Tunnel portal road icing early warning method based on digital twinning

Abstract

The invention relates to a tunnel portal road icing early warning method based on digital twinning, and the method comprises the steps: generating a clean data set through integrating rock mass fracture, underground water level and meteorological monitoring data; filling the monitoring blind area based on the spatial self-correlation weight to form a dynamic water seepage distribution diagram; combining laser radar point cloud data to construct a freezing early warning digital twinborn body, and dynamically mapping a rock mass fracture network and water seepage space distribution; and finally, inputting the real-time weather and the road surface temperature field into the twinborn body, quantifying the icing probability of the tunnel portal pavement microporous structure through a thermodynamic model, and generating an icing risk spatial distribution diagram. The method solves the core defects of geology and tunnel portal road monitoring splitting, three-dimensional space perception deficiency, response lag and insufficient probability quantification in traditional early warning, realizes whole-process dynamic early warning from rock mass water seepage migration to road surface ice coagulation, improves the identification precision and response speed of tunnel portal road sudden ice coagulation risk, and improves the early warning efficiency. And a reliable basis is provided for intelligent traffic decision making.

Description

Technical Field

[0001] This invention relates to the field of road safety technology, specifically to a method for early warning of road icing at tunnel entrances based on digital twins. Background Technology

[0002] Icing at tunnel entrances is a major safety hazard in winter traffic. Influenced by multiple factors including geological seepage, sudden changes in microclimate, and the thermodynamic properties of the road surface, localized icing zones often form at tunnel entrances. These zones are characterized by their sudden onset and uneven spatial distribution, easily leading to vehicle loss of control accidents. Traditional icing warning methods primarily rely on meteorological station data and road surface temperature monitoring. However, these methods have inherent limitations: tunnel entrance icing is a phase transition process that occurs under specific temperature and humidity conditions after geological seepage migrates to the road surface. Current technology fails to reveal the complete physical chain of "rock seepage - road icing." Especially when seepage migrates through rock fissures, its spatial distribution is controlled by geological structure; once seepage reaches the road surface, the probability of icing is further regulated by both the road surface microporous structure and microclimate. This complex, multi-medium, multi-scale process makes accurate early warning difficult with traditional single-monitoring methods.

[0003] Current icing warning technologies suffer from the following key shortcomings: First, geological seepage monitoring and road surface monitoring are disconnected; rock fissure sensors can only provide point readings and cannot reflect the dynamic path of seepage water migrating from the rock mass to the road surface. Second, they lack the ability to perceive the entire three-dimensional space of tunnel entrances, especially leaving blind spots in monitoring high-risk areas such as shaded sides and structural joints. Third, existing models are mostly static empirical formulas that cannot integrate real-time geological changes and microclimate fluctuations, leading to a lag in response to sudden icing events. Fourth, warning results can only provide a simple risk level and fail to quantify the actual probability of icing at different locations, making it difficult to support precise de-icing decisions. These shortcomings result in a high false alarm rate and large positioning errors in existing systems, failing to meet the safety requirements of modern intelligent transportation. Summary of the Invention

[0004] Based on this, the purpose of this invention is to provide a tunnel entrance road icing early warning method based on digital twins that can connect the entire chain of geological water seepage and road icing, realize three-dimensional spatial dynamic monitoring of tunnel entrances, and accurately quantify the probability of icing.

[0005] The objective of this invention is achieved through the following solution:

[0006] In a first aspect, the present invention provides a method for early warning of road icing at tunnel entrances based on digital twins, comprising the following steps:

[0007] S1: Process the real-time data and historical seepage records collected by the rock fracture sensor, groundwater level gauge and meteorological station at the tunnel entrance, unify the time reference and spatial reference system of the multi-source data, eliminate measurement anomalies, and generate a standardized clean dataset.

[0008] S2: Process the discrete seepage monitoring points in the standardized clean dataset, calculate the spatial autocorrelation weights to fill the monitoring blind spots, expand the point seepage readings into a continuously distributed area seepage field, and generate a dynamic seepage distribution map.

[0009] S3: Based on the dynamic distribution map of seepage and the 3D point cloud data of the tunnel entrance obtained by lidar scanning, a digital twin model is dynamically constructed. The model details are adjusted according to the geological risk level, and the rock fracture network and seepage spatial distribution characteristics are integrated to generate a digital twin for icing early warning.

[0010] S4: Input the acquired real-time meteorological data stream and road surface temperature field into the icing warning digital twin for icing risk extrapolation processing, calculate the probability of water seepage freezing in the microporous structure of the road surface, and generate a spatial distribution map of icing warning at the tunnel entrance. The spatial distribution map is used to indicate the spatial differentiation characteristics of icing risk levels in various areas at the tunnel entrance.

[0011] In one embodiment, S1 of the tunnel entrance road icing early warning method based on digital twin provided by the present invention specifically includes the following steps:

[0012] S11: Process the raw data collected by the rock fracture width sensor, align the sampling time of different devices to the Coordinated Universal Time standard, and generate time-synchronized fracture data;

[0013] S12: Perform spatial reference system transformation on the data collected by the groundwater level gauge and meteorological station, transforming the local coordinate system of the equipment to a globally unified spatial reference coordinate system, and generating spatially aligned water level and meteorological data.

[0014] S13: Perform anomaly cleaning on time-synchronized crack data, spatially aligned water level and meteorological data, and stored historical seepage records. Remove offset data points that deviate from the main distribution characteristics using the three-standard-deviation criterion to generate a standardized clean dataset.

[0015] In one embodiment, S2 of the tunnel entrance road icing early warning method based on digital twin provided by the present invention specifically includes the following steps:

[0016] S21: Perform spatial autocorrelation modeling on the seepage monitoring points in the standardized clean dataset, calculate the spatial correlation weights between sensors based on the Gaussian semi-variogram, and generate a spatial weight distribution map.

[0017] S22: Perform grid node interpolation on discrete seepage monitoring points in the standardized clean dataset, fill in the seepage rate values ​​of monitoring blind areas based on the spatial weight distribution map, and generate a preliminary seepage distribution field;

[0018] S23: The edge effect correction is applied to the preliminary seepage distribution field. The data abrupt change in the tunnel entrance boundary area is smoothed by the boundary spline function to generate a dynamic seepage distribution map. The dynamic seepage distribution map is used to indicate the continuous spatial variation characteristics of the seepage volume and rate in the entire tunnel entrance area.

[0019] In one embodiment, S3 of the tunnel entrance road icing early warning method based on digital twin provided by the present invention specifically includes the following steps:

[0020] S31: The outlier filtering process is performed on the 3D point cloud data of the tunnel entrance obtained by LiDAR scanning. Noise data in the point cloud data is removed by statistical filtering algorithm to generate denoised 3D point cloud data.

[0021] S32: Perform geological risk zoning on the denoised 3D point cloud data, divide high-risk and low-risk areas based on the fracture density distribution and groundwater depth value of the seepage dynamic distribution map, and generate a geological risk level map.

[0022] S33: The geological risk level map is processed with detailed level selection. High-risk areas are generated using the Poisson surface reconstruction algorithm to generate centimeter-level precision models, while low-risk areas are generated using the Delaunay triangulation algorithm to generate meter-level simplified models. A digital twin for icing early warning is generated, which is used to dynamically map the coupling relationship between the rock mass structure characteristics at the tunnel entrance and the spatial distribution of seepage.

[0023] In one embodiment, S4 of the tunnel entrance road icing early warning method based on digital twin provided by the present invention specifically includes the following steps:

[0024] S41: Extract and process microclimate features from the real-time data stream collected by the meteorological monitoring system, calculate wind speed gradient, humidity change rate and air pressure fluctuation index, and generate a microclimate parameter set.

[0025] S42: Perform local temperature field reconstruction processing on the road surface temperature field acquired by the temperature sensor network, fill in the temperature values ​​of the uncovered areas by radial basis function interpolation, and generate a complete road surface temperature field.

[0026] S43: Input the microclimate parameter set and the complete road surface temperature field into the icing warning digital twin for phase change probability calculation. Based on the heat conduction model of the pavement material pore structure, solve the probability of water seepage and icing, and generate a spatial distribution map of road icing warning at the tunnel entrance.

[0027] Secondly, the present invention provides a tunnel entrance road icing early warning system based on digital twins, which is configured with the following modules:

[0028] The multi-source data processing module is used to process real-time data and historical seepage records collected by rock fracture sensors, groundwater level gauges and meteorological stations at the tunnel entrance, unify the time reference and spatial reference system of multi-source data, eliminate measurement anomalies, and generate standardized clean datasets.

[0029] The seepage field spatial expansion module is used to process discrete seepage monitoring points in the standardized clean dataset, calculate spatial autocorrelation weights to fill monitoring blind spots, expand point seepage readings into a continuously distributed area seepage field, and generate a dynamic seepage distribution map.

[0030] The digital twin model construction module is used to dynamically construct a digital twin model based on the dynamic distribution map of seepage and the three-dimensional point cloud data of the tunnel entrance obtained by lidar scanning. The model details are adjusted according to the geological risk level, and the rock fracture network and seepage spatial distribution characteristics are integrated to generate a digital twin for icing early warning.

[0031] The icing risk simulation module is used to input the acquired real-time meteorological data stream and road surface temperature field into the icing early warning digital twin for icing risk simulation processing, calculate the probability of water seepage freezing in the microporous structure of the road surface, and generate a spatial distribution map of icing early warning at the tunnel entrance. The spatial distribution map is used to indicate the spatial differentiation characteristics of the icing risk level in each area at the tunnel entrance.

[0032] Thirdly, this application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the above-mentioned methods for early warning of road icing at tunnel entrances based on digital twins.

[0033] Fourthly, this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-mentioned methods for early warning of road icing at tunnel entrances based on digital twins.

[0034] In summary, the tunnel entrance road icing early warning method based on digital twin provided in this application achieves deep fusion of multi-dimensional data, including rock fissures, groundwater levels, meteorological elements, and historical seepage records, at the data layer through the unification of spatiotemporal benchmarks and anomaly cleaning of multi-source heterogeneous data. This eliminates the fragmentation defects of traditional point-based monitoring. At the spatial modeling layer, the seepage field reconstruction technology based on spatial autocorrelation weights enables precise expansion from discrete monitoring points to continuous seepage distribution surfaces, addressing the monitoring gaps in blind areas such as the shaded side of the tunnel entrance and structural joints. At the digital twin layer, the feature-driven modeling method, which integrates laser point clouds and dynamic seepage distribution, constructs a coupled mapping relationship between the rock fissure network and the seepage path, enabling full-chain visual tracking of the migration of geological seepage to the road surface. At the risk estimation layer, the dynamic analysis of real-time meteorological and road temperature fields using a microporous structure thermodynamic model accurately quantifies the seepage icing probability gradient of different road surface materials, overcoming the static limitations of empirical formulas. The resulting spatial distribution map of icing risk enables high-resolution analysis of the micro-scale risk differentiation characteristics of the road surface at tunnel entrances. By integrating the coupling mechanism of geological seepage migration paths and the thermodynamic response of pavement microporous structures, the spatial boundaries of areas prone to sudden icing are located. This provides a spatial decision-making basis for active traffic control measures such as variable speed limit control and lane-level closures, thereby significantly compressing the response time window for icing events and improving the all-weather safety resilience of traffic infrastructure under special scenarios at tunnel entrances.

[0035] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0036] Figure 1 A flowchart illustrating a method for early warning of road icing at tunnel entrances based on digital twins, provided in an embodiment of this application;

[0037] Figure 2 This is a schematic diagram of the process for generating a digital twin for icing warning provided in an embodiment of this application;

[0038] Figure 3 This is a schematic diagram of a tunnel entrance road icing early warning system based on digital twin, provided as another embodiment of this application. Detailed Implementation

[0039] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0041] In one embodiment, such as Figure 1 As shown, a method for early warning of road icing at tunnel entrances based on digital twins is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0042] S1: Process real-time data and historical seepage records collected by rock fracture sensors, groundwater level gauges and meteorological stations at the tunnel entrance, unify the time reference and spatial reference system of multi-source data, eliminate measurement anomalies, and generate a standardized clean dataset.

[0043] Specifically, the system processes real-time data and historical seepage records collected by rock fracture sensors, groundwater level gauges, and weather stations at the tunnel entrance. Preferably, the system employs time synchronization technology to unify the time reference of data collected by each device, eliminating time discrepancies between devices through timestamp calibration and ensuring data consistency in the time dimension. Regarding the spatial reference system, the system uses a unified geodetic coordinate system as the standard, utilizing a coordinate transformation model to transform the spatial coordinates of data collected by different devices, achieving correspondence between the spatial locations of each monitoring point. For eliminating measurement anomalies, the system can use statistical methods for preliminary screening, calculating the mean and standard deviation of the dataset, identifying data exceeding a reasonable range as anomalies and removing them. If the proportion of anomalies exceeds a certain percentage, the system can use box plots for secondary verification, using the interquartile range as a benchmark to remove data points that do not meet the requirements, ultimately generating a standardized clean dataset containing parameters such as rock fracture seepage volume, groundwater level depth, air temperature, and humidity.

[0044] S2: Process the discrete seepage monitoring points in the standardized clean dataset, calculate the spatial autocorrelation weights to fill the monitoring blind spots, expand the point seepage readings into a continuously distributed area seepage field, and generate a dynamic seepage distribution map.

[0045] Specifically, the system processes discrete seepage monitoring points in the standardized clean dataset, calculates the spatial autocorrelation weights of these points, and uses spatial autocorrelation analysis to determine the spatial distribution characteristics of the monitoring point data. After determining the spatial distribution characteristics, the system calculates weights based on corresponding algorithms. During the weight allocation process, the orientation of rock fractures is incorporated, and a direction factor is introduced based on fracture orientation information obtained from previous geological surveys to improve the rationality of the weight calculation. For monitoring blind areas, the system uses co-kriging interpolation to fill in the gaps, using groundwater level changes and rock porosity as covariates. Multiple linear regression is used to determine the weights of these covariates, thus reasonably supplementing the blind area data. Through the above processing, the system expands point-like seepage readings into a continuously distributed planar seepage field, generating a dynamic seepage distribution map. This map undergoes cross-validation to ensure data accuracy meets subsequent modeling requirements. Simultaneously, the system captures abrupt changes in seepage rates through time series analysis, marking areas with abnormal seepage rate changes to provide a basis for subsequent risk assessment.

[0046] S3: Based on the dynamic distribution map of seepage and the 3D point cloud data of the tunnel entrance obtained by lidar scanning, a digital twin model is dynamically constructed. The model details are adjusted according to the geological risk level, and the rock fracture network and seepage spatial distribution characteristics are integrated to generate a digital twin for icing early warning.

[0047] Specifically, the system dynamically constructs a digital twin model based on the dynamic distribution map of seepage and the 3D point cloud data of the tunnel entrance obtained by lidar scanning. The 3D point cloud data obtained by lidar scanning needs to be preprocessed, with non-road points removed by a ground point filtering algorithm, and multi-station cloud registration completed using a point cloud registration algorithm to construct a 3D geometric model of the tunnel entrance. Preferably, during the dynamic construction of the digital twin model, the system can adopt hierarchical detailing technology, adjusting the model detail level according to the geological risk level. The geological risk level is divided by the risk matrix method, with high-risk areas corresponding to higher model detail levels to fully preserve the geometric features of the fractures, while medium- and low-risk areas use simplified model detail levels to balance model accuracy and computational efficiency.

[0048] The rock mass fracture network fusion is achieved through finite element mesh generation. The system treats fractures as permeable boundaries, and the permeability coefficient is dynamically calculated based on the fracture aperture. Simultaneously, the rock mass fracture network is coupled with the spatial distribution characteristics of seepage water. The coupling process employs Darcy's law and the mass conservation equation to solve simultaneously, ensuring the physical rationality of the coupling results. The final generated ice-freezing early warning digital twin comprises three layers: a physical layer, a data layer, and a model layer. The physical layer is a three-dimensional geometric model, the data layer contains real-time monitoring data interfaces, and the model layer is a seepage-heat transfer coupled model. The system supports dynamic updates of the digital twin and ensures efficient model loading through lightweight processing.

[0049] S4: Input the acquired real-time meteorological data stream and road surface temperature field into the icing warning digital twin for icing risk extrapolation processing, calculate the probability of water seepage freezing in the microporous structure of the road surface, and generate a spatial distribution map of icing warning at the tunnel entrance. The spatial distribution map is used to indicate the spatial differentiation characteristics of icing risk levels in various areas at the tunnel entrance.

[0050] Specifically, the system inputs the acquired real-time meteorological data stream and road surface temperature field into the icing early warning digital twin for icing risk simulation. The real-time meteorological data stream includes elements such as air temperature, humidity, wind speed, and solar radiation intensity, and is accessed by the digital twin via an API interface. The road surface temperature field is collected by distributed sensors and then input into the system. The icing risk simulation is based on a thermodynamic-seepage coupling model. The system calculates the migration time of seepage water in the microporous structure of the road surface. The calculation process combines Darcy's law and porosity parameters to ensure that the migration time calculation conforms to the seepage law. Subsequently, the time threshold for the road surface temperature to drop to the freezing point is calculated through the energy balance equation. The energy balance equation considers factors such as latent heat exchange, convective heat transfer, and radiative heat transfer to reflect the thermodynamic process of road surface temperature change.

[0051] Preferably, the icing probability calculation can employ a Logistic regression model. Input parameters include the time it takes for infiltrated water to reach the road surface, the difference between the road surface temperature and the freezing point, relative humidity, and wind speed. The system corrects the icing probability based on the actual values ​​of relative humidity and wind speed. Through the above processing, the system generates a spatial distribution map of road icing warnings at tunnel entrances. This spatial distribution map is expressed in a raster format and is used to indicate the spatial differentiation characteristics of icing risk levels in different areas of the tunnel entrance. Information related to high-risk areas is marked on the map, providing spatial positioning basis for accurate de-icing decisions and supporting the formulation and implementation of targeted de-icing measures.

[0052] In summary, the tunnel entrance road icing early warning method based on digital twin provided in this application achieves deep fusion of multi-dimensional data, including rock fissures, groundwater levels, meteorological elements, and historical seepage records, at the data layer through the unification of spatiotemporal benchmarks and anomaly cleaning of multi-source heterogeneous data. This eliminates the fragmentation defects of traditional point-based monitoring. At the spatial modeling layer, the seepage field reconstruction technology based on spatial autocorrelation weights enables precise expansion from discrete monitoring points to continuous seepage distribution surfaces, addressing the monitoring gaps in blind areas such as the shaded side of the tunnel entrance and structural joints. At the digital twin layer, the feature-driven modeling method, which integrates laser point clouds and dynamic seepage distribution, constructs a coupled mapping relationship between the rock fissure network and the seepage path, enabling full-chain visual tracking of the migration of geological seepage to the road surface. At the risk estimation layer, the dynamic analysis of real-time meteorological and road temperature fields using a microporous structure thermodynamic model accurately quantifies the seepage icing probability gradient of different road surface materials, overcoming the static limitations of empirical formulas. The resulting spatial distribution map of icing risk enables high-resolution analysis of the micro-scale risk differentiation characteristics of the road surface at tunnel entrances. By integrating the coupling mechanism of geological seepage migration paths and the thermodynamic response of pavement microporous structures, the spatial boundaries of areas prone to sudden icing are located. This provides a spatial decision-making basis for active traffic control measures such as variable speed limit control and lane-level closures, thereby significantly compressing the response time window for icing events and improving the all-weather safety resilience of traffic infrastructure under special scenarios at tunnel entrances.

[0053] In one embodiment, S1 of the tunnel entrance road icing early warning method based on digital twin provided by the present invention specifically includes the following steps:

[0054] S11: Process the raw data collected by the rock fracture width sensor, align the sampling time of different devices to the Coordinated Universal Time standard, and generate time-synchronized fracture data.

[0055] Specifically, the system receives raw data collected by rock fracture width sensors, which includes fracture width measurements and corresponding acquisition time information. Preferably, the system extracts the timestamps from each sensor's data and compares them with Coordinated Universal Time (UTC). For data with time discrepancies, the system adjusts it using a time calibration algorithm. This algorithm corrects the acquisition time of each data point based on the deviation record between the sensor's built-in clock and UTC. During time alignment, the system processes the differences in sampling frequencies between different sensors. For data with different sampling frequencies, interpolation is used to supplement missing time points, ensuring the continuity of all data in the time series. After time calibration, the system integrates the data, arranges it chronologically, and generates time-synchronized fracture data. Each record in this fracture data includes a fracture width measurement, a corresponding UTC timestamp, and sensor identification information, for subsequent correlation processing with other data.

[0056] S12: Perform spatial reference system transformation processing on the data collected by the groundwater level gauge and the meteorological station, transform the local coordinate system of the equipment to the globally unified spatial reference coordinate system, and generate spatially aligned water level and meteorological data.

[0057] Specifically, the system reads the local coordinate system parameters accompanying the data collected by the groundwater level gauge and the weather station. These parameters consist of the plane coordinates and elevation information calibrated during equipment installation. Preferably, the system can call a preset coordinate transformation model, which includes the rotation, translation, and scaling parameters required for the reference coordinate system transformation. The system inputs the local coordinate parameters of the groundwater level gauge and the weather station into the transformation model, completes the coordinate transformation through matrix operations, and maps the position information of the equipment in the local coordinate system to a globally unified spatial reference coordinate system.

[0058] During the conversion process, the system performs a consistency check on the conversion results. By comparing the coordinate deviations of known benchmark points in the same area before and after the conversion, it confirms that the conversion accuracy meets the preset requirements. If the check fails, the system recalculates the conversion model parameters and performs the conversion operation again until the accuracy requirements are met. Finally, the system generates water level and meteorological data with unified spatial location information, where the coordinates of each monitoring point correspond to their spatial location in the global benchmark coordinate system.

[0059] S13: Perform anomaly cleaning on time-synchronized crack data, spatially aligned water level and meteorological data, and stored historical seepage records. Remove offset data points that deviate from the main distribution characteristics using the three-standard-deviation criterion to generate a standardized clean dataset.

[0060] Specifically, the system integrates time-synchronized fracture data, spatially aligned water level and meteorological data, and stored historical seepage records to form a comprehensive dataset. The system groups the comprehensive dataset by data type and calculates the statistical characteristics of each group, including the mean and standard deviation. The system determines the reasonable distribution range of each group based on the three-standard-deviation criterion. The lower limit of the reasonable distribution range is the mean minus three standard deviations, and the upper limit is the mean plus three standard deviations. The system evaluates each data point in the dataset one by one. If a data point's value exceeds the reasonable distribution range of its corresponding group, it is identified as an offset data point deviating from the main distribution characteristics. Preferably, the system removes the identified offset data points from the dataset. For local data gaps caused by the removal of data points, the system can use linear interpolation of adjacent valid data to fill in the gaps, maintaining the integrity of the dataset. After the above processing, the system generates a standardized clean dataset covering fracture width, groundwater level, meteorological parameters, and historical seepage information. The data within the dataset meets the requirements for subsequent processing in terms of time, space, and numerical distribution.

[0061] In one embodiment, S2 of the tunnel entrance road icing early warning method based on digital twin provided by the present invention specifically includes the following steps:

[0062] S21: Perform spatial autocorrelation modeling on the seepage monitoring points in the standardized clean dataset, calculate the spatial correlation weights between sensors based on the Gaussian semi-variogram, and generate a spatial weight distribution map.

[0063] Specifically, the system calls upon seepage monitoring point data from a standardized cleaning dataset. This data includes the spatial coordinates of each monitoring point and corresponding seepage volume and rate information. The system performs spatial autocorrelation modeling on this data, extracting the spatial coordinates of all monitoring points and establishing a spatial location matrix. Preferably, the system introduces a Gaussian semivariogram, which describes the correlation between monitoring point data at different spatial locations and its variation with distance. The system calculates the spatial distance between any two monitoring points, inputs the distance parameter into the Gaussian semivariogram, and obtains the corresponding semivariogram value. Based on the semivariogram value, the system calculates the spatial correlation weights between sensors. The weight corresponds to the strength of the spatial correlation between monitoring points; the stronger the correlation, the larger the weight. The system associates the calculated spatial correlation weights with the corresponding spatial locations, generating a spatial weight distribution map. This distribution map visually presents the influence weights of each monitoring point within the tunnel entrance area on the surrounding space.

[0064] S22: Perform grid node interpolation on discrete seepage monitoring points in the standardized clean dataset, fill in the seepage rate values ​​in the monitoring blind area based on the spatial weight distribution map, and generate a preliminary seepage distribution field.

[0065] Specifically, the system uses discrete seepage monitoring point data from a standardized clean dataset, combined with a spatial weight distribution map, to perform grid node interpolation, dividing the tunnel entrance area into regular grid cells, with each grid node corresponding to a spatial location. For each grid node, the system determines the spatial correlation weight of its surrounding effective monitoring points based on the spatial weight distribution map; the range of effective monitoring points is determined by a spatial correlation threshold. The system weights the seepage rate values ​​of each effective monitoring point with their corresponding spatial weights to obtain an estimated seepage rate for that grid node. For grid nodes in monitoring blind spots—areas without directly surrounding monitoring points—the system expands the range of effective monitoring points and interpolates based on the weights of more distant monitoring points in the spatial weight distribution map to fill in the seepage rate values ​​in these areas. Through interpolation processing of all grid nodes, the system generates a preliminary seepage distribution field, which contains seepage rate data for all grid nodes across the entire tunnel entrance area, presenting the spatial distribution trend of seepage volume.

[0066] S23: The edge effect correction is applied to the preliminary seepage distribution field. The data abrupt change in the tunnel entrance boundary area is smoothed by the boundary spline function to generate a dynamic seepage distribution map. The dynamic seepage distribution map is used to indicate the continuous spatial variation characteristics of the seepage volume and rate in the entire tunnel entrance area.

[0067] Specifically, the system performs edge effect correction on the generated preliminary seepage distribution field. Because the monitoring points in the tunnel entrance boundary area are relatively sparse, the preliminary distribution field is prone to data abrupt changes at the boundary. The system introduces a boundary spline function, which smooths the seepage rate values ​​of the grid nodes in the boundary area by fitting data from effective monitoring points near the boundary. The system extracts the grid node data from the tunnel entrance boundary area in the preliminary seepage distribution field, inputs it along with the monitoring point data near the boundary into the boundary spline function, calculates the corrected seepage rate values, and replaces the original abrupt data points. During the correction process, the system ensures that the data changes in the boundary area conform to the overall distribution trend, avoiding local abnormal fluctuations. After edge effect correction, the system generates a dynamic seepage distribution map. This map displays the seepage volume and seepage rate across the entire tunnel entrance area in a continuous field distribution form, clearly indicating the magnitude of seepage volume, the speed of seepage rate, and their continuous spatial variation characteristics in different areas.

[0068] In one embodiment, such as Figure 2 As shown, S3 of the tunnel entrance road icing early warning method based on digital twin provided by the present invention specifically includes the following steps:

[0069] S31: The outlier filtering process is performed on the 3D point cloud data of the tunnel entrance obtained by LiDAR scanning. Noise data in the point cloud data is removed by statistical filtering algorithm to generate denoised 3D point cloud data.

[0070] Specifically, the system receives 3D point cloud data of the tunnel entrance obtained by LiDAR scanning. This data includes the 3D coordinate information of the rock surface at the tunnel entrance and the surrounding environment. The system preprocesses the 3D point cloud data, extracts the spatial coordinate parameters of all points, and establishes a point cloud data index. Preferably, the system can use a statistical filtering algorithm to filter out outliers. First, a neighborhood search range is set for each point, and the k-nearest neighbor algorithm is used to find neighboring points around each point. The number of neighboring points is dynamically adjusted according to the point cloud density. The system calculates the Euclidean distance between each point and its neighboring points, denoted as d. i (i = 1, 2, ..., n, where n is the number of neighborhood points), and calculate the mean μ and standard deviation σ of these distances using the formula:

[0071]

[0072] Where, d i Let represent the Euclidean distance between the target point and its i-th neighboring point, where n is the total number of neighboring points. Based on statistical results, the system determines the distance threshold T using the following formula:

[0073] T=μ+kσ

[0074] Here, k is a preset coefficient used to control the strictness of the threshold. When the average distance between a point and its neighboring points exceeds T, the point is determined to be noisy data. The system marks and removes the identified noisy points, and simultaneously re-indexes the remaining point cloud data to ensure data continuity. After processing, the system generates denoised 3D point cloud data, which retains the main geometric features of the tunnel entrance rock mass and eliminates outliers caused by measurement errors or environmental interference.

[0075] S32: Perform geological risk zoning on the denoised 3D point cloud data, and divide high-risk and low-risk areas based on the fissure density distribution and groundwater depth value of the seepage dynamic distribution map to generate a geological risk level map.

[0076] Specifically, the system calls upon denoised 3D point cloud data, and simultaneously imports a dynamic distribution map of seepage and groundwater level monitoring data. From the dynamic distribution map of seepage, it extracts fracture density distribution information, reflecting the density of rock fractures in different areas of the tunnel entrance, denoted as FD (number of fractures per unit area). From the groundwater level monitoring data, it obtains the groundwater level depth value D (vertical distance between the groundwater level and the road surface) at each location. The system spatially correlates the fracture density distribution with the groundwater level depth value, establishing a coupled analysis model between the two, and calculates the geological risk index RI using the formula:

[0077] RI = ω1·FD + ω2·(1 / D)

[0078] In this system, ω1 and ω2 are weighting coefficients, satisfying ω1 + ω2 = 1, and their values ​​are determined through multiple regression analysis of historical icing accident data. 1 / D is the reciprocal of the groundwater level depth, used to quantify the impact of water level depth on risk. The system zons the tunnel entrance area according to preset classification rules: when the RI (Risk Index) of a certain area reaches a set threshold, the area is classified as a high-risk area; when the RI is below another set threshold, the area is classified as a low-risk area. The system smooths the zoning boundaries using a spatial interpolation algorithm to avoid abrupt changes in zoning results, ultimately generating a geological risk level map. This map uses different identifiers to distinguish between high-risk and low-risk areas, clearly presenting the spatial distribution characteristics of geological risks at the tunnel entrance.

[0079] S33: The geological risk level map is processed with detailed level selection. High-risk areas are generated using the Poisson surface reconstruction algorithm to generate centimeter-level precision models, while low-risk areas are generated using the Delaunay triangulation algorithm to generate meter-level simplified models. A digital twin for icing early warning is generated, which is used to dynamically map the coupling relationship between the rock mass structure characteristics at the tunnel entrance and the spatial distribution of seepage.

[0080] Specifically, the system loads a geological risk level map and identifies the spatial extent of high-risk and low-risk areas. For high-risk areas, the system calls a subset of the corresponding area from the denoised 3D point cloud data and processes it using a Poisson surface reconstruction algorithm. This algorithm fits the point cloud data by constructing an implicit surface function φ(x,y,z), satisfying the Laplace equation:

[0081]

[0082] in, Here, φ(x,y,z) is the Laplacian operator, a scalar function in three-dimensional space, and its zero isosurface represents the reconstructed rock mass surface. This algorithm generates a three-dimensional model containing fine structures, accurately reflecting details such as fracture morphology and unevenness on the rock mass surface. For low-risk areas, the system extracts a subset of the point cloud from the corresponding area and processes it using the Delaunay triangulation algorithm. This algorithm forms a triangular mesh by connecting discrete points in the point cloud, satisfying the empty circle property, meaning that the circumcircle of any triangle does not contain other point cloud data points. The volume of the mesh cell is calculated using the formula:

[0083]

[0084] Where p1, p2, p3, and p4 are the vertex coordinates of the tetrahedral mesh. This algorithm simplifies secondary features of the rock mass surface, generating a simplified model that reduces data volume while maintaining basic geometry. The system stitches together the detailed model of the high-risk area and the simplified model of the low-risk area, ensuring continuity at the model joint through coordinate alignment, ultimately generating a digital twin for icing risk early warning. This digital twin integrates rock mass structural feature data and seepage spatial distribution data at the tunnel entrance, dynamically mapping the coupling relationship between the two, providing a digital physical carrier for subsequent icing risk simulation.

[0085] In one embodiment, S4 of the tunnel entrance road icing early warning method based on digital twin provided by the present invention specifically includes the following steps:

[0086] S41: Extract and process microclimate features from the real-time data stream collected by the meteorological monitoring system, calculate wind speed gradient, humidity change rate and air pressure fluctuation index, and generate a microclimate parameter set.

[0087] Specifically, the system receives real-time data streams collected by the meteorological monitoring system. These streams contain wind speed, humidity, air pressure, and corresponding timestamps at different altitudes. The system segments the real-time data into time series segments and extracts microclimate features in fixed time windows. For wind speed gradient calculation, the system selects wind speed monitoring values ​​from two adjacent altitudes, denoted as the lower-level wind speed v1 and the upper-level wind speed v2, with a corresponding altitude difference of Δz. The wind speed gradient G is then calculated using a formula. v :

[0088]

[0089] For example, in the calculation of the humidity change rate, the system calculates the time derivative of the humidity data H(t) at the same monitoring point, using the following formula:

[0090]

[0091] Where Δt is the sampling time interval. The air pressure fluctuation index is obtained by calculating the standard deviation of the air pressure sequence per unit time, as shown in the formula:

[0092]

[0093] Among them, P i This refers to the air pressure measurement value per unit time. The average air pressure during this time period is denoted by , and N is the number of sampling points during this time period. The system integrates the calculated wind speed gradient, humidity change rate, and air pressure fluctuation index to generate a microclimate parameter set, which includes the spatial coordinates of each monitoring point and the corresponding microclimate characteristic values.

[0094] S42: Perform local temperature field reconstruction processing on the road surface temperature field acquired by the temperature sensor network, and fill in the temperature values ​​of the uncovered areas by radial basis function interpolation to generate a complete road surface temperature field.

[0095] Specifically, the system acquires road surface temperature data collected by a temperature sensor network. This data includes temperature measurements and spatial coordinates of the sensor deployment points. The system performs spatial distribution analysis on the temperature data to identify areas not covered by sensors. For filling in temperature values ​​in uncovered areas, the system uses a radial basis function interpolation algorithm, selecting valid sensor data around the interpolation point as sample points. Let the coordinates of the interpolation point be x, and the coordinates of the sample points be x'. i (i = 1, 2, ..., m, where m is the number of sample points), temperature value is T i Radial basis function selection Where r = ||xx i ‖ represents the Euclidean distance between the interpolation point and the sample point, and ε is the shape parameter. The interpolation function expression is:

[0096]

[0097] Where, λ i The weighting coefficients are obtained by solving a system of linear equations. Confirmed. The system performs the above interpolation calculation on all uncovered areas, fuses the interpolation results with the sensor measurement data, and generates a complete road surface temperature field. This temperature field is presented in grid form, containing temperature values ​​at all locations in the entire tunnel entrance area, and cross-validation ensures that the interpolation error is within the allowable range, providing complete temperature boundary conditions for icing phase change analysis.

[0098] S43: Input the microclimate parameter set and the complete road surface temperature field into the icing warning digital twin for phase change probability calculation. Solve the water infiltration and freezing probability based on the heat conduction model of the pavement material pore structure to generate a spatial distribution map of the road icing warning at the tunnel entrance.

[0099] Specifically, the system inputs the microclimate parameter set and the complete road surface temperature field into the icing early warning digital twin, initiating the phase change probability calculation process. The calculation is based on a heat conduction model of the pavement material's pore structure, which considers the heat exchange process of water seepage within the pores. The heat conduction equation is:

[0100]

[0101] Where T is the infiltration temperature, t is time, α = k / (ρc) is the thermal diffusivity, k is the thermal conductivity of the pavement material, ρ is the infiltration density, c is the specific heat capacity of the infiltration, and Q is the heat exchange term of the microclimate environment, including the influence of wind speed and humidity on heat dissipation. The system obtains the variation law of infiltration temperature with time by solving the above equations. When the temperature drops to the freezing point, the icing probability P is calculated by combining the pavement microporous structure parameters, as shown in the formula:

[0102]

[0103] Where t0 is the time when the seepage water reaches the road surface, t f At the current moment, E is the activation energy for water to freeze, and k is the energy for water to freeze. B is the Boltzmann constant. The system performs the above calculations on each grid cell at the tunnel entrance, mapping the icing probability to spatial coordinates to generate a spatial distribution map of road icing warnings at the tunnel entrance. This map identifies the icing risk of each area with different probability intervals, directly reflecting the spatial differentiation characteristics of icing probability and providing a quantitative basis for de-icing decisions.

[0104] In the embodiments of this application, the system receives real-time data and historical seepage records collected by rock fracture sensors, groundwater level gauges, and weather stations, while simultaneously acquiring 3D point cloud data of the tunnel entrance scanned by lidar, forming an initial input information stream. The system then unifies the time reference and converts the spatial reference system for the multi-source data, generating a standardized clean dataset through anomaly cleaning. Based on this dataset, the system generates a continuous dynamic seepage distribution map through spatial autocorrelation modeling and grid interpolation. The dynamic seepage distribution map is fused with the 3D point cloud data, and the system constructs a digital twin for icing early warning through geological risk zoning and detail level selection. Subsequently, the system inputs... Real-time meteorological data streams and road surface temperature fields are fed into a digital twin. Phase transition probability calculations are used to generate a spatial distribution map of road icing warnings at tunnel entrances. This map reflects the risk level and spatial differentiation characteristics of icing in the vehicle traffic area at tunnel entrances. Based on this spatial distribution map, observers can accurately locate high-risk icing areas and deploy de-icing equipment in advance, prioritizing icing-prone road sections. It also provides a basis for traffic control, guiding vehicles to avoid high-risk areas by setting up warning signs or temporary traffic restrictions, reducing vehicle loss of control accidents caused by road icing, and ensuring vehicle safety during passage through tunnel entrances.

[0105] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0106] Based on the same inventive concept, this application also provides a digital twin-based tunnel entrance road icing warning device for implementing the aforementioned digital twin-based tunnel entrance road icing warning method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the digital twin-based tunnel entrance road icing warning device provided below can be found in the limitations of the digital twin-based tunnel entrance road icing warning method described above, and will not be repeated here.

[0107] Preferably, such as Figure 3 As shown, this invention provides a tunnel entrance road icing early warning system 500 based on digital twins, which is configured with the following modules:

[0108] The multi-source data processing module 510 is used to process real-time data and historical seepage records collected by rock fracture sensors, groundwater level gauges and meteorological stations at the tunnel entrance, unify the time reference and spatial reference system of multi-source data, eliminate measurement anomalies, and generate standardized clean datasets.

[0109] The seepage field spatial expansion module 520 is used to process discrete seepage monitoring points in the standardized clean dataset, calculate spatial autocorrelation weights to fill monitoring blind spots, expand point seepage readings into a continuously distributed area seepage field, and generate a dynamic seepage distribution map.

[0110] The digital twin model construction module 530 is used to dynamically construct a digital twin model based on the dynamic distribution map of seepage and the three-dimensional point cloud data of the tunnel entrance obtained by lidar scanning. The model details are adjusted according to the geological risk level, and the rock fracture network and seepage spatial distribution characteristics are integrated to generate a digital twin for icing early warning.

[0111] The icing risk simulation module 540 is used to input the acquired real-time meteorological data stream and road surface temperature field into the icing early warning digital twin for icing risk simulation processing, calculate the probability of water seepage freezing in the microporous structure of the road surface, and generate a spatial distribution map of icing early warning at the tunnel entrance. The spatial distribution map is used to indicate the spatial differentiation characteristics of the icing risk level in each area at the tunnel entrance.

[0112] Preferably, the multi-source data processing module 510 provided in this application is configured with the following units:

[0113] The fracture data time synchronization unit is used to process the raw data collected by the rock fracture width sensor, align the sampling time of different devices to the Coordinated Universal Time standard, and generate time-synchronized fracture data.

[0114] The data spatial alignment unit is used to perform spatial reference system transformation processing on the data collected by groundwater level gauges and meteorological stations, transforming the local coordinate system of the equipment to a globally unified spatial reference coordinate system, and generating spatially aligned water level and meteorological data.

[0115] The multi-source data anomaly cleaning unit is used to clean up anomaly data from time-synchronized crack data, spatially aligned water level and meteorological data, and stored historical seepage records. It removes offset data points that deviate from the main distribution characteristics by using the three-standard-deviation criterion, and generates a standardized clean dataset.

[0116] Preferably, the seepage field space expansion module 520 provided in this application is configured with the following units:

[0117] The spatial modeling unit for seepage monitoring points is used to perform spatial autocorrelation modeling on seepage monitoring points in a standardized clean dataset. It calculates the spatial correlation weights between sensors based on a Gaussian semi-variogram and generates a spatial weight distribution map.

[0118] The monitoring blind zone filling unit is used to perform grid node interpolation on discrete seepage monitoring points in the standardized clean data set, fill in the seepage rate values ​​of the monitoring blind zone according to the spatial weight distribution map, and generate a preliminary seepage distribution field.

[0119] The edge effect correction unit is used to correct the edge effect of the initial seepage distribution field. It smooths the data abrupt changes in the tunnel entrance boundary area through the boundary spline function and generates a dynamic seepage distribution map. The dynamic seepage distribution map is used to indicate the continuous spatial variation characteristics of the seepage volume and rate in the entire tunnel entrance area.

[0120] Preferably, the digital twin model building module 530 provided in this application is configured with the following units:

[0121] The 3D point cloud data denoising unit is used to filter out outliers in the 3D point cloud data of the tunnel entrance obtained by LiDAR scanning. It removes noise data in the point cloud data through statistical filtering algorithm to generate denoised 3D point cloud data.

[0122] The geological risk zone division unit is used to perform geological risk zoning processing on the denoised 3D point cloud data. Based on the fracture density distribution of the seepage dynamic distribution map and the groundwater level depth value, high-risk and low-risk areas are divided to generate a geological risk level map.

[0123] The digital twin detail level construction unit is used to perform detail level selection processing on the geological risk level map. High-risk areas are generated by the Poisson surface reconstruction algorithm to generate centimeter-level precision models, while low-risk areas are generated by the Delaunay triangulation algorithm to generate meter-level simplified models. This generates an ice freezing early warning digital twin, which is used to dynamically map the coupling relationship between the rock mass structure characteristics at the tunnel entrance and the spatial distribution of seepage.

[0124] Preferably, the icing risk simulation module 540 provided in this application is configured with the following units:

[0125] The microclimate feature extraction unit is used to extract and process microclimate features from the real-time data stream collected by the meteorological monitoring system, calculate wind speed gradient, humidity change rate and air pressure fluctuation index, and generate a microclimate parameter set.

[0126] The road surface temperature field reconstruction unit is used to perform local temperature field reconstruction processing on the road surface temperature field acquired by the temperature sensor network. It fills in the temperature values ​​of the uncovered areas by radial basis function interpolation to generate a complete road surface temperature field.

[0127] The icing risk probability calculation unit is used to input the microclimate parameter set and the complete road surface temperature field into the icing early warning digital twin for phase change probability calculation. Based on the heat conduction model of the pavement material pore structure, it solves the probability of water seepage and icing, and generates a spatial distribution map of icing early warning at the tunnel entrance.

[0128] In one embodiment, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method for early warning of road icing at tunnel entrances based on digital twins.

[0129] In one embodiment, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for early warning of road icing at tunnel entrances based on digital twins.

[0130] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0131] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0132] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for early warning of road icing at tunnel entrances based on digital twins, characterized in that, Includes the following steps: S1: Process the real-time data and historical seepage records collected by the rock fracture sensor, groundwater level gauge and meteorological station at the tunnel entrance, unify the time reference and spatial reference system of the multi-source data, eliminate measurement anomalies, and generate a standardized clean dataset. S2: Process the discrete seepage monitoring points in the standardized cleaning dataset, calculate the spatial autocorrelation weights to fill the monitoring blind spots, expand the point seepage readings into a continuously distributed area seepage field, and generate a dynamic seepage distribution map. S3: Based on the dynamic distribution map of seepage and the three-dimensional point cloud data of the tunnel entrance obtained by lidar scanning, a digital twin model is dynamically constructed. The model details are adjusted according to the geological risk level, and the rock fracture network and seepage spatial distribution characteristics are integrated to generate a digital twin for ice freezing early warning. S4: Input the acquired real-time meteorological data stream and road surface temperature field into the icing warning digital twin for icing risk extrapolation processing, calculate the probability of water seepage freezing in the microporous structure of the road surface, and generate a spatial distribution map of icing warning at the tunnel entrance. The spatial distribution map is used to indicate the spatial differentiation characteristics of icing risk levels in various areas at the tunnel entrance.

2. The method according to claim 1, characterized in that, S1 includes: S11: Process the raw data collected by the rock fracture width sensor, align the sampling time of different devices to the Coordinated Universal Time standard, and generate time-synchronized fracture data; S12: Perform spatial reference system transformation on the data collected by the groundwater level gauge and meteorological station, transforming the local coordinate system of the equipment to a globally unified spatial reference coordinate system, and generating spatially aligned water level and meteorological data. S13: Perform abnormal data cleaning processing on the time-synchronized crack data, the spatially aligned water level and meteorological data, and the stored historical seepage records. Remove offset data points that deviate from the main distribution characteristics by using the three-standard-deviation criterion to generate a standardized clean dataset.

3. The method according to claim 1, characterized in that, S2 includes: S21: Perform spatial autocorrelation modeling on the seepage monitoring points in the standardized clean dataset, calculate the spatial correlation weights between sensors based on the Gaussian semi-variogram, and generate a spatial weight distribution map. S22: Perform grid node interpolation on the discrete seepage monitoring points in the standardized clean data set, fill in the seepage rate values ​​of the monitoring blind area according to the spatial weight distribution map, and generate a preliminary seepage distribution field; S23: The preliminary seepage distribution field is corrected for edge effects. The data abrupt changes in the tunnel entrance boundary area are smoothed by the boundary spline function to generate a dynamic seepage distribution map. The dynamic seepage distribution map is used to indicate the continuous spatial variation characteristics of the seepage volume and rate in the entire tunnel entrance area.

4. The method according to claim 1, characterized in that, S3 includes: S31: The outlier filtering process is performed on the 3D point cloud data of the tunnel entrance obtained by LiDAR scanning. Noise data in the point cloud data is removed by statistical filtering algorithm to generate denoised 3D point cloud data. S32: Perform geological risk zoning processing on the denoised 3D point cloud data, and divide high-risk and low-risk areas according to the fracture density distribution and groundwater depth value of the seepage dynamic distribution map to generate a geological risk level map. S33: The geological risk level map is processed by selecting the level of detail. The high-risk area is generated by the Poisson surface reconstruction algorithm to generate a model with centimeter-level accuracy, and the low-risk area is generated by the Delaunay triangulation algorithm to generate a simplified model with meter-level accuracy. The icing early warning digital twin is generated. The icing early warning digital twin is used to dynamically map the coupling relationship between the rock mass structure characteristics at the tunnel entrance and the spatial distribution of seepage.

5. The method according to any one of claims 1, characterized in that, S4 includes: S41: Extract and process microclimate features from the real-time data stream collected by the meteorological monitoring system, calculate wind speed gradient, humidity change rate and air pressure fluctuation index, and generate a microclimate parameter set. S42: Perform local temperature field reconstruction processing on the road surface temperature field acquired by the temperature sensor network, fill in the temperature values ​​of the uncovered areas by radial basis function interpolation, and generate a complete road surface temperature field. S43: Input the microclimate parameter set and the complete road surface temperature field into the icing warning digital twin for phase transition probability calculation, solve the water seepage and freezing probability based on the heat conduction model of the pavement material pore structure, and generate a spatial distribution map of icing warning at the tunnel entrance.

6. A tunnel entrance road icing early warning system based on digital twins, characterized in that, The system includes: The multi-source data processing module is used to process real-time data and historical seepage records collected by rock fracture sensors, groundwater level gauges and meteorological stations at the tunnel entrance, unify the time reference and spatial reference system of multi-source data, eliminate measurement anomalies, and generate standardized clean datasets. The seepage field spatial expansion module is used to process the discrete seepage monitoring points in the standardized cleaning dataset, calculate the spatial autocorrelation weights to fill the monitoring blind spots, expand the point seepage readings into a continuously distributed area seepage field, and generate a dynamic seepage distribution map. The digital twin model construction module is used to dynamically construct a digital twin model based on the dynamic distribution map of seepage and the three-dimensional point cloud data of the tunnel entrance obtained by lidar scanning. The model details are adjusted according to the geological risk level, and the rock fracture network and seepage spatial distribution characteristics are integrated to generate a digital twin for ice freezing early warning. The icing risk simulation module is used to input the acquired real-time meteorological data stream and road surface temperature field into the icing early warning digital twin for icing risk simulation processing, calculate the probability of water seepage freezing in the microporous structure of the road surface, and generate a spatial distribution map of the road icing early warning at the tunnel entrance. The spatial distribution map is used to indicate the spatial differentiation characteristics of the icing risk level in each area of ​​the tunnel entrance.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 5.

Images