Highway and municipal road full-life-cycle intelligent construction and management system based on digital twin and BIM (Building Information Modeling)

Abstract

The invention relates to the technical field of electronic data digital processing, and particularly discloses a road and municipal road full-life-cycle intelligent construction and management system based on digital twin and BIM. The system comprises a physical perception layer, a data fusion and twinborn construction layer, an intelligent analysis decision-making layer and a full-period management application layer, and constructs a high-fidelity digital twinborn synchronous with a physical road through unified space-time reference registration, multi-modal data analysis association and dynamic twinborn updating. And structure health prediction, traffic efficiency evaluation and collaborative maintenance decision optimization are carried out on the basis. According to the invention, full-life-cycle digital and intelligent closed-loop management of roads and municipal roads from design to maintenance is realized, and the management efficiency and decision scientificity are improved.

Description

Technical Field

[0001] This invention belongs to the field of electronic data digital processing technology, specifically relating to an intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM. Background Technology

[0002] Digital twin technology, as a core enabling technology connecting the physical world and virtual space, is profoundly transforming the paradigm of urban planning, construction, and management. It achieves the simulation, monitoring, diagnosis, and prediction of the real world by constructing high-fidelity virtual mappings of physical entities. Building Information Modeling (BIM) technology, as a key carrier of three-dimensional geometric and semantic information, provides a solid foundation for the digital representation of infrastructure. The combination of these two technologies brings new possibilities for the full lifecycle management of large-scale infrastructure such as highways and municipal roads.

[0003] In existing technologies, highway and municipal road management systems based on BIM and digital twins have achieved preliminary integration of design models and construction information. However, after roads enter a decades-long operation and maintenance period, the system faces severe challenges: massive amounts of heterogeneous data from traffic monitoring, pavement inspection, environmental sensing, and manual inspections, with varying formats, frequencies, and semantic standards, result in significant bottlenecks in real-time and efficient fusion and synchronization within the digital twin platform. Traditional data integration methods struggle to dynamically parse and correlate these unstructured or semi-structured data with static component information in the BIM model, creating "data silos" and preventing the virtual model from accurately and promptly reflecting the actual service status and performance degradation process of the road structure.

[0004] This lag and inefficiency in data fusion directly restricts the intelligent analysis capabilities based on digital twins. Managers struggle to quickly extract key decision-making information from massive amounts of raw data, such as accurately identifying the development trend of road surface defects, assessing the impact of traffic load on structural safety, or predicting the optimal timing for maintenance needs. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent construction and management system for the entire life cycle of highways and municipal roads based on digital twins and BIM, in order to solve the technical contradiction in the existing technology that the massive amount of heterogeneous data from multiple sources is difficult to integrate and synchronize with the static BIM model in real time during the decades-long operation and maintenance period of highways and municipal roads. This results in the digital twin failing to accurately reflect the real state of the physical entity, thus restricting the ability of intelligent analysis and decision-making.

[0006] To achieve the above objectives, this invention proposes an intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM. This system comprises a physical perception layer, a data fusion and twin construction layer, an intelligent analysis and decision-making layer, and a full-cycle management application layer.

[0007] The physical sensing layer is used to deploy and operate various types of sensing devices throughout the entire lifecycle of highways and municipal roads to continuously collect multi-source, heterogeneous physical world data. During the design and construction phases, sensing devices include positioning modules, stress and strain sensors, and high-definition cameras deployed on construction machinery and critical structural components to collect image data on component installation accuracy, material stress states, and construction progress.

[0008] During the operation and maintenance phase, the sensing equipment further includes a distributed fiber optic sensor network embedded within the road surface structure, traffic flow monitoring equipment deployed on and above the roadside, automated inspection vehicles equipped with multispectral imagers and lidar, and environmental monitoring stations along the road. The distributed fiber optic sensor network collects real-time data on road surface temperature field, strain field, and vibration spectrum at a density of 1024 sampling points per kilometer.

[0009] Traffic flow monitoring equipment collects data on traffic volume, vehicle type, vehicle speed, and lane occupancy at a rate of 30 frames per second. Automatic inspection vehicles collect road surface images and 3D point cloud data twice daily at a preset frequency. Environmental monitoring stations collect data on temperature, humidity, precipitation, and wind speed once per minute. All sensing devices upload the collected raw data streams in real time to the data fusion and twin construction layer via a 5G communication network or an industrial IoT private network.

[0010] The data fusion and twin construction layer receives and processes multi-source heterogeneous data streams from the physical sensing layer, driving high-fidelity, synchronized, and traceable digital twins for highways and municipal roads. The core of this layer includes a unified spatiotemporal reference registration module, a multimodal data parsing and association module, and a dynamic twin update engine. The unified spatiotemporal reference registration module first assigns a unified world coordinate system and a high-precision BeiDou timestamp to all incoming sensing data streams, eliminating spatiotemporal deviations introduced by device clock asynchrony and coordinate system differences. The multimodal data parsing and association module incorporates an ontology knowledge base for the highway and municipal road domain and a set of adaptive parsing rules.

[0011] The ontology knowledge base defines a semantic relationship network between road components, material properties, performance indicators, defect types, traffic events, and environmental elements. The adaptive parsing rule set is configured for the characteristics of each type of data source. For time-series signals from a distributed fiber optic sensor network, the rule set uses pre-defined wavelet transform and feature extraction algorithms to parse them into the mean temperature, maximum strain, and dominant vibration frequency of the corresponding pavement section. For multispectral images from automated inspection vehicles, the rule set uses a pre-trained deep learning segmentation model to identify and locate defects such as cracks, potholes, and ruts, and calculate their geometric dimensions and severity levels. For traffic flow video data, the rule set uses object detection and tracking algorithms to parse out vehicle trajectories and traffic flow statistics for each lane. The parsed structured data fragments, based on their spatial coordinates and timestamps, are automatically associated with corresponding road component instances in the digital twin BIM model by querying the ontology knowledge base. For example, strain data of a certain pavement section is associated with the structural layer components of that pavement, and identified cracks are associated with specific pavement panel components.

[0012] The dynamic twin update engine maintains digital twins of highways and municipal roads, which are based on the initial BIM models from the design and construction phases for both geometry and semantics. The engine receives associated data fragments from the multimodal data parsing and association module and updates the twins in real-time or near real-time, according to data type and update strategy. For high-frequency changing data such as traffic flow and ambient temperature, the engine directly updates the attribute tables of the corresponding components in the twin using a streaming processing approach.

[0013] For data reflecting the evolution of road surface defects and structural strain, the engine integrates them with historical data sequences to construct a time-series state curve for the component in the twin. Based on the curve trend, it predicts the state value at a future moment, achieving advanced simulation of the twin's state. All data injection, state update, and prediction operations on the twin are recorded in an immutable log based on blockchain technology, forming a full lifecycle data traceability chain.

[0014] The intelligent analysis and decision-making layer is used to perform in-depth state assessment, performance prediction, and decision optimization analysis based on the digital twin provided by the data fusion and twin construction layer, which is synchronized with the physical world and rich in semantics. This layer includes a structural health diagnosis and prediction module, a traffic operation efficiency assessment module, and a collaborative maintenance decision optimization module. The structural health diagnosis and prediction module accesses real-time strain, vibration, and apparent defects data of key components such as pavements, bridges, and tunnels in the digital twin.

[0015] The module incorporates a hybrid model based on physical mechanisms and data-driven approaches. First, it calculates the theoretical stress distribution of the component based on the constitutive relations of material mechanics and real-time traffic load data. Then, it compares the theoretical stress with the measured strain from fiber optic sensing, identifying structural anomaly areas through residual analysis. Furthermore, the module utilizes a long short-term memory network, taking the component's historical state sequence and future traffic load predictions as input, to perform rolling predictions of key performance indicators such as crack propagation rate and material fatigue life, outputting performance degradation curves and remaining service life estimates for the next 30, 90, and 180 days.

[0016] The traffic operation efficiency assessment module calculates a dynamic traffic operation index based on real-time traffic flow data, historical accident data, and road geometry information from a digital twin. This index integrates multiple dimensions, including average vehicle speed, congestion duration, and accident risk probability. The module uses a microscopic traffic simulation model to simulate different traffic control schemes within the digital twin, such as signal timing adjustments, lane function changes, or speed limit strategies, and predicts the improvement effect of each scheme on the traffic operation index, providing quantitative data for traffic management.

[0017] The collaborative maintenance decision optimization module receives a maintenance requirement list from the structural health diagnosis and prediction module and a traffic impact prediction from the traffic operation efficiency assessment module. The module abstracts the road network as a graph model, where nodes represent road segments to be maintained and edge weights represent the traffic delay costs caused by maintenance work. The module constructs a multi-objective optimization model with the goal of minimizing total social cost. The objective function includes direct maintenance costs, user travel delay costs, and the increase in structural risk costs due to unaddressed delays. This model is solved using an improved genetic algorithm in a simulation environment provided by a digital twin, outputting the optimal maintenance work scheduling plan, resource allocation plan, and traffic diversion plan for future maintenance cycles.

[0018] The full-cycle management application layer provides a web-based collaborative work platform and visual interactive interface for different user roles in the design, construction, operation, and maintenance phases. In the design phase, the platform supports reverse feedback based on historical performance data from the digital twin, optimizing and iterating the design parameters of newly constructed roads. In the construction phase, the platform compares construction progress and quality inspection data with 4D construction simulation models to achieve real-time early warnings of schedule deviations and quality risks. In the operation phase, the platform centrally displays the structural health, traffic operation index, and environmental status of the entire road in a dashboard format, and pushes early warning information and decision-making suggestions from the intelligent analysis and decision-making layer. In the maintenance phase, the platform visualizes and simulates the solutions generated by the collaborative maintenance decision optimization module, and generates digital maintenance instruction packages containing work orders, drawings, and process requirements, which are directly sent to the mobile terminals of maintenance units to guide on-site operations. All management instructions, approval processes, and execution feedback generated in all phases are recorded in the platform and linked to the data traceability chain of the digital twin, forming a complete management closed loop.

[0019] In one embodiment of the present invention, the adaptive parsing rule set in the multimodal data parsing and association module is deployed using a containerized microservice architecture. The parsing algorithm corresponding to each data source is encapsulated as an independent microservice, such as a fiber optic signal parsing microservice, an image defect recognition microservice, and a traffic flow parsing microservice. These microservices are registered and discovered through a lightweight API gateway. When a new type of sensing device is connected to the system, only the corresponding parsing microservice needs to be developed and deployed to the container cluster, and its rules registered with the gateway through the configuration center. This enables plug-and-play parsing support for new data sources without requiring a complete reconstruction of the entire data fusion layer.

[0020] Furthermore, the data traceability chain based on blockchain technology in the dynamic twin update engine is implemented as follows: The system creates a dedicated consortium blockchain network for digital twins of highways and municipal roads, with participating nodes including construction units, design units, construction units, operation units, and regulatory agencies. Each data update operation on the twin, including the data source, timestamp, hash value of the operation content, and associated component identifier, is packaged into a data block. This block is added to the chain after being verified by the consensus mechanism. Any participating party can, with authorization, query the status data of any component at any historical moment and its complete update record chain, ensuring the authenticity, non-repudiation, and auditability of the data throughout its entire lifecycle.

[0021] Furthermore, the hybrid model in the structural health diagnosis and prediction module employs an attention-enhanced spatiotemporal graph convolutional network in its data-driven portion. This network uses the road network topology as a graph structure, sensor data from each road segment as node features, and traffic flow data as edge features. The graph convolutional layers extract the spatial correlation of the road network, the temporal convolutional layers extract the temporal dependence of the sensor data, and the attention mechanism dynamically weights the importance of different time steps and different spatial neighbors to the current prediction, thereby more accurately capturing the spatiotemporal propagation patterns of performance degradation and improving the accuracy of long-term predictions.

[0022] Furthermore, the multi-objective optimization model of the collaborative maintenance decision optimization module introduces a robust optimization strategy based on digital twin simulation. The model not only considers the predicted average traffic flow but also simulates various possible traffic demand scenarios using a digital twin, including peak hours, severe weather, and unexpected accidents. During the solution process, the optimization algorithm ensures that the increase in total social cost does not exceed a preset threshold under the most unfavorable simulation scenario, thereby generating a more robust maintenance decision scheme with stronger anti-interference capabilities and improving the reliability of decisions in real-world complex environments.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention systematically solves the challenge of integrating massive amounts of multi-source heterogeneous data with static BIM models during the operational phase by constructing a data fusion and twin construction layer that includes unified spatiotemporal benchmark registration, multimodal data parsing and association, and a dynamic twin update engine. Utilizing an ontology knowledge base and adaptive parsing rule set, unstructured sensor data and image data are transformed in real time into structured information with clear semantics, and automatically associated with specific components in the digital twin. This process achieves a leap from physical world data to information, and then to digital twin knowledge, making the digital twin no longer a static model, but a high-fidelity dynamic mirror that can evolve synchronously with physical roads and accurately map their actual service status, providing a reliable data foundation for advanced intelligent analysis.

[0024] 2. The intelligent analysis and decision-making layer constructed in this invention deeply utilizes a synchronously updated digital twin, achieving a leap from condition monitoring to predictive decision-making. Structural health diagnosis employs a hybrid model combining mechanism and data-driven approaches, improving the accuracy and interpretability of anomaly identification and lifespan prediction. Traffic operation efficiency assessment is simulated and extrapolated within the digital twin environment, enabling quantitative pre-assessment of management strategies. In particular, the collaborative maintenance decision optimization module incorporates structural safety, traffic impact, and economic costs into a unified multi-objective optimization framework, solving and robustly verifying these objectives within the simulation environment provided by the digital twin, generating a globally optimal and interference-resistant maintenance plan. This fundamentally changes the traditional experience-dependent and delayed-response maintenance model, achieving a radical transformation in highway and municipal road maintenance from passive response to proactive prediction, and from local optimization to global coordination.

[0025] 3. This invention, through a full-cycle management application layer, deeply integrates management activities at each stage—design, construction, operation, and maintenance—with a digital twin, forming a closed-loop digital management system covering the entire lifecycle of highways and municipal roads. Data and knowledge generated at each stage are continuously accumulated within the digital twin and used to optimize preceding stages, achieving continuous improvement based on real performance feedback. Simultaneously, a blockchain-based data traceability mechanism ensures the credibility and auditability of data across the entire chain. This system is not only a monitoring and analysis tool but also a collaborative work platform and decision support system, significantly improving the overall efficiency, precision, and long-term economic benefits of highway and municipal road planning, construction, management, and maintenance. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall technical solution architecture of the intelligent construction and management system for the entire life cycle of highways and municipal roads based on digital twins and BIM proposed in this invention. Figure 2 This is a schematic diagram of the core principle framework of the data fusion and twin construction layer in this invention; Figure 3 This is a logical flow diagram of the intelligent analysis and decision-making layer in this invention; Figure 4 This is a schematic diagram of the multi-level interaction relationship and data flow between the physical perception layer, the data fusion and twin construction layer and the intelligent analysis and decision-making layer in this invention; Figure 5 This is a schematic diagram comparing the principle and effect of the collaborative maintenance decision optimization module in the digital twin simulation environment. Detailed Implementation

[0027] This embodiment details the specific implementation of an intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and building information models. Please refer to the appendix. Figures 1 to 5The system is logically divided into four layers, from bottom to top: the physical perception layer, the data fusion and digital twin construction layer, the intelligent analysis and decision-making layer, and the full-cycle management application layer. These four layers work together to form a complete closed loop, from physical world data acquisition to high-fidelity mapping in the digital space, to intelligent analysis and decision-making, and finally to guiding full-cycle management practices.

[0028] The physical sensing layer serves as the interface between the system and physical highways and municipal roads. Its core task is to deploy and operate a series of heterogeneous sensing devices throughout all stages of the road's lifecycle to continuously and multidimensionally collect data streams reflecting the physical state of the road. During the design and construction phases, sensing deployment focuses on monitoring the quality and progress of the construction process. Specifically, high-precision GPS modules and inertial measurement units are installed on key construction machinery such as large pavers and rollers to track the machinery's trajectory and attitude in real time, ensuring that the construction process meets design requirements.

[0029] In structural components such as precast bridge beams, tunnel segments, and key sections of roadbeds, resistive or fiber optic stress-strain sensors are pre-embedded or installed. These sensors continuously monitor stress changes in components during hoisting, installation, and initial load-bearing processes at a sampling frequency of 10 times per second, preventing overstress damage. Simultaneously, high-definition spherical network cameras are deployed at the main construction work surfaces to continuously capture construction images at 1080p resolution and a rate of 25 frames per second. The image stream is transmitted in real-time to the data center via on-site 5G communication base stations or industrial-grade wireless LAN access points.

[0030] Entering the operation and maintenance phase, the density and complexity of the sensing network are significantly increased, aiming to achieve comprehensive and three-dimensional perception of the service status of road infrastructure. Distributed fiber optic sensor networks are embedded within the pavement structure, running longitudinally and laterally along the road. This network continuously measures temperature and strain along the fiber optic path with a spatial resolution of 1024 sampling points per kilometer. The sensing fibers are typically laid between the bottom of the asphalt surface layer and the top of the base layer, or near the joints of cement concrete pavements. A fiber optic sensor demodulator acquires the raw optical signal at a rate of 100 times per second and demodulates the temperature and strain values ​​at each point along the fiber in real time using wavelet transform, forming a high-density temperature and strain field data matrix covering the entire road section.

[0031] An array of traffic flow monitoring equipment, consisting of high-definition checkpoint cameras, microwave traffic detectors, and laser scanners, is deployed along the roadside and overhead. The checkpoint cameras capture vehicle images at 30 frames per second, running license plate recognition and vehicle type classification algorithms via local edge computing units. Microwave traffic detectors scan 60 times per second to detect vehicle presence, speed, and occupancy in each lane. Laser scanners provide 3D point cloud data of vehicle outlines. These devices, via a dedicated fiber optic network or 5G network, aggregate the processed structured traffic flow data, including lane flow rates, average vehicle speeds, vehicle type composition, and time occupancy in seconds, to the regional traffic control center.

[0032] To acquire road surface conditions, the system is equipped with an automated inspection vehicle featuring a multispectral imager and a 3D LiDAR. The inspection vehicle travels along a predetermined route according to a pre-set schedule of twice daily, or as part of a special inspection plan triggered by weather events. The multispectral imager acquires road surface images in the visible and near-infrared bands, achieving an image resolution of 4096×2160 pixels. The 3D LiDAR scans the road surface at a rate of 320,000 points per second, generating 3D point cloud data with millimeter-level precision.

[0033] The vehicle integrates both a global positioning system and an inertial navigation system, assigning precise geographic coordinates and attitude information to each frame of image and each piece of point cloud data. Inspection data is transmitted in real time via the onboard 5G communication module, or transmitted in batches via a high-speed wireless network after the inspection is completed.

[0034] In addition, an environmental monitoring station is deployed every 2 kilometers along the road. Each station integrates a temperature sensor, a relative humidity sensor, a tipping bucket rain gauge, and an ultrasonic anemometer. The temperature and humidity sensors collect data once per minute, the rain gauge records the cumulative rainfall per minute, and the anemometer outputs a set of data every 10 seconds. All environmental data is uploaded to a cloud platform via narrowband IoT or 4G networks.

[0035] The data fusion and twin construction layer is the core of the system. Its responsibility is to receive and process massive, multi-source, and heterogeneous raw data streams from the physical sensing layer, and to drive a high-fidelity, traceable digital twin that evolves synchronously with the physical path. Please refer to the appendix. Figure 2 This layer consists of three core modules: a unified spatiotemporal benchmark registration module, a multimodal data parsing and association module, and a dynamic twin update engine.

[0036] The unified spatiotemporal reference registration module is the first processing gate for all data flowing into the system. This module maintains a unified world coordinate system based on the national 2000 geodetic coordinate system and connects to the high-precision time synchronization service of the BeiDou Navigation Satellite System, generating timestamps with nanosecond-level accuracy. For each incoming raw data stream, the module first performs spatiotemporal tag attachment and alignment operations.

[0037] For data that inherently carries GPS coordinates, such as data collected by automated inspection vehicles, the module transforms its coordinates to a unified world coordinate system. For data with only relative or logical position information, such as strain sensor data with a specific number, the module queries a pre-defined device space registry to map it to its absolute position in the world coordinate system. Simultaneously, the module injects a unified BeiDou timestamp into all data packets, overwriting or calibrating any potentially asynchronous local clocks of the devices. This process completely eliminates spatiotemporal reference differences introduced by variations in sensing device brand, model, and deployment location, ensuring that all subsequent processing is performed within a consistent spatiotemporal framework.

[0038] The multimodal data parsing and association module is responsible for transforming raw data into structured information with clear engineering semantics. The core of this module consists of two parts: a deeply customized ontology knowledge base for the highway and municipal road domain, and a set of adaptive parsing rules based on a containerized microservice architecture. The ontology knowledge base is constructed using a resource description framework and a web ontology language, formally defining all entity types, attributes, and relationships involved in highways and municipal roads. Entity types include roads, lanes, pavements, roadbeds, bridges, tunnels, drainage facilities, traffic signs, and sensors. Attributes include geometric dimensions, material type, design strength, construction time, and maintenance history. Relationships include "located in," "contains," "bears load," "monitored," and "has defects." This knowledge base provides a semantic graph foundation for the association of data fragments.

[0039] The adaptive parsing rule set is implemented as a microservice. The system develops and deploys independent parsing microservices for each type of mainstream data source. For example, the fiber optic signal parsing microservice specifically handles raw light intensity or wavelength offset signals from distributed fiber optic sensor networks. This microservice encapsulates signal preprocessing, noise filtering, wavelet transform demodulation, and feature extraction algorithms. It receives the raw signal stream with spatiotemporal labels, first performs moving average filtering to suppress high-frequency noise, then applies discrete wavelet transform to decompose the signal into different frequency bands, extracts the effective components reflecting temperature and strain from specific frequency bands, and finally calculates the average temperature, maximum strain, minimum strain, and root mean square value of each sensing point over the past second. The output is a structured temperature and strain record, with each record containing fields such as location coordinates, timestamp, temperature value, and strain value.

[0040] The image-based defect recognition microservice specifically processes road surface images from automated inspection vehicles or fixed cameras. This microservice loads a pre-trained deep learning semantic segmentation model based on an encoder-decoder architecture, trained using hundreds of thousands of road surface images labeled with defect types such as cracks, potholes, repairs, and ruts. Upon receiving an image, the service first performs geometric correction and illumination normalization preprocessing before inputting the image into the model. The model outputs a defect category probability map for each pixel. A post-processing algorithm generates bounding boxes and polygonal outlines of defect instances based on these probability maps and calculates their geometric features, such as the length, maximum width, and area of ​​cracks; the area and depth estimates of potholes; and the cross-sectional curve and average depth of ruts. Finally, a structured defect detection report is output, including a unique defect identifier, type, severity level, polygonal outline coordinates, size measurement, and discovery time.

[0041] The traffic flow analysis microservice targets video streams and microwave radar data. It runs lightweight object detection and multi-object tracking algorithms, such as the YOLO series of deep learning algorithms and the SORT tracking framework. The service processes video frame by frame, identifying the bounding box and vehicle type of each vehicle, and associating the same vehicle across consecutive frames to form a trajectory. Combined with radar speed measurement data, it calculates the vehicle's speed, acceleration, and lane location for each trajectory. The service aggregates statistics every 1 second, outputting a traffic flow snapshot containing fields such as time slice start time, lane number, flow rate, average speed, time occupancy, and vehicle type distribution.

[0042] All parsing microservices are registered, discovered, and load-balanced through a lightweight application programming interface (API) gateway. The gateway provides a unified API endpoint. When the data fusion layer receives a new data packet, it routes it to the corresponding parsing microservice through the gateway based on its data source type identifier. The parsed structured data fragments, called "information atoms," carry semantic information extracted from the original data but are not yet associated with the digital twin.

[0043] The association process follows the parsing process. Each information atom carries precise spatial coordinates and a timestamp. The association engine uses these coordinates as a center to perform a spatial query in the digital twin's building information model spatial database, searching for the nearest road component instance or one that spatially contains that point. For example, a crack information atom located near the centerline of a lane will be associated with the building information model component representing the pavement layer of that lane. A strain information atom from a fiber optic sensor embedded near a bridge pier foundation will be associated with that pier component. The association process not only relies on spatial location but also performs semantic verification by querying the ontology knowledge base. For example, an information atom of the "crack" defect type will only be associated with component types defined as "prone to cracks" in the knowledge base, such as pavement or bridge decks, and not with streetlights. After successful association, the information atom is assigned a globally unique identifier pointing to the target building information model component, thus completing the binding from data to a specific component in the digital twin.

[0044] The dynamic twin update engine is the "heart" of the digital twin, maintaining digital twin instances of highways and municipal roads. This twin uses the detailed building information model delivered during the design and construction phases as its initial geometric and semantic basis. The engine continuously monitors the associated information atom streams output from the multimodal data parsing and association module. Based on the data type of the information atoms and the preset update strategy, the engine performs update operations on the twin in different modes.

[0045] For frequently changing dynamic attributes, such as traffic flow, average vehicle speed, and ambient temperature, the engine employs a streaming update mode. Upon arrival of an information atom, the engine directly locates the corresponding component's attribute table in the twin, writing the new value or continuously updating the attribute. For example, the traffic flow attribute of a lane will be continuously refreshed with a 1-second cycle.

[0046] For data reflecting state evolution or cumulative effects, such as structural strain values, defect sizes, and material performance indicators, the engine employs a time-series integration and trend update model. The engine not only records current values ​​but also creates and maintains a time-series database within the digital twin, storing historical values ​​for that indicator. When new information atoms arrive, the engine first appends their value to the historical sequence, and then, based on this sequence, applies prediction algorithms to estimate future states. Specifically, for pavement strain, the engine might use an autoregressive integral moving average model or a long short-term memory network to predict the strain trend at that point in the next hour. For crack length, the engine might use an extended model based on fracture mechanics or a data-driven regression model to predict its extension length over the next 30 days. These predicted values, as "advanced states," are also written into the digital twin, enabling the digital twin to not only reflect the present but also predict the near future to some extent, achieving advanced state simulation.

[0047] All operations on the digital twin, including data injection, attribute updates, and state prediction, are considered "transactions." The engine submits key information about these transactions, including the transaction's unique identifier, timestamp, operation type, globally unique identifier of the associated component, data hash value, and the digital signature of the operation initiator, to the data traceability chain system. This system is built on a consortium blockchain network. Nodes in the network are jointly maintained by the construction unit, design unit, construction unit, operation unit, maintenance unit, and government regulatory agencies. Transactions are packaged into blocks, verified through a practical Byzantine fault-tolerant consensus mechanism, and then appended to the chain. This chain constitutes an immutable and traceable audit log of all historical state changes of the digital twin. Any authorized party can query this chain to trace the state and changes of any component at any historical moment, ensuring the authenticity and credibility of the data throughout its entire lifecycle.

[0048] The intelligent analysis and decision-making layer, based on a real-time synchronized and semantically rich digital twin provided by the data fusion and twin construction layer, performs in-depth analysis, evaluation, and optimization calculations. Please refer to the appendix. Figure 3 This layer contains three core functional modules: structural health diagnosis and prediction module, traffic operation efficiency assessment module, and collaborative maintenance decision optimization module.

[0049] The Structural Health Diagnosis and Prediction module focuses on the physical condition assessment and life prediction of road infrastructure. The module subscribes to real-time sensor data and historical state sequences of key components, such as pavement structural layers, bridge main beams, piers, and tunnel linings, from digital twins. Its core is a hybrid model integrating physical mechanisms and data-driven methods. The mechanism part is based on finite element analysis, calculating the theoretical stress-strain distribution field of components based on their geometric dimensions, material parameters, and real-time traffic loads recorded in the digital twin. The data-driven part employs an attention-enhanced spatiotemporal graph convolutional network. This network abstracts the road network as a graph, where each monitored road segment or component is a node. Node features include time series data from sensors such as strain, vibration frequency, and temperature at that location. Edges connecting nodes represent spatial adjacency or functional relationships, and edge features can include traffic flow.

[0050] The graph convolutional layers of the network are responsible for capturing the spatial dependencies between different nodes in the road network. For example, congestion on an upstream road segment may lead to changes in vehicle load distribution on a downstream road segment. Temporal convolutional layers are responsible for extracting temporal evolution patterns from each node's own historical data sequence. An attention mechanism is introduced into the network's temporal and spatial dimensions, dynamically calculating the contribution weights of different historical time points to the current prediction, as well as the influence weights of spatially different neighboring nodes on the target node. This enables the network to more accurately identify abnormal propagation paths and key drivers of performance degradation.

[0051] The workflow of the hybrid model is as follows: First, real-time sensor strain data is compared with the theoretical strain calculated by the mechanistic model to calculate residuals. The spatial distribution and temporal patterns of the residuals are input into a spatiotemporal graph convolutional network. The network analyzes these residual patterns, and if it identifies spatiotemporal characteristics similar to historical damage cases or typical failure modes, it triggers an anomaly alarm and locates the abnormal area. Second, for identified or potentially weak components, the module utilizes their complete historical state sequence, combined with predictions of future traffic loads, to perform rolling predictions through the spatiotemporal graph convolutional network. The network outputs future evolution curves of key performance indicators, such as crack width, material stiffness reduction factor, and fatigue damage accumulation. Based on these curves, the module can estimate the remaining service life of the component under specified performance thresholds, such as predicting the cumulative number of standard axle loads or time days required for a road surface to reach a certain critical value under current traffic loads.

[0052] The traffic operation efficiency assessment module focuses on analyzing the networked operation service level of roads. The module takes as input real-time traffic flow data, historical traffic accident records, road geometry, and signal control schemes from a digital twin. Its core function is to calculate a comprehensive dynamic traffic operation index. This index is a weighted composite of multiple sub-indicators, including the ratio of average travel speed to free-flow speed, the proportion of congestion duration, the travel time reliability index, and the probability of accident risk predicted based on historical data and real-time conditions. Each sub-indicator is normalized, and the final index value ranges from 0 to 100, with higher values ​​indicating better operational efficiency.

[0053] Furthermore, this module integrates a micro-traffic simulation engine. When traffic managers consider adjusting signal timings in a certain area, implementing tidal flow lanes, or changing speed limits, they can create simulation scenarios in a digital twin. The module imports current traffic demand and control schemes into the simulation engine and performs simulations within the digital twin's environment. The simulation outputs predicted traffic flow states after implementing the new scheme and recalculates traffic operation indices. By comparing the changes in indices before and after the implementation of the scheme, the module can quantitatively evaluate the potential effects of different management strategies, providing intuitive and quantitative evidence for decision-making.

[0054] The collaborative maintenance decision optimization module is a high-level decision center that connects structural safety and traffic operations, enabling optimized allocation of global resources. Please refer to the appendix. Figure 5 The input to this module comes from two aspects: first, the maintenance requirement list for a future period of time output by the structural health diagnosis and prediction module, which specifies the road sections that need maintenance, the recommended maintenance type, the estimated operation time, the time window requirements, and the urgency level; second, the traffic flow prediction and delay sensitivity analysis of the road network for each time period and each road section provided by the traffic operation efficiency assessment module.

[0055] The module abstracts the entire road network as a graph model. Each node in the graph represents a road segment or component to be maintained, and nodes have attributes such as maintenance cost, operation time, and optimal construction period. The edges connecting nodes represent the actual connectivity in the road network, and each edge is assigned a weight that represents the additional traffic delay cost caused to the path of the destination node when maintenance is performed on the path of the starting node. The delay cost is calculated through microscopic traffic simulation, taking into account factors such as increased detour distance, reduced speed, and increased congestion, and is monetized as social travel cost.

[0056] The module's decision-making objective is to schedule all necessary maintenance work within a future planning period, such as the next three months, and determine the specific start time and required resources for each work, while generating corresponding traffic management plans. This is a typical multi-objective optimization problem. The objective function of the optimization model aims to minimize the total social cost, which consists of three parts: the first part is the direct maintenance cost, including labor, material, and machinery costs; the second part is the user travel delay cost, which is the economic cost of social travel time loss caused by maintenance work occupying road resources; and the third part is the risk cost, which is the potential cost increment caused by further deterioration of the structural condition due to the postponement of certain emergency maintenance work, which may lead to larger-scale repairs or safety accidents in the future.

[0057] The optimization model is solved in a simulation environment provided by a digital twin. The solution algorithm employs an improved genetic algorithm. The algorithm first generates a set of random maintenance task scheduling schemes as an initial population. Each scheme is a chromosome, with genes encoding information such as the start time and resource allocation for each task. Then, the algorithm launches a simulation instance for each chromosome scheme in the digital twin, simulating traffic flow changes when the scheme is executed in the actual road network, thereby accurately calculating the total social cost corresponding to the scheme as its fitness value. The population is iteratively evolved through genetic operations such as selection, crossover, and mutation. To enhance the robustness of the decision schemes, the model introduces a simulation-based robust optimization strategy. When evaluating each scheme, not only average traffic demand scenarios are simulated, but also a set of extreme or unfavorable scenarios, such as weekday morning rush hour, heavy rain, and sudden accidents on adjacent road sections. Optimization constraints require that, under the most unfavorable simulation scenario, the increase in the total social cost of the finally selected scheme compared to the baseline scheme must not exceed a preset threshold. After thousands of iterations, the algorithm converges to one or a set of Pareto optimal solutions. The system selects the optimal solution that balances cost, robustness, and operability as the final collaborative maintenance decision. This decision plan details the work order, time window, required equipment and personnel, specific traffic closure and diversion routes, and backup contingency plans for each maintenance task.

[0058] The full-cycle management application layer is a unified interactive portal for all types of users. It is a collaborative work platform built on web technology. The platform provides customized visual interfaces and workflow tools for designers, construction companies, operators, maintenance providers, and regulators.

[0059] During the design phase, the platform provides reverse query and analysis capabilities for historical data from digital twins. Designers can query data such as the actual performance, common defects, and maintenance frequency of similar road sections after many years of operation. This real-world feedback data can serve as an important basis for optimizing the design parameters of newly constructed roads, such as adjusting the thickness of pavement structural layers, selecting more durable materials, and optimizing drainage design, thus achieving iterative design based on performance feedback.

[0060] During the construction phase, the platform integrates 4D construction management functions. It links the construction schedule with the building information model to generate dynamic construction simulation animations. On-site, actual progress data is collected through the perception layer, such as the completion status of component installation and material arrival records.

Claims

1. A smart construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM, characterized in that: include: The physical sensing layer is used to deploy and operate various types of sensing devices at all stages of the entire life cycle of highways and municipal roads to continuously collect multi-source heterogeneous physical world data and upload the collected raw data streams in real time. The data fusion and twin construction layer is used to receive and process multi-source heterogeneous data streams from the physical sensing layer, and drive high-fidelity, synchronized, and traceable digital twins of highways and municipal roads. The intelligent analysis and decision-making layer is used to perform status assessment, performance prediction, and decision optimization analysis based on the digital twin provided by the data fusion and twin construction layer. The full-cycle management application layer provides a web-based collaborative work platform and visual interactive interface for different user roles in the design, construction, operation and maintenance stages. It is used to realize reverse design feedback based on digital twins, early warning of construction progress and quality risks, centralized display of operation status and push of decision suggestions, as well as visualization of maintenance plan simulation and digital command issuance.

2. The intelligent construction and management system for the entire life cycle of highways and municipal roads based on digital twins and BIM as described in claim 1, wherein the data fusion and twin construction layer includes a unified spatiotemporal benchmark registration module, a multimodal data parsing and association module, and a dynamic twin update engine; The unified spatiotemporal reference registration module is used to assign a unified world coordinate system and a high-precision BeiDou timestamp to all accessed sensing data streams. The multimodal data parsing and association module has a built-in ontology knowledge base for highway and municipal road fields and a set of adaptive parsing rules. It is used to parse the raw data stream into structured data fragments and automatically associate the structured data fragments with the corresponding road component instances in the digital twin BIM model by querying the ontology knowledge base based on their spatial coordinates and timestamps. The dynamic twin update engine is used to maintain digital twins of highways and municipal roads, and receives associated data fragments from the multimodal data parsing and association module. It updates the twins in real time or near real time according to the data type and update strategy, and records all data injection, state update and prediction operations on the twins in an immutable log based on blockchain technology, forming a full life cycle data traceability chain.

3. The intelligent construction and management system for the entire life cycle of highways and municipal roads based on digital twins and BIM as described in claim 2, wherein the intelligent analysis and decision-making layer includes a structural health diagnosis and prediction module, a traffic operation efficiency evaluation module, and a collaborative maintenance decision optimization module; The structural health diagnosis and prediction module is used to access real-time data of key components in the digital twin and to perform structural anomaly identification and rolling prediction of performance indicators based on the built-in physical mechanism and data-driven hybrid model. The traffic operation efficiency evaluation module is used to calculate a dynamic traffic operation index based on real-time traffic flow data, historical accident data and road geometry information in the digital twin, and to predict the improvement effect of each scheme by simulating different traffic control schemes in the digital twin. The collaborative maintenance decision optimization module receives a maintenance requirement list from the structural health diagnosis and prediction module and a traffic impact prediction from the traffic operation efficiency assessment module. It constructs a multi-objective optimization model with the goal of minimizing total social cost, and solves the model using an improved genetic algorithm in the simulation environment provided by the digital twin, outputting the optimal maintenance operation scheduling scheme.

4. The intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM as described in claim 3, characterized in that, In the physical sensing layer, the sensing devices deployed during the design and construction phases include positioning modules, stress and strain sensors, and high-definition camera devices deployed on construction machinery and key structural parts. The sensing equipment deployed during the operation and maintenance phase further includes distributed fiber optic sensor networks embedded in the road surface structure, traffic flow monitoring equipment deployed on and above the roadside, automated inspection vehicles equipped with multispectral imagers and lidar, and environmental monitoring stations deployed along the road.

5. The intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM as described in claim 4, characterized in that, The adaptive parsing rule set in the multimodal data parsing and association module is deployed using a containerized microservice architecture; the parsing algorithm corresponding to each data source is encapsulated as an independent microservice, and the microservices are registered and discovered through a lightweight API gateway.

6. The intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM as described in claim 5, characterized in that, The data traceability chain based on blockchain technology in the dynamic twin update engine is implemented as follows: The system creates a dedicated consortium blockchain network for digital twins of highways and municipal roads, with participating nodes including construction units, design units, construction units, operation units, and regulatory agencies; each data update operation on the twin is packaged into a data block, which is then added to the chain after being verified by the consensus mechanism.

7. The intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM as described in claim 6, characterized in that, The hybrid model in the structural health diagnosis and prediction module employs a spatiotemporal graph convolutional network with an attention mechanism enhanced in its data-driven part. The network uses the road network topology as a graph structure, with sensor data from each road segment as node features and traffic flow data as edge features; In the network, graph convolutional layers extract the spatial correlation of the road network, temporal convolutional layers extract the temporal dependence of sensor data, and the attention mechanism dynamically weights the importance of different time steps and different spatial neighbors to the current prediction.

8. The intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM as described in claim 7, characterized in that, The working process of the structural health diagnosis and prediction module is as follows: First, based on the constitutive relations of mechanics of materials and real-time traffic load data, the theoretical stress distribution of the component is calculated; Then, the theoretical stress is compared with the strain measured by fiber optic sensing, and the abnormal areas of the structure are identified through residual analysis. By further utilizing long short-term memory networks, and taking the historical state sequence of components and future traffic load predictions as inputs, key performance indicators are predicted in a rolling manner.

9. The intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM as described in claim 8, characterized in that, The multi-objective optimization model of the collaborative maintenance decision optimization module introduces a robust optimization strategy based on digital twin simulation. The model simulates various possible traffic demand scenarios using digital twins. When optimizing the algorithm, it is constrained to ensure that the increase in total social cost does not exceed a preset threshold under the most unfavorable simulation scenario.

10. The intelligent construction and management system for the entire lifecycle of highways and municipal roads based on digital twins and BIM as described in claim 9, characterized in that, The multi-objective optimization model constructed by the collaborative maintenance decision optimization module has objective functions including direct maintenance costs, user travel delay costs, and increased structural risk costs due to unprocessed delays.

Images