Highway engineering quality real-time monitoring method and system based on digital twinning

Abstract

The invention provides a highway engineering quality real-time monitoring method and system based on digital twinning, and the method comprises the steps: collecting construction data, and carrying out the preprocessing of the construction data, and obtaining a behavior time sequence and a coverage rate index sequence; constructing a behavior graph, performing quality prediction on each behavior segment node through a graph neural network to obtain a predicted quality score, and performing coverage rate index sequence correction to obtain an actual quality score; mapping the actual quality score to a spatial position to obtain a spatial quality tensor, wherein the spatial position is a three-dimensional position obtained by combining equipment position data and a compaction horizon number; performing structural consistency analysis on the spatial quality tensor, identifying and marking a topology abnormal region, and outputting a structural consistency scoring tensor; and generating a visual twin layer based on the spatial quality tensor and the structural consistency scoring tensor, and performing graph level labeling and section mapping on the abnormal region.

Description

Technical Field

[0001] This invention belongs to the field of engineering construction, and in particular relates to a method and system for real-time monitoring of highway engineering quality based on digital twins. Background Technology

[0002] In modern highway construction, construction quality directly affects the service life, safety performance, and subsequent maintenance costs of road structures. This is especially true in critical stages such as subgrade compaction and multi-layer paving, where the intensity distribution of compaction, inter-layer continuity, and the rationality of equipment construction paths collectively determine the final project quality. Currently, mainstream quality monitoring methods still rely on periodic inspections and on-site sampling, failing to achieve real-time perception and full-site coverage of the continuous compaction status during construction. This makes it difficult to detect hidden structural defects such as localized uncompacted areas, excessive compaction, and faults. Furthermore, existing digital twin systems largely focus on construction progress synchronization and scheduling visualization, lacking dynamic modeling and evolution mechanisms for construction quality itself, and thus failing to accurately represent the continuous impact of construction behavior on the physical structure. While some systems connect to IoT terminals to collect equipment operation data, they do not form a closed loop from behavioral sequences to quality inference, and the data structure is disconnected from the spatial model. This prevents prediction results from being used to dynamically drive the physical field updates of the twin system, and the system cannot promptly express quality evolution trends or structural deviation risks. In addition, existing quality inspection results are mostly discrete point data, which lack the ability to construct spatial continuous representations or structural consistency identification, and lack the topological modeling ability to move from single-point judgment to regional anomaly judgment. Summary of the Invention

[0003] The purpose of this invention is to design a method and system for real-time monitoring of highway engineering quality based on digital twins. It has high spatial representation capability, regional anomaly identification capability and platform integration capability, and is applicable to scenarios of real-time monitoring of highway quality and systematic risk identification in multi-layer filling and irregular construction scenarios.

[0004] To achieve the above objectives, a method for real-time monitoring of highway engineering quality based on digital twins is provided in a first aspect of the present invention, the method comprising:

[0005] Collect construction data, including the operating status of construction equipment and construction parameters; preprocess the construction data to obtain behavioral time series and coverage index series.

[0006] A behavior graph is constructed based on the behavior time series and coverage index sequence. A graph neural network is used to predict the quality of each behavior segment node to obtain the predicted quality score, and the actual quality score is obtained by correcting it with the coverage index sequence.

[0007] The actual quality score is mapped to a spatial location to obtain a spatial quality tensor, where the spatial location is a three-dimensional location obtained by combining the equipment location data and the compaction layer number.

[0008] Structural consistency analysis is performed on the spatial mass tensor to identify and mark topologically anomalous regions, and a structural consistency score tensor is output.

[0009] A visual twin layer is generated based on the spatial quality tensor and the structural consistency score tensor, and layer-level annotation and section mapping are performed on abnormal areas.

[0010] Furthermore, the operating status of the construction equipment includes location data, real-time speed, and rolling pressure. The construction parameters include the compaction layer number to which the construction equipment belongs, which is a process parameter planned in advance in the construction management platform.

[0011] Furthermore, the coverage index sequence is constructed by using two-dimensional linear interpolation to create a continuous pressure field and then dividing it into grids for integral calculation. This is used to measure whether the current rolling pressure at that location exceeds the set effective compaction threshold.

[0012] Furthermore, the nodes of the behavior graph are composed of sequence points in the behavior time series, and the edge set of the behavior graph is constructed based on the temporal sequence and spatial proximity of the construction behaviors.

[0013] Furthermore, the main structure of the graph neural network is a three-layer graph attention network; a physical consistency constraint term is introduced into the graph neural network, which suppresses excessive propagation across layers or regions by dynamically adjusting the weights between nodes.

[0014] Furthermore, the spatial mass tensor structure is a three-dimensional tensor, with dimensions corresponding to the construction area design drawing, and the unit is a normalized mass level value.

[0015] Furthermore, the spatial quality tensor is obtained by using a spatial interpolation method that weights the actual quality score with a Gaussian kernel and adaptively adjusts the compacted trajectory density. A trajectory sparsity adjustment factor is introduced, which represents the number of trajectory points within a radius R centered on the spatial location coordinates, and is used to measure whether the segment is in a sparse data region.

[0016] Furthermore, a structure regularization term is introduced into the spatial mass tensor to impose gradient constraints on the entire spatial mass tensor, enabling it to possess smoothness and reasonable variation consistent with the actual physical compaction state.

[0017] Furthermore, the structural consistency analysis includes: scoring the structural anomaly tendency of each spatial unit, identifying spatial anomaly structures through Laplace transform, suppressing false anomalies caused by missing data through coverage weights to obtain a sensitivity score, generating a structural anomaly mask by binarizing the sensitivity score, and filtering out continuous anomaly regions by combining a spatial connectivity judgment mechanism.

[0018] A second aspect of the invention provides a real-time monitoring system for highway engineering quality based on digital twins, the system comprising:

[0019] A data acquisition unit is used to collect construction data, including the operating status of construction equipment and construction parameters; the construction data is preprocessed to obtain a behavioral time series and a coverage index series.

[0020] The quality prediction unit is used to construct a behavior graph based on the behavior time series and coverage index sequence, perform quality prediction on each behavior segment node through a graph neural network to obtain a predicted quality score, and obtain the actual quality score through correction by the coverage index sequence.

[0021] A structural mapping unit is used to map the actual quality score to a spatial location to obtain a spatial quality tensor, wherein the spatial location is a three-dimensional location obtained by combining equipment location data and compaction layer number.

[0022] An anomaly identification unit is used to perform structural consistency analysis on the spatial mass tensor, identify and mark topologically anomalous regions, and output a structural consistency score tensor.

[0023] The result output unit is used to generate a visual twin layer based on the spatial quality tensor and the structural consistency score tensor, and to perform layer-level annotation and section mapping for abnormal areas.

[0024] The beneficial technical effects of the present invention are at least as follows:

[0025] To address the aforementioned issues, this invention provides a method and system for real-time monitoring of highway engineering quality based on digital twins. The system collects time-series inputs, including the trajectory, pressure, speed, and layer information of construction equipment during construction. A graph structure representing construction behavior units, incorporating spatial, sequential, and inter-layer logic, is constructed, and a graph neural network is used for node-level inference of compaction quality. The prediction results are mapped to a multi-level spatial quality tensor using Gaussian kernel interpolation, and boundary area reliability is corrected by combining trajectory density and coverage information. Furthermore, a topological sensitivity score is constructed through tensor-level structural gradient analysis. Based on Laplace variation and construction coverage weighting, continuous structural anomaly zones are identified, and a structural consistency layer is generated using connected area rules. Finally, the system jointly projects the quality tensor and the anomaly structure layer into the twin model, generating a multi-dimensional quality expression layer that supports platform heatmap visualization, structural anomaly marking, and section quality summary forms.

[0026] This solution establishes a complete technology chain from construction behavior to spatial quality and structural consistency. It has high spatial expression capabilities, regional anomaly identification capabilities, and platform integration capabilities, and is suitable for real-time monitoring and systematic risk identification of highway quality in multi-layer filling and irregular construction scenarios. Attached Figure Description

[0027] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0028] Figure 1 This is a flowchart of the real-time monitoring method for highway engineering quality based on digital twins, as described in this invention.

[0029] Figure 2 This is a framework diagram of the real-time monitoring system for highway engineering quality based on digital twins, as described in this invention. Detailed Implementation

[0030] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0031] In one or more embodiments, such as Figure 1 As shown, a method for real-time monitoring of highway engineering quality based on digital twins is disclosed, the method comprising the following:

[0032] S1: Collect construction data, including the operating status of construction equipment and construction parameters; preprocess the construction data to obtain behavioral time series and coverage index series;

[0033] Specifically, in this embodiment, the compaction process at the construction site is dynamic and continuous, and the operating status and construction parameters of the road roller in different areas and layers are the core factors determining the quality of the roadbed. Therefore, the first task is to collect these behavioral characteristics in a realistically obtainable manner in engineering, forming a structured compaction behavior time series. This sequence will provide a solid data foundation for subsequent quality state prediction and structure mapping. We define each s in the behavior sequence. t It is a quintuple containing the current spatial location (x) t ,y t Roller speed v t Rolling pressure p t and the compaction layer number l t This data structure covers the kinematic information of construction activities, reflects the quality factors, and can correspond one-to-one with the sections and layers in the construction design drawings.

[0034] The data acquisition process is based on standard engineering equipment and is highly feasible to implement. Specifically, location data (x t ,y t The data is obtained through a GNSS receiver installed on top of the road roller. The system configuration typically uses a dual-frequency GNSS module with RTK differential capability (such as the Trimble R12i or Huace T300 series), and outputs spatiotemporal data via the NMEA protocol. Real-time speed v t The rolling pressure (p) is measured by the speed sensor in the roller's travel drive system, typically integrated into the hydraulic pump / motor unit and connected to the main controller (such as the BOMAG or SANY built-in controller), and can be retrieved via the CAN bus interface. t The acquisition of pressure primarily relies on pressure sensors installed in the hydraulic circuit of the roller. These are typically strain gauge sensors with a pressure range of 0–25 MPa, which output analog voltage or digital signals in real time. After A / D conversion, the unit pressure currently applied to the ground is recorded. Layer number l t Instead of being acquired through sensors, the process parameters are pre-planned in the construction management platform. For example, if one section is the first layer of fill and another section is the second layer of base course, the system automatically associates this number with the task plan during equipment operation and distributes it synchronously on the roller terminal equipment.

[0035] The raw data collected often suffers from inconsistent time intervals, instantaneous outliers, and broken trajectories, thus requiring normalization to form a stable time series of compressed behavior. The normalization method employs a fixed-time-window resampling strategy, setting the sampling period Δt = 1 second. For each window [t-Δt / 2, t+Δt / 2], median processing is used to generate representative points s. t That is, the median value of the compaction speed and pressure within this time window is taken as the representative value of the current behavior. The location and floor number are recorded at the midpoint of the time period. This approach effectively eliminates extreme value shifts in the field data caused by equipment vibration and communication interruptions, while preserving the true trend of the trajectory. For example, if the pressure data within a certain window is [0.35, 0.38, 0.29, 0.40] MPa, the median value is taken. This indicates the representative pressure of the segment.

[0036] Furthermore, to quantify the effectiveness of construction coverage in physical space for each action, a regional coverage index κ is introduced here. t This indicator is used to measure the current rolling pressure p. t Is it within the spatial range Ω? t It is fully distributed within the interior. The definition is as follows:

[0037]

[0038] Among them, Ω t The area A represents the construction zone of the compaction section. δ(·) is a Boolean indicator function indicating whether the rolling pressure at this location exceeds the set effective compaction threshold τ, typically set to 0.25–0.30 MPa according to construction specifications. The physical meaning of this indicator is whether a certain section has been "compacted sufficiently," rather than simply "passed through." In actual calculations, a continuous pressure field is constructed using two-dimensional linear interpolation, followed by gridded integration calculations, achieving good accuracy within a 10m × 10m area.

[0039] S2: Construct a behavior graph based on the behavior time series and coverage index sequence, perform quality prediction on each behavior segment node through a graph neural network to obtain the predicted quality score, and obtain the actual quality score through correction by the coverage index sequence.

[0040] Specifically, in this embodiment, the objective of this step is to analyze the acquired stress test behavior time series. With coverage index series Based on this, a graph structure with engineering semantics and structural dependencies is constructed. And through graph neural networks Quality prediction is performed for each behavior segment node. This step is the key hub of the entire patent solution - it not only transforms the actual construction behavior into a computable graph model structure, but also enables the digital twin system to subsequently dynamically obtain the quality status estimation of each construction unit segment by embedding construction physical information, thereby driving the real-time evolution of the twin. Compared with traditional methods based on grid, thermal interpolation, and single-point prediction, the innovation of this step lies in constructing a graph model input structure that includes behavioral causality, spatial continuity, physical coverage constraints, and engineering hierarchical numbering, and coupling it with the deep graph network mechanism, which is significantly superior to existing methods in terms of expression ability and prediction accuracy.

[0041] Graph structure The node set of is composed of each s in the behavior sequence t and the feature vector of each node v t is defined as where: is the median rolling speed in the t-th time window, sourced from Step 1; is the median pressure in the same window, sourced from Step 1; l t is the layer number, from the on-site task system; κ t is the coverage rate index, sourced from the spatial integration result of Step 1; d t is the Euclidean distance between the spatial positions of this behavior segment and its previous segment, for the physical verification of the edge weights in the subsequent graph structure.

[0042] The construction method of the edge set E is based on the temporal sequence and spatial proximity of construction behaviors. If two nodes v i and v j satisfy i < j and d ij < ∈, then a directed edge e ij is added, indicating that in the construction process, segment i may have an impact on the quality inheritance of segment j. The selection of ∈ is based on the single-trip spacing of construction equipment (such as 3 - 6 meters) to limit the boundary diffusion range and avoid invalid connections.

[0043] Quality prediction model The main structure of is a three-layer graph attention network (GAT), and each layer structure is: "attention calculation → adjacency aggregation → non-linear activation". The model performs state transfer and update at the node level, and its most core creativity is reflected in introducing a physical consistency constraint term in the node quality prediction process, making the model not only pursue the minimum error between the predicted value and the observed value, but also consider construction continuity and coverage integrity. This physical regularization term is defined as follows:

[0044]

[0045] in: q is the quality grade score predicted by the model for the t-th behavior segment, with an output value ranging from [0,1]; t λ is the quality level label of the actual detection in the training data (e.g., from the discretized results of the drop weight deflection); λ is the regularization coefficient, which controls the weight of the physical consistency term; γ ij =α|l i -l j |+β|d i -d j | is the construction physical discontinuity function, where l i l i It is the floor number, d i d j It is the spatial coordinate distance, used to penalize overfitting in discontinuous regions.

[0046] The regularization term here suppresses excessive propagation across layers or regions by dynamically adjusting the weights between nodes, thus solving the problem of "artificial height" prediction in overlapping areas of multi-layer construction. In actual construction, even adjacent layers may have different materials, compaction times, and equipment paths. Traditional models are prone to mismapping continuous trajectories as continuous quality states, while this regularization term explicitly suppresses such structural errors.

[0047] Furthermore, after the model prediction is completed, to enhance data stability and engineering interpretability, the coverage κ from step one is introduced. t As a correction coefficient for the prediction result, the final fragment quality expression is output.

[0048]

[0049] This correction ensures that even if the model's prediction score is high, if a section has insufficient compaction coverage, it cannot be considered a reliable quality area in the twin system. This establishes a two-factor fusion standard of "prediction rationality + construction coverage authenticity" from a mechanistic perspective. This strategy is particularly crucial in highway engineering scenarios because many "edge areas" have construction trajectories, but the actual rolling pressure is far below the standard. Therefore, it is necessary to remove such "virtual compaction" segments from the system.

[0050] S3: Map the actual quality score to the spatial location to obtain the spatial quality tensor, wherein the spatial location is a three-dimensional location obtained by combining the equipment location data and the compaction layer number;

[0051] Specifically, in this embodiment, the core task of this step is to express the quality of each compaction behavior obtained in the previous step. Spatial location (x) t ,y t ) and compaction layer number lt Transformed into a three-dimensional tensor with well-defined structures in both the construction space and the floor level. The construction of this tensor is not only a mapping of the quality prediction results at a single point, but also a high-fidelity structured expression of the quality distribution of the entire construction area. It is a key technical support for promoting the digital twin system from abstract prediction to spatial visual modeling.

[0052] To transform the discrete compression behavior into a structured spatial tensor, this step constructs a spatial grid Ω covering the entire construction area, dividing each layer l into X×Y spatial elements. Each element (x, y, l) represents the physical location of a point in a specific embankment layer in the actual road design. In order to... Mapped to This step designs a spatial interpolation mechanism that combines Gaussian kernel weighting and adaptive adjustment of compaction trajectory density.

[0053] First, define the basic interpolation expression:

[0054]

[0055] Where: δ(l=l) t ) is the Kronecker indicator function, with a value of 1 if fragment t belongs to the current layer l, and 0 otherwise, ensuring that prediction results between different layers do not interfere with each other; ω t (x,y) is the spatial influence weight of the compaction section t on the position (x,y).

[0056] Furthermore, to enhance the construction adaptability of spatial interpolation, this step introduces a penalty term reflecting trajectory density into the traditional Gaussian kernel weight:

[0057]

[0058] The first term is a standard two-dimensional Gaussian kernel function, which controls the spatial diffusion range of the compaction section's influence; σ can be set to 1.0–1.5 meters based on the compaction width. The second term is an innovative trajectory sparsity adjustment factor, N. t (x,y) represents the expression with (x,y) t ,y t The number of trajectory points within the center radius r is used to measure whether the segment is in a sparse data region; the coefficient μ is the adjustment weight, usually between 0.2 and 0.5. The larger the value, the more emphasis is placed on the weighted amplification of interpolation in sparse regions; this combined weight mechanism maintains smooth interpolation in dense data regions, while enhancing the predictive presence in sparse data regions, solving the problem of insufficient compaction of conventional interpolation in "corner blocks".

[0059] Generated tensors While the spatial tensor forms a continuous field, abnormal conditions such as under-compaction and repeated compaction along the construction equipment path can cause abrupt changes in quality values ​​or overfitting in certain regions. Therefore, we further designed a tensor-level structural regularization mechanism to apply gradient constraints to the entire spatial tensor, ensuring it possesses smoothness and reasonable variation consistent with the actual physical compaction state.

[0060]

[0061] Where: gradient term and κ'(x,y,l) represents the local rate of change of the tensor in the spatial direction; κ'(x,y,l) is the mapped spatial coverage tensor (derived from κ in step one). t (Mapped from elsewhere), reflecting the reliability of each point being compacted and covered; the weight 1-κ' serves as a sparse region penalty factor, preventing drastic jumps in the tensor within the "low coverage area," thereby improving the interpretability and stability of the heatmap in the twin system. This regularization term does not participate in interpolation calculations but is performed as a one-time filtering operation in the tensor post-processing stage, similar to edge-preserving filtering in image denoising. Its core design is bound to the physical coverage area of ​​the construction trajectory, forming a structural constraint logic that "only allows free variation within reliable areas."

[0062] S4: Perform structural consistency analysis on the spatial mass tensor, identify and mark topologically anomalous regions, and output a structural consistency score tensor;

[0063] Specifically, in this embodiment, this step aims to modify the spatial mass tensor constructed in step three. Structural consistency analysis is performed to identify topological anomalies, such as uncompacted strips, local faults, and discontinuous compaction surfaces—typical construction defects. Within the entire patent system, it serves as the "closed-loop verification of quality structure," representing the core transformation point from "model prediction reliability" to "physical spatial structural rationality." The tensor output from the previous step... The quality level of each spatial unit at a specific compaction layer has been clearly expressed, but it has not been explicitly determined whether it forms a structurally continuous and engineering-acceptable compacted entity. Therefore, this step requires structural analysis to ensure that the twin system not only "looks complete" but is also "logically sound."

[0064] Input variables are It is constructed from step three based on trajectory point interpolation and weighted fusion of compacted sections; and κ(x,y,l), which is the compaction coverage index κ from step one. t The compaction reliability tensor is constructed through spatial mapping. To ensure the reliability and engineering relevance of the analysis results, this step uses κ as the source of structural reliability weights for topological analysis, avoiding the generation of pseudo-structural features in sparse data regions.

[0065] In actual construction, especially during the roadbed compaction stage, structural anomalies such as "compaction interruption zones" or "reverse overlapping compaction zones" often occur. These areas are... The quality of a spatial unit may suddenly drop or rise, forming a clear boundary with the surrounding area. If these boundaries form a closed loop or a large-area gradient abrupt change, it may constitute a structural risk. Therefore, we designed a topological sensitivity scoring mechanism to score the "structural anomaly tendency" of each spatial unit. The calculation formula is as follows:

[0066]

[0067] Where: S(x,y,l) represents the structural sensitivity score at that point; κ(x,y,l) is the spatial two-dimensional Laplacian operator of the mass tensor, which measures the average gradient difference between the point and its surroundings, representing the degree of "short structural change"; κ(x,y,l) is the coverage confidence of the point, the higher the value, the more complete the compacted data, and the larger 1-κ is, the more likely the data is incomplete; α is the adjustment exponent, used to amplify the gradient anomaly response in low-coverage areas (e.g., α = 1.5).

[0068] This scoring formula considers both the "intensity of structural change" of the mass field and the "information reliability" of compaction behavior. It identifies spatially anomalous structures through Laplace transform and suppresses spurious anomalies caused by missing data through covering weights, making the scoring results more engineering interpretable.

[0069] Furthermore, structural consistency analysis is performed based on S(x,y,l). The scoring results are binarized to generate a structural anomaly mask R(x,y,l). Then, combined with a spatial connectivity determination mechanism, physically meaningful continuous anomaly regions are selected. The anomaly region identification is expressed as:

[0070]

[0071] in: The tensor is used to label structural anomalies; a value of 1 indicates an anomaly, and 0 indicates normality. s The sensitivity scoring threshold (e.g., selecting the 95th percentile); A(x,y,l) represents the area of ​​the connected anomaly region to which this point belongs; A min It is the minimum effective structural anomaly area threshold (e.g., 4 square meters); two The function is used for joint logical judgment: that is, point anomaly + region significance.

[0072] The identification logic here not only determines whether a single point is abnormal, but also introduces connected area filtering to achieve the evolution from "numerical abnormality" to "structural abnormality", which is in line with the judgment criteria of "segmented compaction logic" in road engineering.

[0073] S5: Generate a visual twin layer based on the spatial quality tensor and structural consistency score tensor, and perform layer-level annotation and section mapping on abnormal areas.

[0074] Specifically, in this embodiment, this step is based on the spatial mass tensor generated in step four. With structural consistency tensor A visual twin layer that can be directly accessed by the construction monitoring platform is generated, and layer-level annotations and work section mappings are performed on abnormal areas. This step is not simply image rendering, but rather the effective organization of "model prediction results" and "structural physical anomalies" into a layered data structure, which is then mapped back to the construction scheduling platform through a spatial work section coding system, achieving a "task closed loop" from AI inference to the engineering management system. This is the endpoint of the entire system's "prediction-identification-reflection" technology chain, and also the fundamental support for achieving implementable and controllable results in the digital twin platform.

[0075] The primary task of this step is to... Mapped to a heatmap layer T recognizable by the twin platform Q The (x, y, l) layer reflects quality levels using color gradations. To improve the clarity of information hierarchy, a color grading mapping function is designed here, incorporating structural anomaly factors. Differentiate different structural states in visual representation:

[0076]

[0077] Where: φ(·) is a linear piecewise mapping function that maps the quality level range [0,1] to color coding (such as blue-green-yellow-red grading); the second term This is used to directly overlay and highlight areas of structural anomalies in the layer, where β is the color enhancement coefficient for the anomaly area, controlling the display intensity (usually set to the maximum color level value). This expression ensures that on the twin platform, both the overall compaction quality and local abnormal structures can be presented, allowing monitoring personnel to easily identify potential problem areas.

[0078] Furthermore, to achieve the schedulability and archiving capability of the monitoring platform, we need to map layer content to structured data forms at the work section level. To this end, we designed a work section aggregation function Γ(g,l), which represents the comprehensive quality and anomaly statistics at a given logical work section g and layer l.

[0079]

[0080] Where: P gThis represents all spatial points (x, y) in section g; the first output is the average compaction quality grade of this layer in this section; the second output is the total number of structural anomalies, which is the direct basis for determining whether rework or enhanced compaction is required; this structured index form will be used in the monitoring platform for various purposes such as section quality audit, process acceptance, and construction report generation.

[0081] The final output consists of two objects: (1) Digital twin quality layer T Q (x,y,l), supports displaying the compaction quality and anomalies of each layer in the form of three-dimensional color block heat in the twin platform; (2) Section-level quality summary table Γ(g,l), supports automated scheduling, historical record archiving and quality control auditing of the scheduling system. The platform can export layers as GeoTIFF, Shapefile or platform-customized formats through configuration interfaces, and export form data as CSV, JSON and other structured documents to achieve interoperability with multi-source systems.

[0082] In one or more embodiments, such as Figure 2 As shown, a real-time monitoring system for highway engineering quality based on digital twins is disclosed. The system includes:

[0083] A data acquisition unit is used to collect construction data, including the operating status of construction equipment and construction parameters; the construction data is preprocessed to obtain a behavioral time series and a coverage index series.

[0084] The quality prediction unit is used to construct a behavior graph based on the behavior time series and coverage index sequence, perform quality prediction on each behavior segment node through a graph neural network to obtain a predicted quality score, and obtain the actual quality score through correction by the coverage index sequence.

[0085] A structural mapping unit is used to map the actual quality score to a spatial location to obtain a spatial quality tensor, wherein the spatial location is a three-dimensional location obtained by combining equipment location data and compaction layer number.

[0086] An anomaly identification unit is used to perform structural consistency analysis on the spatial mass tensor, identify and mark topologically anomalous regions, and output a structural consistency score tensor.

[0087] The result output unit is used to generate a visual twin layer based on the spatial quality tensor and the structural consistency score tensor, and to perform layer-level annotation and section mapping for abnormal areas.

[0088] It is worth noting that the specific workflow of the real-time monitoring system for highway engineering quality based on digital twins provided in this embodiment of the invention is the same as that of the real-time monitoring method for highway engineering quality based on digital twins described in the above embodiments, and will not be repeated here.

[0089] This invention also provides a real-time monitoring device for highway engineering quality based on digital twins, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the above embodiments of the real-time monitoring method for highway engineering quality based on digital twins, for example... Figure 1 The steps S1 to S5 described above; or, when the processor executes the computer program, it implements the functions of each module in the above system embodiments.

[0090] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the digital twin-based real-time monitoring device for highway engineering quality.

[0091] The real-time monitoring device for highway engineering quality based on digital twins can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. This device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the device may also include input / output devices, network access devices, buses, etc.

[0092] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the digital twin-based real-time monitoring equipment for highway engineering quality, connecting all parts of the equipment via various interfaces and lines.

[0093] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the digital twin-based real-time monitoring equipment for highway engineering quality by running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the operation of the air conditioning controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, SmartMemory Card (SMC), Secure Digital Card (SD), Flash Card, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0094] The modules integrated into the digital twin-based real-time monitoring equipment for highway engineering quality, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0095] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0096] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring of highway engineering quality based on digital twins, characterized in that, The method includes: Collect construction data, including the operating status of construction equipment and construction parameters; preprocess the construction data to obtain behavioral time series and coverage index series. A behavior graph is constructed based on the behavior time series and coverage index sequence. A graph neural network is used to predict the quality of each behavior segment node to obtain the predicted quality score, and the actual quality score is obtained by correcting it with the coverage index sequence. The actual quality score is mapped to a spatial location to obtain a spatial quality tensor, where the spatial location is a three-dimensional location obtained by combining the equipment location data and the compaction layer number. Structural consistency analysis is performed on the spatial mass tensor to identify and mark topologically anomalous regions, and a structural consistency score tensor is output. A visual twin layer is generated based on the spatial quality tensor and the structural consistency score tensor, and layer-level annotation and section mapping are performed on abnormal areas.

2. The method for real-time monitoring of highway engineering quality based on digital twins according to claim 1, characterized in that, The operating status of the construction equipment includes location data, real-time speed, and rolling pressure. The construction parameters include the compaction layer number to which the construction equipment belongs, which is a process parameter planned in advance in the construction management platform.

3. The method for real-time monitoring of highway engineering quality based on digital twins according to claim 1, characterized in that, The coverage index sequence is constructed by using two-dimensional linear interpolation to create a continuous pressure field and then dividing it into grids for integral calculation. It is used to measure whether the current rolling pressure at this location exceeds the set effective compaction threshold.

4. The method for real-time monitoring of highway engineering quality based on digital twins according to claim 1, characterized in that, The nodes of the behavior graph are composed of sequence points in the behavior time series, and the edge set of the behavior graph is constructed based on the temporal sequence and spatial proximity of the construction behaviors.

5. The method for real-time monitoring of highway engineering quality based on digital twins according to claim 1, characterized in that, The main structure of the graph neural network is a three-layer graph attention network. The graph neural network introduces a physical consistency constraint term, which suppresses excessive propagation across layers or regions by dynamically adjusting the weights between nodes.

6. The method for real-time monitoring of highway engineering quality based on digital twins according to claim 1, characterized in that, The spatial mass tensor structure is a three-dimensional tensor, with dimensions corresponding to the construction area design drawing, and the unit is a normalized quality level value.

7. The method for real-time monitoring of highway engineering quality based on digital twins according to claim 1, characterized in that, The spatial quality tensor is obtained by using a spatial interpolation method that weights the actual quality score with a Gaussian kernel and adaptively adjusts the compacted trajectory density. A trajectory sparsity adjustment factor is introduced, which represents the number of trajectory points within a radius R centered on the spatial location coordinates. This factor is used to measure whether the segment is in a sparse data region.

8. The method for real-time monitoring of highway engineering quality based on digital twins according to claim 1, characterized in that, The spatial mass tensor also incorporates a structure regularization term to impose gradient constraints on the entire spatial mass tensor, ensuring that it possesses smoothness and reasonable variation consistent with the actual physical compaction state.

9. The method for real-time monitoring of highway engineering quality based on digital twins according to claim 1, characterized in that, The structural consistency analysis includes: scoring the structural anomaly tendency of each spatial unit, identifying spatial anomaly structures through Laplace transform, suppressing false anomalies caused by missing data through coverage weights to obtain a sensitivity score, generating a structural anomaly mask by binarizing the sensitivity score, and filtering out continuous anomaly regions by combining a spatial connectivity judgment mechanism.

10. A real-time monitoring system for highway engineering quality based on digital twins, characterized in that, The system includes: A data acquisition unit is used to collect construction data, including the operating status of construction equipment and construction parameters; the construction data is preprocessed to obtain a behavioral time series and a coverage index series. The quality prediction unit is used to construct a behavior graph based on the behavior time series and coverage index sequence, perform quality prediction on each behavior segment node through a graph neural network to obtain a predicted quality score, and obtain the actual quality score through correction by the coverage index sequence. A structural mapping unit is used to map the actual quality score to a spatial location to obtain a spatial quality tensor, wherein the spatial location is a three-dimensional location obtained by combining equipment location data and compaction layer number. An anomaly identification unit is used to perform structural consistency analysis on the spatial mass tensor, identify and mark topologically anomalous regions, and output a structural consistency score tensor. The result output unit is used to generate a visual twin layer based on the spatial quality tensor and the structural consistency score tensor, and to perform layer-level annotation and section mapping for abnormal areas.

Images