series residual data is less than or equal to the residual threshold. Finally, the final environmental characteristic data and the final equivalent extreme load equivalent data are output.

[0077] The final environmental characteristic data, the final equivalent extreme load equivalent data, and the measured condition data are input into the pavement decay prediction model for prediction, and the final prediction data of the current node is output.

[0078] Determine whether each time-series predicted value in the final predicted data is greater than the preset maintenance threshold. If not, mark the target road segment as a road segment to be maintained.

[0079] After traversing all road segments within the target area, the maintenance priority sequence is determined based on the final predicted data corresponding to each road segment to be maintained, and resource optimization allocation is performed.

[0080] The aforementioned disease morphology feature data aims to characterize the physical deformation of the target road segment after loading. Specifically, it includes spatial topological feature data containing at least one disease morphology feature within the target road segment, which is used to invert the equivalent extreme load at the road segment level.

[0081] The initial environmental characteristic data is intended to characterize the intrinsic physical resistance and external natural service conditions of the target road segment. Specifically, it includes multi-dimensional coupled characteristic data containing the mechanical properties of the target road segment materials and spatiotemporal climate parameters. This data is used to provide the physical constraint boundary for inverting the equivalent extreme load at the road segment level and serves as the environmental loss benchmark for time-series decay prediction.

[0082] The initial condition data is intended to characterize the overall macro-health level of the target road segment at the previous historical monitoring node. Specifically, it is quantitative benchmark data containing several comprehensive performance evaluation indicators of the target road segment, which is used as the benchmark starting point for evolution and deduction, and to predict and output the initial prediction data of the current node.

[0083] The comprehensive performance evaluation indicators include at least the road surface damage index, road surface structural strength, road rutting depth, and smoothness index.

[0084] Among them, the preset residual threshold refers to the maximum allowable deviation between the model prediction result and the measured data. Specifically, it can be set according to the road network prediction accuracy requirements or historical error statistics, and is used to trigger the closed-loop iterative correction condition of the road surface decay prediction model parameters.

[0085] Among them, the preset maintenance threshold refers to the minimum service limit at which the pavement performance deteriorates to the point where engineering intervention is necessary. Specifically, it can be set according to the industry's highway maintenance technical specifications or operational safety baseline. It is used to anchor the expected failure node in the predicted time-series deterioration trajectory and trigger the calculation of maintenance priority.

[0086] The process of determining the initial equivalent extreme load data characterizing the posterior distribution features of the extreme load includes the following steps:

[0087] The disease morphology feature data and the initial environmental feature data are subjected to feature fusion processing to generate a conditional semantic vector representing known physical boundary constraints;

[0088] Random sampling is performed in a pre-defined standard normal latent variable space to obtain a latent variable feature sequence for simulating unknown dynamic load variables;

[0089] Using the conditional semantic vector as the mapping guide condition, the latent variable feature sequence is input into the disease source tracing inversion model to perform inverse bijective transformation, and the implicit extreme load probability distribution data representing the posterior probability density of extreme load is output.

[0090] The integral expectation of the implicit extreme load probability distribution data within a preset high confidence interval is calculated to obtain the initial equivalent extreme load data; the calculation formula is as follows:

[0091] ,

[0092] in, This represents the calculated initial equivalent extreme load equivalent data. This represents a continuous independent variable representing the physical intensity of the extreme load corresponding to the probability distribution data of the implicit extreme load. This represents the probability distribution data of hidden extreme loads, specifically represented by a probability density function curve. and These represent the upper and lower boundaries of the preset high-confidence interval, respectively.

[0093] The preset high confidence interval is used to filter out outliers caused by non-physical long-tail effects generated during the reverse mapping transformation of the disease source inversion model, so as to ensure that the final determined initial equivalent extreme load equivalent data can accurately coincide with the real physical and mechanical boundary of the target road section while having statistical robustness.

[0094] Preferably, the preset high confidence interval is set by setting a preset significance level. (For example, 0.05, which can be adjusted reasonably according to the actual situation) to define the tolerance limit for non-physical long-tail noise generated by the reverse mapping, and based on the cumulative distribution characteristics of the implicit extreme load probability distribution data, respectively cut off the cumulative probabilities on both sides of the probability density curve. The extreme regions are thus determined to satisfy the integral area within the interval. The lower and upper boundaries of the boundary precisely lock the load calculation domain within a physically effective area that contains 95% of the core probability distribution;

[0095] Specifically, feature fusion processing involves introducing a cross-attention mechanism or a multimodal cascade operator to achieve deep coupling between the geometric damage features of the morphology of the disease and the mechanical boundary information in the initial environmental features, thereby suppressing the failure of physical constraints caused by the lack of single-source features.

[0096] The standard normal latent variable space is a set of mathematical features that follow a multivariate standard normal distribution, constructed to address the uncertainty of extreme loads. In the physical and mechanical mechanisms of road damage, dynamic factors such as the vehicle's speed, braking impact force, and tire contact eccentricity are latent variables that are unobservable after the incident. Through this standard normal latent variable space, the aforementioned extremely chaotic unknown dynamic physical disturbances are uniformly and quantitatively encoded into a standard Gaussian probability distribution that is mathematically easy to sample. By performing random sampling within this space, such as Monte Carlo sampling, all possible unknown dynamic driving states are mathematically exhaustively simulated. Furthermore, through the inverse mapping of a reversible neural network, these dimensionless random seeds are restored to the real extreme load probability distribution in the physical world.

[0097] Furthermore, the disease tracing and inversion model can employ a conditionally reversible neural network model; the model construction includes the following steps:

[0098] Several sets of sample feature data characterizing the relationship between the evolution state of road segment defects and historical extreme loads were collected by using accelerated loading test equipment and finite element forward simulation platform to obtain the first sample dataset; the first sample data in the first sample dataset includes at least pavement defect morphology data, environmental feature data and corresponding extreme load data.

[0099] Construct an initial conditional reversible neural network model and set the first training data ratio, such as 8:2 or 7.5:2.5, which can be adjusted reasonably according to the actual situation.

[0100] The first sample dataset is divided according to the first training data ratio to obtain the first training dataset and the first test dataset.

[0101] Set a first training error threshold, such as 5%-10%, which can be adjusted reasonably according to the actual situation; input the training data in the first training dataset into the initial conditional reversible neural network model for training, and continuously adjust the parameters of the initial conditional reversible neural network model according to the training results until the training error is less than the first training error threshold or the number of training times is greater than the first maximum number of training times, and obtain the trained conditional reversible neural network model.

[0102] Set a first test precision, such as 90%-95%, which can be adjusted reasonably according to the actual situation; input the test data in the first test dataset into the trained conditional reversible neural network model for testing, calculate the accuracy of the test results, if the accuracy of the test results is greater than the first test precision, then the disease tracing and inversion model is obtained; otherwise, retrain until the accuracy of the test results is greater than the first test precision.

[0103] The structure of the initial condition reversible neural network model can be seen in Table 1 below:

[0104] Model Name Model type Model Structure Initial condition invertible neural network model Conditional invertible neural network model Multidimensional Conditional Fusion Module: Input Layer: Receives disease morphology feature data (as conditional input) and environmental feature data (as physical constraints); Deep Encoding Layer: Contains 3-5 fully connected layers, each with 128-256 neurons, using the LeakyReLU activation function to avoid gradient vanishing, and outputs a high-dimensional conditional semantic vector. Reversible Coupled Mapping Module: Cascades 8-12 affine coupling blocks. The internal sub-network uses a shallow multilayer perceptron to calculate translation and scaling coefficients, and configures the Swish activation function to ensure high-order continuous differentiability, constructing a strict bijective transformation between the physical extreme load space and the Gaussian latent variable space. Probability Inversion Module: Performs Monte Carlo sampling based on the latent variable space of the standard normal distribution, combines it with the conditional semantic vector, and outputs the latent extreme load probability distribution data through the inverse mapping link. Parameter Settings: Loss Function: Composed of a weighted composite of negative log-likelihood loss and a physical commonsense penalty term; Optimizer: AdamW (initial learning rate 1e-4) combined with a cosine annealing decay strategy; Initialization: Initialized using the Kaiming normal distribution.

[0105] Table 1

[0106] The calculation of the time-series residual data between the initial prediction data and the measured condition data includes the following steps:

[0107] The initial equivalent extreme load equivalent data, initial environmental characteristic data, and initial condition data are input into a preset pavement decay prediction model for prediction, and the initial prediction data of the current node is output; wherein, the initial prediction data represents the multi-dimensional pavement performance index of the current node generated based on the physical decay law simulation.

[0108] Obtain measured condition data of the target road segment at the current node; the measured condition data represents the actual multi-dimensional pavement performance indicators of the current node obtained through on-site monitoring;

[0109] The multidimensional pavement performance index includes at least a macroscopic pavement condition evaluation index, microscopic disease physical characteristics, and hidden structural mechanical indicators, and the measured condition data and the initial prediction data have the same feature dimensions.

[0110] Calculate the normalized relative deviations of the initial prediction data and the measured data on each corresponding feature dimension;

[0111] The normalized relative deviation is weighted and aggregated according to the preset sensitivity weights of each feature dimension to obtain the time-series residual data between the initial prediction data and the measured data.

[0112] Weighted aggregation refers to mapping multidimensional normalized relative deviations to a unified error quantization space. Specifically, it can be implemented using weighted accumulation operators, square root summation synthesis, or attention-weighted pooling algorithms. This is used to suppress random interference from fluctuations in insensitive features and enhance the contribution of key disease indicators to time series prediction correction.

[0113] The output of final environmental characteristic data and final equivalent extreme load data includes the following steps:

[0114] Determine whether the time-series residual data is greater than a preset residual threshold;

[0115] If not, it means that the current prediction result meets the preset residual accuracy requirements, and the initial environmental feature data and the initial equivalent extreme load equivalent data are respectively used as the final environmental feature data and the final equivalent extreme load equivalent data;

[0116] If so, a corresponding decay compensation coefficient is generated based on the deviation of the time-series residual data, and the parameters of the initial environmental feature data are updated using the decay compensation coefficient to obtain the corrected environmental feature data;

[0117] Using the corrected environmental feature data as a new input benchmark, the steps of determining the initial equivalent extreme load equivalent data, outputting the initial prediction data, and calculating the time series residual data are returned and executed repeatedly until the output time series residual data is less than or equal to the preset residual threshold. The environmental feature data and equivalent extreme load equivalent data corresponding to the current round are respectively used as the final environmental feature data and the final equivalent extreme load equivalent data.

[0118] The road surface degradation prediction model can adopt the GBDT model; the model construction includes the following steps:

[0119] By extracting historical pavement service monitoring data, several sets of sample feature data characterizing the road segment decay and evolution process are collected to obtain a second sample dataset; the sample feature data includes at least historical initial equivalent extreme load, historical initial environmental characteristics, historical initial condition data, and historical measured condition data of the corresponding decay nodes.

[0120] Construct the initial GBDT model and set the second training data ratio, such as 8:2 or 7.5:2.5, which can be adjusted reasonably according to the actual situation.

[0121] The second sample dataset is divided according to the second training data ratio to obtain the second training dataset and the second test dataset.

[0122] Set a second training error threshold, such as 5%-10%, which can be adjusted reasonably according to the actual situation. Input the training data in the second training dataset into the initial GBDT model for training. Continuously adjust the parameters of the initial GBDT model according to the training results until the training error is less than the second training error threshold, and obtain the trained GBDT model.

[0123] Set a second test precision, such as 90%-95%, which can be adjusted reasonably according to the actual situation. Input the test data in the second test dataset into the trained GBDT model for testing, and calculate the accuracy of the test results. If the accuracy of the test results is greater than the second test precision, the road surface decay prediction model is obtained; otherwise, retrain until the accuracy of the test results is greater than the second test precision.

[0124] The structure of the initial GBDT model can be seen in Table 2 below:

[0125] Model Name Model type Model Structure Initial GBDT model GBDT model Heterogeneous Feature Integration Module: Input Layer: Receives multi-source heterogeneous dimensional data; Feature Space Construction Layer: Utilizes the natural robustness of tree models, directly constructing a high-dimensional nonlinear optimization space containing continuous and discrete features without mandatory normalization. Regression Tree Base Learner Module: Network Backbone: Employs classification and regression trees as weak learners; Node Splitting Strategy: Performs feature selection and node splitting based on the maximization criterion of mean squared error reduction; Structural Constraints: Sets the maximum tree depth to 3-7 layers, constrains the minimum number of samples per leaf node, and achieves efficient nonlinear mapping of local multidimensional features. Residual Boosting Mapping Module: Serial Iteration Mechanism: Employs a forward step-by-step addition model, calculating the negative gradient of the current strong learner on the training samples in each iteration; Fitting Operator: Trains a new CART tree to fit the pseudo-residual sequence; Output Operator: Weighted summation of the outputs of hundreds of base learners, outputting the predicted data for the target time-series node. Composite optimization and parameter setting: Loss function: The Huber loss function, which is robust to monitoring outliers, is adopted to improve the stability of extreme road condition prediction; Parameter setting: The number of base estimators is set to 100-300; Optimization strategy: The initial learning rate is set to 0.05-0.1, a learning rate reduction mechanism is introduced to smooth the convergence process, and stochastic gradient boosting is configured to prevent model overfitting.

[0126] Table 2

[0127] Furthermore, obtaining the corrected environmental feature data includes the following steps:

[0128] Calculate the difference between the time-series residual data and the preset residual threshold to obtain the out-of-bounds amplitude variable that characterizes the severity of the error;

[0129] The super-boundary amplitude variable is input into a preset nonlinear mapping function with boundary constraints to calculate the decay compensation coefficient, which characterizes the compensation adjustment direction and step size; the calculation formula is as follows:

[0130] ,

[0131] in, Indicates the decay compensation coefficient. This represents a nonlinear mapping function used to smoothly map an infinitely large over-boundary error to the range (-1, 1). This represents the preset smoothing adjustment factor, used to control the error-sensitive mapping slope. This represents the system's extreme gain parameter;

[0132] Based on the decay compensation coefficient, the local components corresponding to each environmental feature dimension in the initial environmental feature data are corrected to obtain the corrected environmental feature data; the correction formula is as follows:

[0133] ,

[0134] in, This indicates the first element in the corrected environmental characteristic data. Local correction components corresponding to each environmental feature dimension This indicates the first element in the initial environmental feature data. Local components corresponding to each environmental feature dimension Indicates the first Sensitivity weights corresponding to each environmental feature dimension This represents the Hadamard product, which is the product of corresponding elements of a vector.

[0135] The system extreme value gain parameter represents the maximum limit of physical variation in environmental characteristic data that the system is allowed to undergo within a single iteration correction period.

[0136] Preferably, during engineering implementation, the system extreme gain parameter can be preset based on prior experience of the mechanical properties of road materials. For example, if the system extreme gain parameter is set to 0.2, the maximum reduction or surge in a single instance is allowed to not exceed 20% of the initial value. This parameter can effectively prevent the correction coefficient from linearly and unboundedly amplifying due to excessively large measured residuals.

[0137] Among them, the sensitivity weights corresponding to each environmental feature dimension represent the differentiated driving contribution of environmental disaster-causing factors of different dimensions to the current micro-deterioration of the road surface (i.e., the generation of the time-series residuals); when the damage of a certain road section is mainly induced by water damage, the system will assign a higher sensitivity weight value to the hydrological feature dimension, so that the local correction component of this dimension will be reduced more significantly; while for weakly correlated environmental dimensions, a sensitivity weight value close to 0 will be assigned to maintain the stability of the boundary conditions of this dimension.

[0138] Preferably, in specific implementation, the sensitivity weights can be statically weighted based on the standard specifications in the highway maintenance expert system; or they can be dynamically generated by pre-introducing machine learning algorithms such as random forests or gradient boosting trees to evaluate the feature importance of the historical regional disease database.

[0139] It should be noted that the correction of the initial environmental characteristic data in the closed-loop feedback is not a falsification of the objective meteorological or physical environment itself, but rather an equivalent parameter adjustment mechanism used to compensate for prediction bias caused by unobserved latent damage that is difficult to monitor.

[0140] Specifically, latent variables such as microscopic fatigue damage and microenvironmental erosion inside the road surface are highly concealed and difficult to continuously acquire by conventional detection equipment; the prediction model is based on the initial solidified health environment characteristic data for forward extrapolation, and cannot perceive the accumulation of these latent damages, which inevitably leads to a large prediction deviation between the prediction results and the actual road conditions.

[0141] Therefore, this invention extracts the temporal residuals characterizing this deviation and generates decay compensation coefficients to correct the initial environmental characteristic data. Its essence is to mathematically transform the internal latent damage that is difficult to quantify into the external constraint boundary degradation that the model can calculate. That is, by artificially correcting the environmental characteristic parameters, the disease evolution progress that is missed due to the lack of latent variables is forcibly compensated, thereby avoiding systematic prediction bias and making the model's inference trajectory accurately coincide with the physical deterioration state after the latent damage is superimposed.

[0142] The process of determining the maintenance priority sequence based on the final predicted data for each road segment to be maintained, and then performing resource optimization allocation, includes the following steps:

[0143] The final environmental characteristic data, the final equivalent extreme load equivalent data, and the measured condition data are input into the road surface decay prediction model for prediction, and the final prediction data of the current node is output.

[0144] If all time-series predicted values ​​in the final predicted data are greater than the preset maintenance threshold, then the target road segment is determined to be a steady-state service road segment that does not require maintenance intervention, and the current monitoring strategy is maintained.

[0145] If not, the target road segment will be marked as a road segment requiring maintenance;

[0146] Select other road segments within the target area as target road segments and repeat the above operations until all road segments within the target area have been traversed. Then, extract the target time-series nodes in the final prediction data corresponding to each road segment to be maintained that are first less than or equal to the preset maintenance threshold.

[0147] Calculate the time span between the target time node and the current node to obtain the expected failure period of the road section to be maintained;

[0148] Based on the expected failure period and the decay gradient of the final predicted data, the maintenance priority sequence of each road segment to be maintained within the target area is determined; wherein, the road segment with a shorter expected failure period and a larger decay gradient has a higher maintenance priority; the calculation formula for the maintenance priority sequence is as follows:

[0149] ,

[0150] in, , and They represent the first The maintenance priority sequence, expected failure period, and decay gradient of the final predicted data for each road section to be maintained; and These represent the decay gradient sensitivity weight and the failure time sensitivity weight, respectively.

[0151] The decay gradient sensitivity weight and failure duration sensitivity weight are used to balance the relative importance of the rate of disease deterioration and the remaining service life in the multi-attribute decision model.

[0152] Preferably, for high-grade highways or traffic-intensive road sections in the early to mid-stages of service, the system tends to increase the weight of the decay gradient sensitivity, such as... 0.6 The value is 0.4; for road networks with severely limited maintenance budgets or nearing the end of their service life, the system tends to increase the weight of the aforementioned failure-to-end sensitivity, such as... 0.4 The value is 0.6; the specific value can be adjusted according to the actual situation.

[0153] Based on the maintenance priority time sequence, and combined with the final prediction data corresponding to each road segment to be maintained, resource optimization and allocation operations are performed.

[0154] The process of performing resource optimization allocation based on the maintenance priority time sequence and combined with the final prediction data corresponding to each road segment to be maintained includes the following steps:

[0155] Set the current iteration number to And the maximum number of iterations is Define a road segment maintenance resource optimization space, and randomly generate resources within that space. Each road segment maintenance resource optimization data point corresponds to a road segment maintenance plan data point, resulting in a road segment maintenance resource optimization dataset.

[0156] An objective function is constructed based on the maintenance priority time series and the final prediction data. The fitness value of each road segment maintenance resource optimization data point in the road segment maintenance resource optimization dataset is then calculated using the objective function. The expression for the objective function is as follows:

[0157] ,

[0158] in, Indicates the first Fitness values ​​of maintenance resource optimization data for each road segment; This indicates the total number of road sections requiring maintenance within the target area; Indicates according to the first The comprehensive urgency index is determined by the maintenance priority sequence corresponding to each road section to be maintained; Indicates according to the first The final predicted data corresponding to the first road section to be maintained is matched with the first... The expected dynamic efficiency gain after maintenance is carried out on the road section maintenance plan data corresponding to the road section maintenance resource optimization data; This represents an adaptive dynamic penalty factor, used to impose a nonlinear penalty on illegal solutions that exceed the total resource limit; Indicates according to the first The final predicted data corresponding to the first road section to be maintained is matched with the first... The expected cost of maintenance for each road segment based on the road segment maintenance plan data corresponding to the road segment maintenance resource optimization data; This indicates the upper limit constraint of the total road network maintenance resources preset within the current planning period;

[0159] In the early stages of algorithm iteration, the adaptive dynamic penalty factor is set to a small initial value, allowing the system to temporarily tolerate some highly efficient combinations that slightly exceed the total resource limit, in order to maintain population diversity and guide the global search. As the iteration progresses into the middle and later stages, the value of the adaptive dynamic penalty factor increases rapidly and non-linearly with the number of iterations.

[0160] The road segment maintenance resource optimization data in the aforementioned road segment maintenance resource optimization dataset is evenly divided into... For each road segment maintenance resource optimization data subset, the road segment maintenance resource optimization data with the highest fitness value in each road segment maintenance resource optimization data subset is selected as the current optimal solution in that road segment maintenance resource optimization data subset;

[0161] The location of the road segment maintenance resource optimization data in each road segment maintenance resource optimization data subset is updated around the location of the corresponding current optimal solution; the location update formula is as follows:

[0162] ,

[0163] in, Indicates the first The first data subset of road section maintenance resource optimization The location after updating the location using the optimized maintenance resource data for each road section. Indicates the first The first data subset of road section maintenance resource optimization The current location of maintenance resource optimization data for each road section. Indicates the first The location of the current optimal solution in a subset of road segment maintenance resource optimization data. This indicates the impact factor of the current optimal solution on the maintenance resource optimization data of other road sections. This represents a random number that follows a uniform distribution between [0,1].

[0164] The current optimal solution in each road segment maintenance resource optimization data subset is updated based on the center position of the corresponding road segment maintenance resource optimization data subset; the position update formula is as follows:

[0165] ,

[0166] in, Indicates the first The location after updating the position of the current best solution in the subset of road segment maintenance resource optimization data. Indicates the first The central location of the subset of road section maintenance resource optimization data. This represents a control factor, used to control the impact of the center position on the position update of the current optimal solution;

[0167] Remove the maintenance resource optimization data of the road segment with the lowest fitness value from each road segment maintenance resource optimization data subset, and use a random perturbation strategy to update the position of the removed road segment maintenance resource optimization data;

[0168] The fitness value of each road segment maintenance resource optimization data after location update is calculated according to the objective function. The road segment maintenance resource optimization data with the highest fitness value in each road segment maintenance resource optimization data subset is selected as the current optimal solution in that road segment maintenance resource optimization data subset. At the same time, if the fitness value of the road segment maintenance resource optimization data after location update is greater than the original fitness value, the new position is used to replace the original position; otherwise, the original position is retained.

[0169] Determine the Is it greater than or equal to the stated If the above Greater than or equal to the If the optimal road segment maintenance plan data is found to be the one with the highest fitness value, then the road segment maintenance plan data corresponding to the optimized road segment maintenance resource data will be output as the optimal road segment maintenance plan data for the target area; otherwise, the iteration will continue until the optimal plan data for the target area is found to be the one with the highest fitness value. Greater than or equal to the .

[0170] The optimization space for road maintenance resources is set based on the scale of the road sections to be maintained within the target area, the combination dimension of candidate maintenance schemes, and the total resource boundary constraints.

[0171] Exemplary system:

[0172] Please see Figure 2 A system for predicting maintenance needs and optimizing resource allocation for highway network facilities includes a data acquisition module, an initial equivalent extreme load equivalent measurement module, a time-series residual calculation module, an environmental characteristic data correction module, and a resource optimization allocation module.

[0173] The data acquisition module is used to acquire the disease morphology characteristics data, initial environmental characteristics data, and initial condition data of the target road section;

[0174] The initial equivalent extreme load equivalent measurement module is used to determine the initial equivalent extreme load equivalent data that characterizes the posterior distribution characteristics of the extreme load based on the disease morphology feature data and the initial environmental feature data combined with the preset disease source tracing inversion model.

[0175] The temporal residual calculation module is used to input the initial equivalent extreme load equivalent data, initial environmental characteristic data and initial condition data into a preset pavement decay prediction model for prediction, and output the initial prediction data of the current node; obtain the measured condition data of the target road segment at the current node, and calculate the temporal residual data between the initial prediction data and the measured condition data.

[0176] An environmental feature data correction module is used to determine whether the time-series residual data is greater than a preset residual threshold. If so, a decay compensation coefficient is generated based on the time-series residual data to correct the initial environmental feature data. The inversion and prediction steps are re-executed based on the corrected environmental feature data until the time-series residual data is less than or equal to the residual threshold. Finally, the final environmental feature data and the final equivalent extreme load equivalent data are output.

[0177] The resource optimization and allocation module is used to input the final environmental characteristic data, the final equivalent extreme load equivalent data, and the measured condition data into the pavement decay prediction model for prediction, and output the final prediction data of the current node; determine whether each time-series prediction value in the final prediction data is greater than the preset maintenance threshold; if not, mark the target road segment as a road segment to be maintained; after traversing all road segments in the target area, determine the maintenance priority time sequence according to the final prediction data corresponding to each road segment to be maintained, and perform resource optimization and allocation operations.

[0178] Exemplary computer-readable medium:

[0179] Embodiments of this application may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps described in the "Exemplary Methods" section above according to the various embodiments of this application.

[0180] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0181] The basic principles of this application have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this application are merely examples and not limitations, and should not be considered as essential features of each embodiment of this application. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the application to the necessity of employing the aforementioned specific details for implementation.

[0182] The block diagrams of devices, apparatuses, devices, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0183] It should also be noted that in the apparatus, equipment, and methods of this application, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered as equivalent solutions of this application.

[0184] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this application. Therefore, this application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0185] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for predicting highway network facility maintenance needs and optimizing resource allocation, characterized in that, Includes the following steps: Acquire the disease morphology characteristics, initial environmental characteristics, and initial condition data of the target road section; Based on the disease morphology data and initial environmental data, combined with the preset disease source tracing and inversion model, the initial equivalent extreme load data characterizing the posterior distribution characteristics of the extreme load is determined. The initial equivalent extreme load equivalent data, initial environmental characteristic data, and initial condition data are input into a preset road surface decay prediction model for prediction, and the initial prediction data of the current node is output. Obtain the measured status data of the target road segment at the current node, and calculate the temporal residual data between the initial prediction data and the measured status data; Determine whether the time-series residual data is greater than a preset residual threshold; If so, then a decay compensation coefficient is generated based on the time-series residual data to correct the initial environmental characteristic data. The inversion and prediction steps are re-executed based on the corrected environmental characteristic data until the time-series residual data is less than or equal to the residual threshold. Finally, the final environmental characteristic data and the final equivalent extreme load equivalent data are output. The final environmental characteristic data, the final equivalent extreme load equivalent data, and the measured condition data are input into the pavement decay prediction model for prediction, and the final prediction data of the current node is output. Determine whether each time-series predicted value in the final predicted data is greater than the preset maintenance threshold. If not, mark the target road segment as a road segment to be maintained. After traversing all road segments within the target area, the maintenance priority sequence is determined based on the final predicted data corresponding to each road segment to be maintained, and resource optimization allocation is performed.

2. The method for predicting highway network facility maintenance needs and optimizing resource allocation according to claim 1, characterized in that, The process of determining the initial equivalent extreme load data characterizing the posterior distribution features of the extreme load includes the following steps: The disease morphology feature data and the initial environmental feature data are subjected to feature fusion processing to generate a conditional semantic vector representing known physical boundary constraints; Random sampling is performed in a pre-defined standard normal latent variable space to obtain a latent variable feature sequence for simulating unknown dynamic load variables; Using the conditional semantic vector as the mapping guide condition, the latent variable feature sequence is input into the disease source tracing inversion model to perform inverse bijective transformation, and the implicit extreme load probability distribution data representing the posterior probability density of extreme load is output. The integral expectation data of the implicit extreme load probability distribution data within a preset high confidence interval is calculated to obtain the initial equivalent extreme load equivalent data.

3. The method for predicting highway network facility maintenance needs and optimizing resource allocation according to claim 1, characterized in that, The calculation of the time-series residual data between the initial prediction data and the measured condition data includes the following steps: The initial equivalent extreme load equivalent data, initial environmental characteristic data, and initial condition data are input into a preset pavement decay prediction model for prediction, and the initial prediction data of the current node is output; wherein, the initial prediction data represents the multi-dimensional pavement performance index of the current node generated based on the physical decay law simulation. Obtain measured condition data of the target road segment at the current node; the measured condition data represents the actual multi-dimensional pavement performance indicators of the current node obtained through on-site monitoring; Calculate the normalized relative deviations of the initial prediction data and the measured data on each corresponding feature dimension; The normalized relative deviation is weighted and aggregated according to the preset sensitivity weights of each feature dimension to obtain the time-series residual data between the initial prediction data and the measured data.

4. The method for predicting highway network facility maintenance needs and optimizing resource allocation according to claim 1, characterized in that, The output of final environmental characteristic data and final equivalent extreme load data includes the following steps: Determine whether the time-series residual data is greater than a preset residual threshold; If not, it means that the current prediction result meets the preset residual accuracy requirements, and the initial environmental feature data and the initial equivalent extreme load equivalent data are respectively used as the final environmental feature data and the final equivalent extreme load equivalent data; If so, a corresponding decay compensation coefficient is generated based on the deviation amplitude of the time-series residual data, and the initial environmental feature data is corrected and compensated using the decay compensation coefficient to obtain the corrected environmental feature data; Using the corrected environmental feature data as a new input benchmark, the steps of determining the initial equivalent extreme load equivalent data, outputting the initial prediction data, and calculating the time series residual data are returned and executed repeatedly until the output time series residual data is less than or equal to the preset residual threshold. The environmental feature data and equivalent extreme load equivalent data corresponding to the current round are respectively used as the final environmental feature data and the final equivalent extreme load equivalent data.

5. The method for predicting highway network facility maintenance needs and optimizing resource allocation according to claim 4, characterized in that, The obtained corrected environmental feature data includes the following steps: Calculate the difference between the time-series residual data and the preset residual threshold to obtain the out-of-bounds amplitude variable that characterizes the severity of the error; The super-boundary amplitude variable is input into a preset nonlinear mapping function with boundary constraints to calculate the decay compensation coefficient that characterizes the compensation adjustment direction and step size. Based on the decay compensation coefficient, the local components corresponding to each environmental feature dimension in the initial environmental feature data are corrected to obtain the corrected environmental feature data.

6. The method for predicting highway network facility maintenance needs and optimizing resource allocation according to claim 1, characterized in that, The process of determining the maintenance priority sequence based on the final predicted data for each road segment to be maintained, and then performing resource optimization allocation, includes the following steps: The final environmental characteristic data, the final equivalent extreme load equivalent data, and the measured condition data are input into the road surface decay prediction model for prediction, and the final prediction data of the current node is output. If all time-series predicted values ​​in the final predicted data are greater than the preset maintenance threshold, then the target road segment is determined to be a steady-state service road segment that does not require maintenance intervention, and the current monitoring strategy is maintained. If not, the target road segment will be marked as a road segment requiring maintenance; Select other road segments within the target area as target road segments and repeat the above operations until all road segments within the target area have been traversed. Then, extract the target time-series nodes in the final prediction data corresponding to each road segment to be maintained that are first less than or equal to the preset maintenance threshold. Calculate the time span between the target time node and the current node to obtain the expected failure period of the road section to be maintained; Based on the expected failure period and the decay gradient of the final predicted data, the maintenance priority sequence of each road segment to be maintained in the target area is determined; wherein, the road segment with the shorter expected failure period and the larger decay gradient has a higher maintenance priority. Based on the maintenance priority time sequence, and combined with the final prediction data corresponding to each road segment to be maintained, resource optimization and allocation operations are performed.

7. The method for predicting highway network facility maintenance needs and optimizing resource allocation according to claim 6, characterized in that, The process of performing resource optimization allocation based on the maintenance priority time sequence and combined with the final prediction data corresponding to each road segment to be maintained includes the following steps: Set the current iteration number to And the maximum number of iterations is Define a road segment maintenance resource optimization space, and randomly generate resources within that space. Each road segment maintenance resource optimization data point corresponds to a road segment maintenance plan data point, resulting in a road segment maintenance resource optimization dataset. A target function is constructed based on the maintenance priority time sequence and the final prediction data, and the fitness value of the maintenance resource optimization data for each road segment is calculated based on the target function. The road section maintenance resource optimization dataset is evenly divided into... For each road segment maintenance resource optimization data subset, the road segment maintenance resource optimization data with the highest fitness value in each road segment maintenance resource optimization data subset is selected as the current optimal solution in that road segment maintenance resource optimization data subset; The maintenance resource optimization data of each road segment in the respective road segment maintenance resource optimization data subset is updated around the position of the corresponding current optimal solution; The current optimal solution in each road segment maintenance resource optimization data subset is updated based on the center position of the corresponding road segment maintenance resource optimization data subset; Remove the maintenance resource optimization data of the road segment with the lowest fitness value from each road segment maintenance resource optimization data subset, and use a random perturbation strategy to update the position of the removed road segment maintenance resource optimization data; The fitness value of each road segment maintenance resource optimization data after location update is calculated according to the objective function. The road segment maintenance resource optimization data with the highest fitness value in each road segment maintenance resource optimization data subset is selected as the current optimal solution in that road segment maintenance resource optimization data subset. At the same time, if the fitness value of the road segment maintenance resource optimization data after location update is greater than the original fitness value, the new position is used to replace the original position; otherwise, the original position is retained. Determine the Is it greater than or equal to the stated If the above Greater than or equal to the If the optimal road segment maintenance plan data is found to be the one with the highest fitness value, then the road segment maintenance plan data corresponding to the optimized road segment maintenance resource data will be output as the optimal road segment maintenance plan data for the target area; otherwise, the iteration will continue until the optimal plan data for the target area is found to be the one with the highest fitness value. Greater than or equal to the .

8. A system for implementing the method for predicting highway network facility maintenance needs and optimizing resource allocation as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire the disease morphology characteristics data, initial environmental characteristics data, and initial condition data of the target road section; The initial equivalent extreme load equivalent measurement module is used to determine the initial equivalent extreme load equivalent data that characterizes the posterior distribution characteristics of the extreme load based on the disease morphology feature data and the initial environmental feature data combined with the preset disease source tracing inversion model. The temporal residual calculation module is used to input the initial equivalent extreme load equivalent data, initial environmental characteristic data and initial condition data into the preset road surface decay prediction model for prediction, and output the initial prediction data of the current node. Obtain the measured status data of the target road segment at the current node, and calculate the temporal residual data between the initial prediction data and the measured status data; An environmental feature data correction module is used to determine whether the time-series residual data is greater than a preset residual threshold. If so, a decay compensation coefficient is generated based on the time-series residual data to correct the initial environmental feature data. The inversion and prediction steps are re-executed based on the corrected environmental feature data until the time-series residual data is less than or equal to the residual threshold. Finally, the final environmental feature data and the final equivalent extreme load equivalent data are output. The resource optimization and allocation module is used to input the final environmental characteristic data, the final equivalent extreme load equivalent data and the measured condition data into the pavement decay prediction model for prediction, and output the final prediction data of the current node; determine whether each time-series prediction value in the final prediction data is greater than the preset maintenance threshold; if not, mark the target road segment as a road segment to be maintained. After traversing all road segments within the target area, the maintenance priority sequence is determined based on the final predicted data corresponding to each road segment to be maintained, and resource optimization allocation is performed.

9. An electronic device comprising a memory and a processor, characterized in that: The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method as described in any one of claims 1-7.

10. A computer storage medium storing computer-executable instructions thereon, characterized in that: When the computer-executable instructions are executed by a processor, they implement the steps of the method as described in any one of claims 1-7.

Images