A cement stabilized macadam base cracking evolution prediction and evaluation system based on digital twinning

Abstract

The application provides a cement stabilized macadam base cracking evolution prediction and evaluation system based on digital twinning, and relates to the technical field of road engineering. The system comprises a data acquisition module for acquiring road state data of a road base; a first model construction module for constructing a base digital twinning model to obtain road state evolution process data; a second model construction module for constructing a road cracking prediction evolution model; a data prediction module for inputting current road state data and road state evolution process data into the road cracking prediction evolution model to output road cracking prediction data; and a cracking evaluation module for calculating base cracking risk indicators and determining a base cracking risk grade. The application can improve the reliability of cement stabilized macadam base crack prediction under complex environmental and long-term load conditions.

Description

Technical Field

[0001] This application relates to the field of road engineering technology, and in particular to a digital twin-based system for predicting and evaluating the cracking evolution of cement-stabilized crushed stone base courses. Background Technology

[0002] Cement-stabilized crushed stone base course, as an important component of semi-rigid road base courses, is characterized by high strength and good stability, and is widely used in highway and urban road construction. However, because cement-stabilized crushed stone base courses are typical semi-rigid materials, they are susceptible to stress accumulation and damage propagation due to cement hydration shrinkage, temperature and humidity changes, and repeated vehicle loads during long-term service, leading to base course cracks. These cracks not only reduce the overall load-bearing capacity of the road structure but may also extend upwards to the asphalt surface layer, affecting road service life and driving safety.

[0003] Currently, existing technologies for evaluating cracking in cement-stabilized crushed stone base courses typically employ methods such as on-site testing, laboratory experiments, and model analysis. For example, the degree of damage is assessed using apparent indicators like crack width, number of cracks, and crack area, or crack prediction models are established by combining parameters such as material strength and shrinkage performance. However, existing technologies still have certain shortcomings: Firstly, current testing methods primarily identify existing cracks, focusing on crack condition evaluation, making it difficult to promptly acquire information on the generation, propagation, and penetration of micro-cracks within the base course, thus failing to achieve continuous prediction of crack evolution trends. Secondly, cracking in cement-stabilized crushed stone base courses is influenced by multiple factors, including material properties, construction quality, environmental conditions, and traffic loads. Existing methods often use single indicators or fixed parameters for analysis, making it difficult to accurately describe the cumulative damage process under the coupled effects of multiple factors. Furthermore, existing prediction models usually rely on historical experimental data or empirical relationships. During actual road service, as environmental conditions and structural states change, model parameters are difficult to adjust in a timely manner, leading to deviations between predicted results and actual crack development, making it difficult to provide accurate and reliable decision-making basis for preventative road maintenance. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this application provides a digital twin-based system for predicting and evaluating the cracking evolution of cement-stabilized crushed stone base courses.

[0005] Firstly, this application provides a digital twin-based system for predicting and evaluating the cracking evolution of cement-stabilized crushed stone base courses, employing the following technical solution: A digital twin-based system for predicting and evaluating the cracking evolution of cement-stabilized crushed stone base courses, the system comprising: The data acquisition module is used to acquire road condition data of the actual road base layer, including base layer structural parameters, material parameters, construction parameters, environmental parameters, traffic load parameters, and base layer crack detection data; The first model construction module is used to construct a digital twin model of the road base layer based on the base layer structure parameters of the actual road base layer. The digital twin model of the base layer is used to establish the spatial mapping relationship and state association relationship between the actual road base layer and the virtual road base layer, and to map the road state data to the corresponding position of the virtual road base layer. The road state data of the virtual road base layer is updated according to the road state data collected at different time nodes to obtain the road state evolution process data of the virtual road base layer. The second model construction module is used to extract historical road state evolution characteristics based on the road state evolution process data of the virtual road base layer, and to construct a road cracking prediction evolution model based on the historical road state evolution characteristics. The data prediction module is used to acquire the current road status data in the grassroots digital twin model, input the current road status data and road status evolution process data into the road cracking prediction evolution model, and output road cracking prediction data. The cracking assessment module is used to calculate the base layer cracking risk index based on road cracking prediction data, compare the base layer cracking risk index with a preset risk level threshold, and determine the base layer cracking risk level.

[0006] Optionally, the first model building module further includes: The spatial division unit is used to establish a spatial coordinate model of the base layer based on the base layer structure parameters. The cement-stabilized crushed stone base layer is divided into multiple base layer state units according to the preset size. The mapping relationship between each base layer state unit and the corresponding virtual area of ​​the virtual road base layer is established. The road state evolution process data of the virtual road base layer is mapped to each base layer state unit to obtain the road state evolution process data of each base layer state unit.

[0007] Optionally, the traffic load parameters include vehicle trajectory and vehicle axle load data; The first model building module further includes: The load mapping unit is used to determine the range of vehicle load action area based on the overlap relationship between the vehicle's running trajectory and the spatial coordinates of the base layer state unit. Based on the range of vehicle load action area, it counts the number of load actions of each base layer state unit. Based on the vehicle axle load data and the number of load actions of each base layer state unit, it calculates the cumulative load damage of each base layer state unit. Based on the cumulative load damage of each base layer state unit, it updates the load action state of the corresponding area of ​​the virtual road base layer. It compares the cumulative load damage of each base layer state unit with a preset load threshold and marks the base layer state units with a crack sensitivity value that is greater than the preset load threshold.

[0008] Optionally, the base layer crack detection data includes road crack detection images; The first model building module further includes: The crack mapping unit is used to perform crack recognition processing on the road crack detection image, extract crack geometric feature data, including crack length, crack width, crack area and crack distribution density; map the crack geometric feature data to the base layer state unit corresponding to the virtual road base layer, and update the road state data of the area corresponding to the virtual road base layer.

[0009] Optionally, the second model building module includes: The graph construction unit is used to obtain the positional relationship between each base-level state unit, and to generate an initial state association graph based on the spatial adjacency relationship of each base-level state unit; to obtain historical road state evolution process data, to calculate the degree of correlation between crack state changes between adjacent base-level state units, to establish crack expansion association edges between adjacent base-level state units whose crack state association degree meets a preset association degree threshold, and to update the initial state association graph based on the crack expansion association relationship to obtain the base-level state graph network.

[0010] Optionally, the second model building module further includes: The feature encoding unit is used to acquire road state evolution process data corresponding to each basic state unit, extract time-series features from the road state evolution process data, and encode the time-series features to obtain the time-series state embedding vector corresponding to each basic state unit. The model building unit is used to construct a road cracking prediction evolution model based on the underlying state graph network and the temporal state embedding vectors of each graph node.

[0011] Optionally, the data prediction module specifically includes: The first unit is used to obtain the current road status data corresponding to each grassroots state unit in the grassroots digital twin model; The second unit is used to extract the historical road state evolution features corresponding to each basic state unit from the road state evolution process data; The third unit is used to input current road condition data and road condition evolution process data into the road cracking prediction evolution model. Based on the spatial correlation between each basic state unit, a basic state map network is constructed to extract the spatial correlation features between the basic state units. Based on the current road condition data and historical road condition evolution features, the spatial correlation features between the basic state units and the evolution features of road condition over time are extracted. Based on the spatial correlation features and time evolution features, predictions are made and road cracking prediction data is output. The road cracking prediction data includes crack prediction status data and crack propagation trend. Crack prediction status data includes crack prediction length, crack prediction area, and crack prediction growth rate. Crack propagation trend includes crack prediction number and crack propagation direction.

[0012] Optionally, the crack assessment module includes: The risk index calculation unit is used to calculate the crack severity index based on the crack prediction length and crack prediction area, the crack activity index based on the crack prediction growth rate, the crack number density index based on the crack prediction number, and the crack dispersion based on the crack extension direction. The crack severity index, crack activity index, crack number density index and crack dispersion are weighted and summed to obtain the base layer cracking risk index. The risk assessment unit compares the base layer cracking risk index with a first preset risk level threshold. If the base layer cracking risk index is greater than the preset risk level threshold, the actual road base layer cracking risk level is determined to be Level 3. If the base layer cracking risk index is not greater than the preset risk level threshold, it continues to determine whether the base layer cracking risk index is greater than a second preset risk level threshold. If it is, the actual road base layer cracking risk level is determined to be Level 2; otherwise, the actual road base layer cracking risk level is determined to be Level 1.

[0013] Optionally, the system further includes: The model update module is used to obtain the actual road cracking data corresponding to the prediction time point of the road cracking prediction data, compare the actual road cracking data with the road cracking prediction data, calculate the prediction deviation, and optimize the model parameters of the road cracking prediction evolution model based on the prediction deviation to obtain the optimized road cracking prediction evolution model.

[0014] Optionally, the system further includes: The road maintenance module is used to select the corresponding strategy from the preset maintenance strategy library based on the road cracking prediction data and the base cracking risk level, and output it as the target maintenance strategy.

[0015] In summary, this application includes at least one of the following beneficial technical effects: 1. This application constructs a digital twin-based prediction and assessment system for the crack evolution of cement-stabilized crushed stone base courses. This system integrates road base course condition perception, digital model construction, crack evolution prediction, risk level assessment, and maintenance decision-making, achieving intelligent management of the entire road base course process from condition monitoring to crack prediction and risk control. Compared to existing methods that rely on single-test data or empirical models for crack evaluation, this application can dynamically acquire information on the condition changes of the base course during its service life. By comprehensively considering the influence of multiple factors such as material properties, environmental factors, and traffic loads, it improves the accuracy of base course crack evolution prediction and risk assessment, providing a reliable basis for preventive road maintenance.

[0016] 2. This application establishes a spatial mapping and state association relationship between the actual road base layer and the virtual road base layer. It maps base layer structural parameters, material parameters, construction parameters, environmental parameters, traffic load parameters, and crack detection data to the virtual road base layer, and continuously updates the virtual model's state based on data collected at different time points. This allows the digital twin model to reflect the actual service state changes of the road base layer. This avoids the problem of traditional detection methods only obtaining state information at a single time point, improving the continuity and realism of the road base layer's state representation.

[0017] 3. This application divides the base course into multiple state units and constructs a base course state map network based on the spatial adjacency relationships between state units and historical road condition evolution data. It then utilizes a road cracking prediction and evolution model to extract spatial correlation and temporal evolution characteristics of the base course damage state, thereby predicting the generation, propagation, and development trends of cracks. Compared to existing methods that use single parameters or empirical relationships for prediction, this application fully considers the regional correlations during crack propagation, improving the accuracy of crack prediction results under complex service conditions.

[0018] 4. This application compares the actual road cracking data at the predicted time point with the predicted results, calculates the prediction deviation, and optimizes the road cracking prediction evolution model based on the prediction deviation, enabling the model to dynamically adjust as the actual road condition changes. Simultaneously, it combines the base layer cracking risk level with corresponding maintenance strategies, achieving feedback optimization between the prediction model and the actual road condition. This improves the adaptability of risk assessment results and the pertinence of maintenance decisions, helping to reduce the risk of further damage to the road structure and extend the road's service life. Attached Figure Description

[0019] Figure 1 This is a flowchart of an embodiment of this application; Figure 2 This is a flowchart of the first model construction module 120 in an embodiment of this application; Figure 3This is a flowchart of the second model construction module 130 in an embodiment of this application; Figure 4 This is a flowchart of the data prediction module 140 in an embodiment of this application; Figure 5 This is a flowchart of the crack assessment module 150 in an embodiment of this application. Detailed Implementation

[0020] The following combination Figures 1 to 5 This application will be described in further detail.

[0021] Example: This example discloses a digital twin-based prediction and evaluation system 100 for predicting the cracking evolution of cement-stabilized crushed stone base courses, such as... Figure 1 As shown, the system 100 includes: a data acquisition module 110 for acquiring road condition data of the road base layer; a first model construction module 120 for constructing a digital twin model of the base layer to obtain road condition evolution process data; a second model construction module 130 for constructing a road cracking prediction evolution model; a data prediction module 140 for inputting the current road condition data and road condition evolution process data into the road cracking prediction evolution model and outputting road cracking prediction data; and a cracking assessment module 150 for calculating the base layer cracking risk index and determining the base layer cracking risk level. This embodiment includes the following steps: The data acquisition module 110 is used to acquire road condition data of the actual road base layer, including base layer structural parameters, material parameters, construction parameters, environmental parameters, traffic load parameters, and base layer crack detection data. The base layer structural parameters include base layer type and thickness; the material parameters include the compressive resilient modulus and compaction degree of the base layer material; the environmental parameters include road surface temperature and annual precipitation; the traffic load parameters include vehicle trajectory and axle load data; and the base layer crack detection data includes road crack detection images. Afterwards, all acquired data undergoes data cleaning and outlier removal, followed by dimension unification and format conversion, and finally, spatiotemporal alignment of the processed data.

[0022] The first model construction module 120 is used to construct a digital twin model of the road base layer based on the base layer structure parameters of the actual road base layer. The digital twin model establishes a spatial mapping relationship and state association relationship between the actual road base layer and the virtual road base layer, maps the road state data to the corresponding position of the virtual road base layer, and updates the road state data of the virtual road base layer based on road state data collected at different time points, thereby obtaining road state evolution process data of the virtual road base layer. The first model construction module 120 includes a spatial partitioning unit 121, an attribute assignment unit 122, an environment mapping unit 123, a load mapping unit 124, and a crack mapping unit 125, as follows: Figure 2 As shown.

[0023] Spatial division unit 121 is used to construct a virtual three-dimensional model of the road base that is consistent with the geometry of the actual road base, based on the base type and thickness data. It establishes a mapping and transformation relationship between the spatial coordinate system of the actual road base and the coordinate system of the virtual road base model, and determines the state correlation between the actual and virtual road bases. The cement-stabilized crushed stone base is divided into multiple base state units according to preset dimensions. A mapping relationship is established between each base state unit and the corresponding virtual area of ​​the virtual road base. The road state evolution process data of the virtual road base is mapped to each base state unit, obtaining the road state evolution process data for each base state unit.

[0024] The attribute assignment unit 122 is used to map material parameters to each base layer state unit and establish the initial mechanical properties of each base layer state unit. Specifically, the compressive resilient modulus and compaction degree data of the base layer material are assigned to each base layer state unit according to the spatial correspondence, so that each base layer state unit obtains initial attribute parameters that characterize its material mechanical properties, forming the initial road state data of the virtual road base layer.

[0025] The environmental mapping unit 123 is used to map environmental parameters to the corresponding areas of the virtual road base layer. Specifically, based on road surface temperature data and annual precipitation data, combined with the spatial location of each base layer state unit and the characteristics of its respective climate zone, the environmental parameters are mapped to the corresponding base layer state units, and the environmental action state data of each base layer state unit are updated.

[0026] The load mapping unit 124 is used to discretize the vehicle's trajectory into a sequence of trajectory points at equal time intervals or equal spatial intervals, and obtain the spatial coordinates of each trajectory point and the corresponding axle load data of the vehicle at that time. The spatial coordinates of each trajectory point are matched with the spatial coordinates of the base layer state units to determine whether the trajectory point falls within the spatial boundary range of the base layer state units, thus determining the set of base layer state units traversed by the vehicle load. Based on the vehicle tire ground contact width and the lateral distribution characteristics of the wheel tracks, a preset width is extended to both sides of the trajectory point to form the vehicle load action area. The intersection operation of the load action area range and the spatial boundary of the base layer state units is performed to determine the coverage ratio of each base layer state unit by the load action area. Based on the coverage ratio, the number of load actions of each base layer state unit is weighted and statistically analyzed. Base layer state units with a coverage ratio greater than a preset ratio threshold are recorded as one effective load action. According to the vehicle axle load data, the single effective load action is converted into a single damage amount according to the fatigue damage equation of the base layer material. The fatigue damage equation uses the standard axle load as a benchmark, converting the actual axle load into a damage coefficient relative to the standard axle load according to the equivalent damage principle. The cumulative damage of each base layer state unit within the statistical period is obtained by summing all single damage amounts. The cumulative damage of each base layer state unit is then mapped to the corresponding area of ​​the virtual road base layer, updating the load application status data of that area. The cumulative damage of each base layer state unit is compared with a preset load threshold, which is determined based on the fatigue life characteristics and design service life of the base layer material. For base layer state units with cumulative damage exceeding the preset load threshold, a crack-sensitive marker is written into their attribute data, and this marker is synchronized to the corresponding area of ​​the virtual road base layer.

[0027] The crack mapping unit 125 is used to process the processed crack detection image using an image segmentation algorithm, distinguishing crack pixels from road background pixels and extracting a crack region mask image. Morphological thinning processing is performed on the crack region mask image to extract crack skeleton lines; the crack skeleton lines are the center lines of the crack region, representing the topological orientation of the crack. The crack length is calculated based on the crack skeleton lines, where the crack length is the product of the number of crack skeleton line pixels and the image spatial resolution calibration coefficient. Along the normal direction of the crack skeleton lines, the crack opening width is measured on the crack region mask image, and the average crack width and maximum crack width are calculated. The crack area is obtained by multiplying the total number of crack pixels in the crack region mask image by the actual area represented by each pixel. The spatial location of the crack in the actual road base layer is determined based on the acquisition station number or geographic coordinates corresponding to the crack detection image. The crack spatial location is matched with the spatial boundary of the base layer state unit to determine the base layer state unit where the crack is located. The crack length, average crack width, maximum crack width, and crack area are assigned to the corresponding base layer state unit. The crack distribution density is obtained by calculating the number of cracks and the total crack length within a unit of base layer state unit of unit length or area. The aforementioned crack geometric feature data are mapped to the base state unit corresponding to the virtual road base layer, and the road state data of the corresponding area of ​​the virtual road base layer is updated.

[0028] The second model construction module 130 is used to extract historical road state evolution features based on road state evolution process data from the virtual road base layer, and to construct a road cracking prediction evolution model based on these historical road state evolution features. It includes a graph construction unit 131, a feature encoding unit 132, and a model construction unit 133, such as... Figure 3 As shown.

[0029] Graph construction unit 131 is used to obtain the spatial coordinates of each base-level state unit and determine the spatial adjacency relationship between each base-level state unit. The spatial adjacency relationship includes shared edge adjacency, shared vertex adjacency, or proximity within a preset spatial distance range; the preset spatial distance is determined according to the division size of the base-level state unit and does not exceed twice the distance between the center points of adjacent units. An initial state association graph is constructed using each base-level state unit as a graph node and spatial adjacency relationships as edges. The node attributes of the initial state association graph include the base-level state unit number and spatial coordinates, while the edge attributes only identify spatial adjacency relationships, and the initial edge weights are uniformly set to a baseline value. For each pair of adjacent base-level state units with spatial adjacency relationships, the correlation degree of crack state changes in their historical time series is calculated. Specifically, crack state change sequences of two adjacent base-level state units within the same time window are extracted, the crack state change sequences are detrended, and the Pearson correlation coefficient or mutual information value between the two sequences is calculated as a quantitative indicator of the correlation degree of crack state changes. The calculated correlation degree of crack state changes is compared with a preset correlation threshold. The preset correlation threshold is determined based on the statistical distribution of the correlation degree between adjacent base-level state units along known crack propagation paths in historical data, taking the upper quartile of the statistical distribution or setting it according to expert experience. For adjacent base-level state units where the correlation degree of crack state change is greater than or equal to the preset correlation threshold, crack propagation correlation edges are established. These crack propagation correlation edges have directional attributes, pointing from base-level state units with a leading crack state change sequence to base-level state units with a lagging change sequence, or from base-level state units with lower crack severity to base-level state units with higher crack severity. The crack propagation correlation edges are superimposed on the initial state correlation graph, updating the edge set and edge attributes to obtain the base-level state graph network. In the base-level state graph network, nodes are base-level state units, and edges include spatial adjacency edges and crack propagation correlation edges. Crack propagation correlation edges use the correlation degree of crack state change as their edge weight, while spatial adjacency edges use a baseline value as their edge weight.

[0030] Feature encoding unit 132 is used to acquire road state evolution process data corresponding to each base layer state unit. The road state evolution process data includes numerical sequences of base layer structural parameters, material parameters, environmental parameters, traffic load parameters, and crack geometric feature data at different times. Temporal features are extracted from the road state evolution process data. For each base layer state unit, statistical features of its temporal data are extracted, including the mean, variance, standard deviation, maximum value, minimum value, and range of each parameter sequence; trend features of the temporal data are extracted, including the linear fitting slope, moving average sequence, and cumulative change of each parameter sequence; dynamic features of the temporal data are extracted, including the rate of change, acceleration, and difference sequence between adjacent times of each parameter sequence; abrupt change features of the temporal data are extracted, including the abrupt change time, abrupt change amplitude, and abrupt change frequency of crack width, cumulative load damage, or road surface temperature. The extracted temporal features are then feature-encoded to obtain the temporal state embedding vector corresponding to each base layer state unit. Specifically, statistical features, trend features, dynamic features, and mutation features are concatenated into vectors to form a high-dimensional original feature vector; the high-dimensional original feature vector is then subjected to feature filtering to remove redundant features whose correlation with the crack state is lower than a preset correlation threshold; the filtered feature vector is then subjected to dimensionality reduction processing through principal component analysis or feature mapping transformation to convert it into a fixed-dimensional temporal state embedding vector.

[0031] Model building unit 133 is used to acquire the basic state graph network. It constructs graph structure data using basic state units as graph nodes and spatial adjacency edges and crack propagation association edges as graph edges. The temporal state embedding vector of each graph node is used as the initial node feature and bound to the graph structure data. The graph structure data with bound node features is input into the graph neural network model. Through the message passing mechanism of the graph neural network, each graph node aggregates the temporal state embedding vector information of its first-order neighboring nodes, propagates the crack state evolution law along the crack propagation association edges, and updates the hidden state feature representation of each graph node. During the message passing process of the graph neural network, information aggregation weights are distinguished according to edge type: for crack propagation association edges, the degree of correlation of crack state change is used as the attention weight for message passing; for spatial adjacency edges, the baseline weight is used for information aggregation. Based on the updated hidden state feature representation of the graph nodes, a temporal prediction network is constructed. The temporal prediction network takes the hidden state feature sequence of each graph node at the current time and several previous historical times as input, extracts node-level temporal dependencies through the temporal modeling layer, and outputs the predicted crack state value of each basic state unit within a future preset time window. The predicted crack state values ​​include predicted crack length, predicted crack width, and crack propagation probability. The road cracking prediction evolution model is trained using historical road state evolution data. The road state evolution data from the previous few time points in the historical data are feature-encoded and used as model input. The actual crack geometric feature data from the next time point are used as supervision labels. By minimizing the mean square error or cross-entropy loss between the predicted and actual crack states, the backpropagation algorithm is used to optimize the model parameters of the graph neural network and the temporal prediction network, resulting in the trained road cracking prediction evolution model. This trained model is then used as the new road cracking prediction evolution model.

[0032] Data prediction module 140 is used to acquire current road state data from the basic digital twin model, input the current road state data and road state evolution process data into the road cracking prediction evolution model, and output road cracking prediction data. It includes a first unit 141, a second unit 142, and a third unit 143, as follows: Figure 4 As shown.

[0033] Unit 141 is used to read the road state data of each base-level state unit at the current moment from the base-level digital twin model, including base-level structural parameters, material parameters, environmental parameters, traffic load parameters, and crack geometric feature data. The read data is standardized to align the dimensions and numerical ranges of each parameter to the model input requirements, forming the node feature matrix of each base-level state unit at the current moment.

[0034] The second unit 142 is used to acquire road state evolution process data of each basic state unit within a preset historical time window, extract statistical features and trend features of each parameter sequence, including mean, rate of change and crack growth slope; and convert the extracted time series features into a fixed-dimensional historical road state evolution feature vector according to feature encoding rules.

[0035] The third unit 143 maps the current road state data obtained by the first unit 141 and the historical road state evolution feature vector extracted by the second unit 142 to the corresponding graph nodes of the base-level state graph network, respectively, to construct the graph structure input for prediction. Spatial correlation features between each base-level state unit are extracted based on the base-level state graph network. The state information of neighboring nodes is aggregated along the graph edges using a graph neural network, enabling each graph node to obtain a hidden state representation that incorporates the influence of spatial neighborhood. Evolutionary features of road state changes over time are extracted based on the current road state data and historical road state evolution features. The current and historical features of each graph node are concatenated temporally, and the temporal modeling layer captures the pattern of road state evolution over time, outputting a temporal evolution feature representation. Spatial correlation features and temporal evolution features are fused to obtain the comprehensive prediction features of each base-level state unit. Based on the comprehensive prediction features, decoding is performed to output the crack prediction state data of each base-level state unit within a future preset time window, including crack prediction length, crack prediction area, and crack prediction growth rate; simultaneously, the crack expansion trend is output, including the number of cracks and the crack expansion direction.

[0036] Cracking assessment module 150 is used to calculate the base layer cracking risk index based on road cracking prediction data, compare the base layer cracking risk index with a preset risk level threshold, and determine the base layer cracking risk level. It includes a risk index calculation unit 151 and a risk assessment unit 152, such as... Figure 5 As shown.

[0037] The risk index calculation unit 151 is used to acquire road cracking prediction data output by the data prediction module 140, including the predicted crack length, predicted crack area, predicted crack growth rate, predicted crack number, and crack propagation direction for each base layer state unit. The predicted crack length and predicted crack area for each base layer state unit are normalized by dividing the predicted crack length by the historical maximum observed crack length for that type of base layer and dividing the predicted crack area by the historical maximum observed crack area for that type of base layer, resulting in normalized length and area indices. The length and area indices are then weighted and summed to obtain the crack severity index for that base layer state unit. The weights of the weighted summation are determined based on the degree of influence of crack length and crack area on the load-bearing capacity of the base layer structure. The crack length weight reflects the degree of weakening of the overall structure by the crack, and the crack area weight reflects the degree of reduction of the effective load-bearing area of ​​the structure by the crack. In this embodiment, the weights of both the length and area indices are 0.5. The predicted crack growth rate is the increase in predicted crack length per unit time relative to the current crack length. The ratio of the predicted crack growth rate to the historical average crack propagation rate of this type of base course under the same environmental and load conditions is used as the crack activity index. When the predicted crack growth rate is greater than the historical average propagation rate, the crack activity index is positive and increases with the degree of deviation; when the predicted crack growth rate is not greater than the historical average propagation rate, the crack activity index is negative and decreases with the degree of mitigation. The number of base course state units predicted to have cracks is counted, and the number of predicted cracks is divided by the total number of base course state units in the road section to obtain the crack number density index. Crack propagation direction data for each base course state unit is obtained, where the crack propagation direction is represented by the azimuth angle relative to the longitudinal direction of the road. The standard deviation of all crack propagation directions is calculated to obtain the crack dispersion. The crack severity index, crack activity index, crack number density index, and crack dispersion are weighted and summed to obtain the base course cracking risk index. In this embodiment, the weights of the crack severity index, crack activity index, crack number density index, and crack dispersion are 0.3, 0.2, 0.4, and 0.1, respectively.

[0038] The risk assessment unit 152 is used to compare the base cracking risk index with the first preset risk level threshold. If the base cracking risk index is greater than the preset risk level threshold, the base cracking risk level of the actual road is determined to be level three. If the base cracking risk index is not greater than the preset risk level threshold, it is further determined whether the base cracking risk index is greater than the second preset risk level threshold. If it is, the base cracking risk level of the actual road is determined to be level two; otherwise, the base cracking risk level of the actual road is determined to be level one.

[0039] The model update module 160 is used to obtain the actual road cracking data corresponding to the prediction time point of the road cracking prediction data, compare the actual road cracking data with the road cracking prediction data, calculate the prediction deviation, and optimize the model parameters of the road cracking prediction evolution model based on the prediction deviation to obtain the optimized road cracking prediction evolution model.

[0040] The road maintenance module 170 is used to select corresponding strategies from a preset maintenance strategy library based on road cracking prediction data and base layer cracking risk levels, and output these as target maintenance strategies. The preset maintenance strategy library stores maintenance strategy entries corresponding to different risk levels and different combinations of crack characteristics; each maintenance strategy entry includes maintenance measure type, maintenance timing suggestions, maintenance resource requirements, and expected maintenance effects. Specifically, a first-level screening is performed based on the base layer cracking risk level. When the risk level is level one, daily inspection strategies are extracted from the preset maintenance strategy library, including increasing crack monitoring frequency, recording crack development dynamics, and maintaining existing traffic load management measures. When the risk level is level two, preventative maintenance strategies are extracted, including crack sealing, local base layer reinforcement, or adjustment of drainage facilities. When the risk level is level three, restorative maintenance strategies are extracted, including local base layer excavation and reconstruction, full-depth base layer repair, or structural layer replacement. A second-level screening is performed based on road cracking prediction data. Under the same risk level, the maintenance strategies are further refined and matched based on crack prediction status data and crack expansion trends. When the predicted crack growth rate exceeds a preset growth rate threshold, measures with rapid crack-stopping capabilities are prioritized in the maintenance strategies corresponding to the risk level. When the predicted number of cracks exceeds a preset number threshold, large-area, comprehensive treatment measures are prioritized. When cracks extend in a highly consistent direction and exhibit a through-crack trend, specific treatment measures targeting through-cracks are prioritized. The maintenance strategies determined after these two levels of screening are output as the target maintenance strategies. These target maintenance strategies include specific maintenance measure types, suggested implementation time windows, a list of required materials and equipment, and traffic management plans, for the maintenance management department to formulate and execute maintenance operation plans.

[0041] In this embodiment, a digital twin-based cement-stabilized crushed stone base course crack evolution prediction and assessment system is constructed. This system integrates road base course condition perception, digital model construction, crack evolution prediction, risk level assessment, and maintenance decision-making, achieving intelligent management of the entire road base course process from condition monitoring to crack prediction and risk control. Compared to existing methods that rely on single-test data or empirical models for crack evaluation, this application can dynamically acquire information on the condition changes of the base course during its service life. By comprehensively considering the influence of multiple factors such as material properties, environmental factors, and traffic loads, it improves the accuracy of base course crack evolution prediction and risk assessment, providing a reliable basis for preventive road maintenance.

[0042] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A digital twin-based system for predicting and evaluating the cracking evolution of cement-stabilized crushed stone base courses, characterized in that, The system includes: The data acquisition module is used to acquire road condition data of the actual road base layer, including base layer structural parameters, material parameters, construction parameters, environmental parameters, traffic load parameters, and base layer crack detection data; The first model construction module is used to construct a digital twin model of the road base layer based on the base layer structure parameters of the actual road base layer. The digital twin model of the base layer is used to establish the spatial mapping relationship and state association relationship between the actual road base layer and the virtual road base layer, and to map the road state data to the corresponding position of the virtual road base layer. The road state data of the virtual road base layer is updated according to the road state data collected at different time nodes to obtain the road state evolution process data of the virtual road base layer. The second model construction module is used to extract historical road state evolution characteristics based on the road state evolution process data of the virtual road base layer, and to construct a road cracking prediction evolution model based on the historical road state evolution characteristics. The data prediction module is used to acquire the current road status data in the grassroots digital twin model, input the current road status data and road status evolution process data into the road cracking prediction evolution model, and output road cracking prediction data. The cracking assessment module is used to calculate the base layer cracking risk index based on road cracking prediction data, compare the base layer cracking risk index with a preset risk level threshold, and determine the base layer cracking risk level.

2. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base courses according to claim 1, characterized in that, The first model building module further includes: The spatial division unit is used to establish a spatial coordinate model of the base layer based on the base layer structure parameters. The cement-stabilized crushed stone base layer is divided into multiple base layer state units according to the preset size. The mapping relationship between each base layer state unit and the corresponding virtual area of ​​the virtual road base layer is established. The road state evolution process data of the virtual road base layer is mapped to each base layer state unit to obtain the road state evolution process data of each base layer state unit.

3. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base courses according to claim 2, characterized in that, The traffic load parameters include vehicle trajectory and vehicle axle load data; The first model building module further includes: The load mapping unit is used to determine the range of vehicle load action area based on the overlap relationship between the vehicle's running trajectory and the spatial coordinates of the base layer state unit. Based on the range of vehicle load action area, it counts the number of load actions of each base layer state unit. Based on the vehicle axle load data and the number of load actions of each base layer state unit, it calculates the cumulative load damage of each base layer state unit. Based on the cumulative load damage of each base layer state unit, it updates the load action state of the corresponding area of ​​the virtual road base layer. It compares the cumulative load damage of each base layer state unit with a preset load threshold and marks the base layer state units with a crack sensitivity value that is greater than the preset load threshold.

4. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base course according to claim 2, characterized in that, The base layer crack detection data includes road crack detection images; The first model building module further includes: The crack mapping unit is used to perform crack recognition processing on the road crack detection image, extract crack geometric feature data, including crack length, crack width, crack area and crack distribution density; map the crack geometric feature data to the base layer state unit corresponding to the virtual road base layer, and update the road state data of the area corresponding to the virtual road base layer.

5. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base courses according to claim 2, characterized in that, The second model building module includes: The graph construction unit is used to obtain the positional relationship between each base-level state unit, and to generate an initial state association graph based on the spatial adjacency relationship of each base-level state unit; to obtain historical road state evolution process data, to calculate the degree of correlation between crack state changes between adjacent base-level state units, to establish crack expansion association edges between adjacent base-level state units whose crack state association degree meets a preset association degree threshold, and to update the initial state association graph based on the crack expansion association relationship to obtain the base-level state graph network.

6. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base courses according to claim 5, characterized in that, The second model building module also includes: The feature encoding unit is used to acquire road state evolution process data corresponding to each basic state unit, extract time-series features from the road state evolution process data, and encode the time-series features to obtain the time-series state embedding vector corresponding to each basic state unit. The model building unit is used to construct a road cracking prediction evolution model based on the underlying state graph network and the temporal state embedding vectors of each graph node.

7. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base courses according to claim 6, characterized in that, The data prediction module specifically includes: The first unit is used to obtain the current road status data corresponding to each grassroots state unit in the grassroots digital twin model; The second unit is used to extract the historical road state evolution features corresponding to each basic state unit from the road state evolution process data; The third unit is used to input current road condition data and road condition evolution process data into the road cracking prediction evolution model. Based on the spatial correlation between each basic state unit, a basic state map network is constructed to extract the spatial correlation features between the basic state units. Based on the current road condition data and historical road condition evolution features, the spatial correlation features between the basic state units and the evolution features of road condition over time are extracted. Based on the spatial correlation features and time evolution features, predictions are made and road cracking prediction data is output. The road cracking prediction data includes crack prediction status data and crack propagation trend. Crack prediction status data includes crack prediction length, crack prediction area, and crack prediction growth rate. Crack propagation trend includes crack prediction number and crack propagation direction.

8. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base course according to claim 7, characterized in that, The crack assessment module includes: The risk index calculation unit is used to calculate the crack severity index based on the crack prediction length and crack prediction area, the crack activity index based on the crack prediction growth rate, the crack number density index based on the crack prediction number, and the crack dispersion based on the crack extension direction. The crack severity index, crack activity index, crack number density index and crack dispersion are weighted and summed to obtain the base layer cracking risk index. The risk assessment unit compares the base layer cracking risk index with a first preset risk level threshold. If the base layer cracking risk index is greater than the preset risk level threshold, the actual road base layer cracking risk level is determined to be Level 3. If the base layer cracking risk index is not greater than the preset risk level threshold, it continues to determine whether the base layer cracking risk index is greater than a second preset risk level threshold. If it is, the actual road base layer cracking risk level is determined to be Level 2; otherwise, the actual road base layer cracking risk level is determined to be Level 1.

9. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base courses according to claim 1, characterized in that, The system also includes: The model update module is used to obtain the actual road cracking data corresponding to the prediction time point of the road cracking prediction data, compare the actual road cracking data with the road cracking prediction data, calculate the prediction deviation, and optimize the model parameters of the road cracking prediction evolution model based on the prediction deviation to obtain the optimized road cracking prediction evolution model.

10. The digital twin-based prediction and evaluation system for crack evolution of cement-stabilized crushed stone base courses according to claim 1, characterized in that, The system also includes: The road maintenance module is used to select the corresponding strategy from the preset maintenance strategy library based on the road cracking prediction data and the base cracking risk level, and output it as the target maintenance strategy.

Images