Short line method cantilever construction bridge line shape prediction and regulation method based on digital twinning

Abstract

The invention provides a digital twinning-based short-line-method cantilever construction bridge line shape prediction and regulation method. The method comprises the steps of obtaining geometric data, physical attribute data, construction behavior data and engineering rule data of short-line-method prefabricated beam segments; establishing an initial digital twinborn model; calibrating parameters of the initial digital twinborn model; constructing and operating a multi-scale hybrid prediction model, and outputting a predicted value and an uncertainty interval of the linear shape of the whole construction process based on a digital twinborn model; judging whether the predicted value has an over-limit deviation or not; when the over-limit deviation exists, generating and optimizing a regulation and control scheme, converting the regulation and control scheme into an executable instruction, and issuing the executable instruction to site construction equipment; and collecting result data after the instruction is executed, and reflowing the result data to update the digital twinborn model and the hybrid prediction model to form closed-loop control. According to the method, the hysteresis problem of post deviation correction of a traditional method is solved, real-time prediction, intelligent diagnosis and active regulation and control of the construction line shape are achieved, and a complete prediction-regulation-optimization closed loop is formed.

Description

Technical Field

[0001] This invention relates to the field of building construction technology, and in particular to a method for predicting and controlling the alignment of cantilever bridges using the short-line construction method based on digital twins. Background Technology

[0002] Precast concrete with cantilever assembly is a common construction method for superstructures such as long-span continuous beams and cable-stayed bridges. This method involves multiple stages, including segmental precasting, transportation, temporary assembly, formwork installation, prestressing tensioning, and closure. However, discrepancies exist between actual geometry, materials, construction conditions, and design assumptions. Factors such as temperature, creep shrinkage, tension loss, formwork deflection, temporary support settlement, adhesive layer thickness fluctuations, construction sequence, and equipment status can all couple and affect the structural alignment. Existing monitoring methods largely rely on offline theoretical calculations before construction and post-construction corrections based on scattered on-site measurements. These methods suffer from problems such as model-to-actual-bridge disconnect, predictive lag, unexplained causes of deviations, reliance on experience for control, and insufficient closed-loop management. This can easily lead to deviations exceeding limits, rework, quality risks, and safety hazards.

[0003] Chinese invention application CN110777669A, published on February 11, 2020, discloses a method for controlling the alignment of prefabricated cantilever beams for short-line matching in high-speed railway construction. This method relates to the field of high-speed railway bridge construction control technology. The method includes beam segment assembly monitoring: marking a pair of measuring points on the centerline of the matching segment, marking another pair of measuring points at equidistant locations on both sides of the pair, and marking three pairs of measuring points at the same locations on the segments to be installed; beam segment assembly adjustment: installing the segments to be installed, then adjusting the elevation and planar position of the six pairs of measuring points, measuring the deviation values ​​of each pair of measuring points and adjusting accordingly, followed by reinforcement; and beam segment assembly correction: calculating and correcting the error of each pair of measuring points on the segments to be installed. However, this invention application not only fails to predict the alignment before or during construction, making it difficult to identify and prevent deviations in advance, but also relies mainly on post-construction measurements and manual corrections. This results in problems such as control lag, reliance on experience, lack of predictability and closed-loop optimization capabilities, making it difficult to achieve high-precision and proactive control of the alignment. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins. This method solves the problem that existing cantilever assembly alignment control methods cannot predict the alignment before or during construction.

[0005] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0006] A method for predicting and controlling the alignment of a cantilever bridge constructed using the short-line method based on digital twins includes the following steps: S1, acquiring geometric data, physical property data, construction behavior data, and engineering rule data of precast beam segments using the short-line method; S2, establishing an initial digital twin model that integrates a geometric model, a structural mechanics model, and construction procedure rules based on the data; S3, deploying a multi-source sensor network to collect real-time data during the construction period, and dynamically calibrating the parameters of the initial digital twin model using a data assimilation algorithm to achieve synchronous evolution between the model and the actual bridge; S4, constructing and running a multi-scale hybrid prediction model, and outputting the predicted values ​​and uncertainty ranges of the alignment throughout the construction process based on the digital twin model; S5, determining whether the predicted values ​​have excessive deviations based on the target alignment and an adaptive threshold; S6, when excessive deviations exist, performing intelligent diagnosis to identify the causes of the deviations, automatically generating and optimizing the control scheme, and converting the optimized scheme into executable instructions to be issued to the on-site construction equipment; S7, collecting the result data after the instructions are executed, and feeding the result data back to update the digital twin model and the hybrid prediction model to form a closed-loop control.

[0007] Further, in step S1, acquiring geometric data includes: performing three-dimensional laser scanning on each prefabricated segment to obtain point cloud data, denoising, simplifying and extracting features from the point cloud data, calculating the geometric parameters of the segment end face and the assembly reference surface, matching them with the design model, and generating a three-dimensional error field to guide assembly.

[0008] Furthermore, in step S1, the construction behavior data includes event-driven process sequence, tensioning sequence and loss value, hanging basket advancement and locking status, and maintenance system; the engineering rule data includes control point target alignment, pre-camber value, deviation threshold and risk level set according to specifications.

[0009] Furthermore, in step S2, the initial digital twin model couples the segmental geometric model, the structural mechanics finite element model considering time-varying effects, and the executable logic encoding the construction behavior data and engineering rule data under a unified spatiotemporal reference.

[0010] Furthermore, in step S3, the data collected by the multi-source sensor network includes: geometric displacement data obtained by total station, displacement gauge, and inclinometer; mechanical data obtained by strain gauge, cable force gauge, and jack pressure sensor; environmental data obtained by temperature and humidity sensor and anemometer; and construction management data reflecting the status of tensioning equipment and the age of concrete.

[0011] Furthermore, step S3 also includes: preprocessing the collected real-time data by cleaning, spatiotemporal alignment, anomaly identification, and credibility assessment; the data assimilation algorithm uses extended Kalman filtering, unscented Kalman filtering, particle filtering, or Bayesian update methods to perform online calibration of material parameters, boundary stiffness, and prestress loss.

[0012] Further, in step S4, the multi-scale hybrid prediction model includes:

[0013] Embedded entity models are used for rapid incremental analysis driven by process events;

[0014] A high-fidelity physical simulation sub-model is used for detailed time history calculations for each construction stage;

[0015] A data-driven sub-model is used to learn and correct the historical prediction residuals of the physical simulation sub-model.

[0016] The hybrid prediction model fuses the outputs of the three types of models using a weighted or Bayesian approach to obtain the final linear prediction value with confidence intervals.

[0017] Furthermore, the data-driven sub-model employs a long short-term memory network, a gradient boosting tree, or a Gaussian process regression algorithm. Its input features include concrete age, structural temperature gradient, measured prestressing tension values, historical alignment deviations, and construction equipment status data.

[0018] Furthermore, in step S6, the intelligent diagnosis identifies the main cause of deviation based on sensitivity analysis or variable importance assessment; the automatically generated control scheme includes one or more combinations of adjusting the subsequent prestressing tension force or sequence, adjusting the temporary support elevation, controlling the thickness of the inter-segment cementing layer, implementing local temperature control, and correcting the hanging basket forward movement parameters; after the generated scheme is evaluated by a multi-objective optimization algorithm, the optimal scheme is converted into operation instructions that can directly control the tensioning system, jacks, or hanging basket.

[0019] Furthermore, in step S7, the returned data is used to update the geometric calibration parameters and mechanical parameters of the digital twin model, retrain the data-driven sub-model, and precipitate the entire process data of diagnosis, decision-making and execution of this regulation into a reusable knowledge base.

[0020] The beneficial effects of this invention are:

[0021] 1. This invention acquires and integrates multi-dimensional data on the geometry, physics, behavior, and rules of precast beam segments to construct an initial digital twin model, solving the problem of the disconnect between the model and the physical data in traditional methods, and laying a reliable data and model foundation for high-precision analysis and control throughout the entire process;

[0022] 2. This invention achieves real-time synchronization and "growth" of the model and the physical bridge state by deploying a multi-source sensor network during the construction period and using a data assimilation algorithm to dynamically calibrate the key parameters of the digital twin model online, which significantly improves the model's realism and reliability.

[0023] 3. This invention constructs a multi-scale hybrid prediction model that integrates embedded entity model, high-fidelity physical simulation model and data-driven correction model, and outputs prediction results with uncertainty range. It effectively combines the accuracy of mechanism and the adaptability of data, and greatly improves the accuracy and robustness of the linear prediction of the entire construction process.

[0024] 4. This invention determines deviations in real time by setting adaptive thresholds and performs intelligent diagnosis of deviations exceeding limits based on methods such as sensitivity analysis to quickly locate the root cause. This changes the traditional lagging mode that relies on human experience and post-event analysis, and achieves rapid and accurate identification of the cause of the problem.

[0025] 5. This invention automatically generates multiple control schemes and uses digital twin models for simulation playback and multi-objective optimization. It can automatically generate the optimal control strategy that corrects the alignment while taking into account structural safety, construction efficiency and cost, thus realizing the transformation from experience-based decision-making to intelligent and optimal decision-making.

[0026] 6. This invention transforms the optimized control scheme into specific executable instructions for equipment and directly issues them to on-site equipment such as tensioning and jacking for execution, establishing a fast and precise connection link from "decision-making" to "execution," reducing human intervention errors, and improving control efficiency and execution.

[0027] 7. This invention drives model parameter updates and algorithm self-optimization by feeding back the results of regulation execution and newly collected data, and accumulates them into a reusable knowledge base, forming a complete "perception-analysis-decision-execution-learning" closed loop. This enables the system to have the ability to continuously evolve and accumulate knowledge, and can serve subsequent engineering projects.

[0028] 8. This invention achieves cross-device, cross-stage, and cross-professional data integration and collaboration through unified data coordinates, time benchmarks, and a fusion visualization platform, thereby improving the transparency and collaborative efficiency of construction management and providing a systematic solution for the digital and intelligent management and control of bridge construction. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart of the method of the present invention;

[0031] Figure 2 This is a schematic diagram of the digital twin architecture and information flow of the present invention;

[0032] Figure 3 This is a schematic diagram of data acquisition and processing according to the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins, as described in Embodiment 1 of this invention, is as follows: Figure 1 , Figure 2 and Figure 3 As shown, its detailed implementation is as follows:

[0035] Step S1, Data Acquisition:

[0036] S11. Segmental Geometric Data Acquisition and Processing: 3D point cloud scanning is performed on each prefabricated segment. The point cloud data is denoised, simplified, and feature lines are extracted. Geometric parameters of segment end faces, assembly reference surfaces, reserved holes, and positioning tenons are calculated and matched with the design geometric model. This matching process can be implemented using an Iterative Closest Point (ICP) algorithm. The output 3D error field between the segment end face and the reference surface will be directly used for error compensation during subsequent assembly. BIM parametric modeling and lightweight processing are completed under a unified coordinate system.

[0037] S12. Physical Properties and Boundary Data Acquisition: Collect material strength and elastic modulus evolution parameters, creep and shrinkage parameters, temperature and heat transfer parameters, prestressing system parameters, design parameters of temporary supports and hanging baskets, as well as load and construction equipment capacity data for each construction stage.

[0038] S13. Behavioral data modeling: sort out the procedures and timestamps of the entire construction process, define event-driven rules such as tensioning sequence and theoretical loss value, hanging basket advancement and locking conditions, curing and temperature control system, and adhesive layer construction technology, and form a process model that can be parsed and executed by computer.

[0039] S14. Rule Data Setting: Based on relevant specifications and project control objectives, set the layout of linear control points, target linear shape and pre-camber values ​​for each stage, deviation threshold and risk level, and unify the semantic definition and time base of all data.

[0040] Step S2, establish the initial digital twin model:

[0041] S21. Combining the geometric model obtained in step S11, the physical properties obtained in step S12, and the boundary conditions, a structural mechanics model considering time-varying effects such as temperature, creep, and shrinkage is established. This structural mechanics model is preferably established using the finite element method, and the prestressing is simulated using a staged activation method, considering prestress loss. This model includes segmental elements, temporary supports, formwork, the prestressing system, and construction loads.

[0042] S22. Behavior and Rule Twin: Write the process events, equipment capabilities and constraints defined in step S13, as well as the control indicators and acceptance rules set in step S14, into the digital twin in the form of executable logic (such as scripts or state machines).

[0043] S23. Unification of coordinates and data bus: Under a unified coordinate and time system, the coupling and visualization integration of the three elements of "geometric model - mechanical model - process rules" are realized to form an initial digital twin model.

[0044] Step S3, Data Acquisition and Dynamic Model Calibration during Construction:

[0045] S31. Deploy a multi-source sensor network: Deploy a sensor network to collect geometric data (such as displacement and tilt angle measured by total station / laser ranging / RTK), mechanical data (such as strain, cable force, jack pressure and displacement), environmental data (such as temperature, humidity, wind speed and sunshine), and management data (such as equipment status, tensioning records, concrete age).

[0046] S32. Data preprocessing: Perform cleaning, spatiotemporal alignment, anomaly detection, imputation, and credibility rating on the collected multi-source data;

[0047] S33. Data Assimilation and Model Growth: Employing one or more algorithms from Extended Kalman Filtering (EKF), Unscented Kalman Filtering (UKF), Particle Filtering, or Bayesian parameter update, and using measured data as observations, dynamic online calibration is performed on key state variables in the digital twin model, such as material parameters, boundary stiffness, prestress loss, and temperature load. Specifically, the data assimilation algorithm can control point displacement and cable force as the main observations, estimating state variables such as material elastic modulus, temporary support stiffness, and prestress loss coefficient, ensuring that the state of the digital twin model remains consistent with the physical bridge and continues to "grow."

[0048] Step S4: Construct a multi-scale hybrid prediction model and output the prediction:

[0049] S41, Embedded Entity Model: Incremental analysis is triggered by process events (such as "tensioning completed" or "hanging basket moved forward") to quickly obtain stage-by-stage linear estimation;

[0050] S42, Physical Simulation Sub-model: Performs high-fidelity time history calculations for each construction stage, and outputs control point alignment, stress, and safety indicators;

[0051] S43. Data-Driven Sub-Model: This sub-model learns and corrects the historical prediction residuals of the physical simulation sub-model, thereby refining the prediction results of the physical simulation model. Algorithms such as Long Short-Term Memory (LSTM), Support Vector Regression (SVR), Gradient Boosting (e.g., XGBoost), or Gaussian Process Regression (GPR) can be used. Input features include age, temperature difference, measured tension values, previous time-period deviations, and equipment status. This sub-model corrects the prediction results of the physical simulation model by learning the patterns of the physical simulation residuals.

[0052] S44. Model Fusion and Uncertainty Assessment: The prediction results of the above-mentioned embedded entity model, physical simulation sub-model and data-driven sub-model are fused by weighted averaging or Bayesian fusion to output the final linear prediction value of the whole process and its uncertainty (confidence) interval.

[0053] Step S5: Set the threshold and determine the deviation:

[0054] S51. Threshold setting and level classification: Based on the standard limits, target alignment and the risk level of the current construction stage, adaptively set the early warning and alarm thresholds for alignment deviation;

[0055] S52. Deviation Judgment and Spatial Distribution Identification: The difference between the predicted alignment and the target alignment is classified and judged, the level of exceeding the limit is given, and the distribution of the exceeding area in the space such as the key span, key section and closure section is identified, thereby locating the key weak parts in the alignment control.

[0056] Step S6: Intelligent diagnosis, generation, and issuance of control instructions:

[0057] S61. Diagnosis of the main causes of deviation: Identify the dominant factors leading to linear deviation based on methods such as sensitivity analysis, variable importance evaluation, or causal relationship diagrams.

[0058] S62. Control scheme generation: Automatically generate multiple candidate control schemes, which may include one or a combination of the following: modifying the prestressing tension and sequence, adjusting the temporary support elevation and shim thickness, controlling the forward movement of the hanging basket and the locking time, adjusting the segment adhesive layer thickness, implementing local heat preservation or cooling, increasing or releasing the reaction force, and modifying the pre-camber parameters.

[0059] S63. Scheme Evaluation and Optimization: Simulation playback and multi-objective optimization are performed on each candidate control scheme in a digital twin. The optimization focuses on minimizing linear deviation, while also satisfying multiple safety constraints such as stress, crack width, and deflection, and taking into account construction period and cost factors. Stress, crack, and deflection safety constraints are applied simultaneously, and construction period and cost are considered. Optimization algorithms may include multi-objective genetic algorithms (NSGA-II), particle swarm optimization, or simulated annealing algorithms.

[0060] S64. Command issuance and human-machine collaboration: The optimized solution is converted into operational instructions and work orders that can be executed by the equipment, and issued to the tensioning system, jacks and hanging basket control devices, and executed after confirmation by the construction personnel;

[0061] Step S7, Data Feedback and Loop Optimization:

[0062] S71. Execution result collection and comparative evaluation: Collect measured data such as alignment and cable force after the execution of control commands, compare them with the prediction results, and record the deviation convergence trajectory and uncertainty changes;

[0063] S72. Model and Knowledge Updates: Update geometric calibration values ​​and model parameters using the reflux data, retrain the data-driven sub-model, and record the model's growth progress. Accumulate the decision-making and effect data of this regulatory process into the project knowledge base.

[0064] S73. Report Generation and Knowledge Reuse: Generate traceable linear control analysis reports. The accumulated knowledge can be used to guide the construction of subsequent segments and provide case references and model reuse for similar projects.

[0065] Example 2: Key Technical Details of Hybrid Prediction and Data Assimilation

[0066] This embodiment further explains some of the key technical means involved in steps S4 (hybrid prediction) and S33 (data assimilation) in Embodiment 1, as a supplement to the technical details of the method.

[0067] (1) Supplement to the physical simulation sub-model of S42: When establishing a high-fidelity physical simulation model, the box girder structure can be simulated by coupling beam elements and shell elements. To more accurately simulate the stress behavior of the box girder, the physical simulation model can be modeled by coupling beam elements and shell elements. To accurately simulate the time-varying effects of concrete, the time-varying modulus method is used to calculate the effects of shrinkage and creep. For the contact problems between segments and the contact problems of temporary supports, nonlinear contact elements can be used.

[0068] (2) Supplement to the data-driven sub-model of S43: This sub-model can use the physical simulation residuals of the previous time period as training labels. Input features include concrete age, temperature difference of structural section, measured values ​​of prestressing tension, equipment status data, and linear deviation of the previous time period. Gaussian process regression (GPR) is one of the optional algorithms, which can simultaneously output the predicted value of the residuals and their uncertainty.

[0069] (3) Supplement to S33 regarding data assimilation: When performing data assimilation calibration, the displacement of control points and key cable forces can be used as the main observations. The state variables to be estimated may include the elastic modulus of the material, the support stiffness of temporary supports, the self-weight eccentricity of the hanging basket, and the prestress loss coefficient, etc. Through a continuous assimilation process, the digital twin is made consistent with the actual bridge, realizing the growth of the model. To improve the robustness of assimilation, the system can automatically reduce the weight of low-quality observation data according to the data credibility rating given in step S32, thereby ensuring the stability of the calibration process.

[0070] Data assimilation uses control point displacement and cable force as observations to estimate state variables such as elastic modulus, temporary support stiffness, hanging basket self-weight eccentricity, and loss coefficient; when the observation quality rating is low, the weight is automatically reduced to ensure robustness.

[0071] Example 3: Core Logic of Intelligent Diagnosis and Control Strategy Generation

[0072] This embodiment further explains the core analysis logic and output format involved in step S6 (intelligent diagnosis and control) in embodiment 1.

[0073] (1) Supplement to the intelligent diagnosis of S61: A combination of global sensitivity analysis and feature importance analysis can be used to systematically identify the key sensitive parameters and main influencing factors affecting the linearity, and an intuitive cause-and-effect analysis diagram can be formed accordingly.

[0074] (2) Supplement to the optimization of schemes and command generation for S63-S64: Multiple automatically generated control schemes can be simulated and verified in a digital twin. The objective function of the optimization process can be set as minimizing the linear deviation, and must meet multiple constraints such as stress, crack, deflection safety, and construction feasibility. Multi-objective genetic algorithms (such as NSGA-II) can be used to select the optimal solution set from numerous schemes. The final output executable commands can specifically include: the target load curve and control steps of the tensioning equipment, the displacement commands and locking sequence of each cylinder of the hanging basket, the thickness specifications and installation number of the temporary support pads, the control range and construction location of the adhesive layer thickness, and the specific parameters of the temperature control measures, etc. The execution of all commands must be recorded and confirmed to form a complete traceable link.

[0075] Output executable instructions, including: the target load curve and control steps of the tensioning equipment, the displacement and locking sequence of the hanging basket cylinder, the thickness and number of the temporary support pads, the control range and position of the adhesive layer thickness, and the parameters of the temperature control measures, and record the execution and signature to form a traceable link.

[0076] Example 4: A deployment method for system hardware and software architecture

[0077] This embodiment describes a possible hardware and software deployment configuration for implementing the above method, corresponding to the system and apparatus mentioned in the invention description of the specification.

[0078] Logically, the system can be divided into edge and cloud components. The edge component, deployed at the construction site, is responsible for connecting sensors and other devices, collecting geometric, mechanical, environmental, and management data in real time, and preprocessing this multi-source data through cleaning, spatiotemporal alignment, and anomaly identification. It also runs fast data assimilation algorithms (such as EKF / UKF) to dynamically calibrate model parameters; runs embedded entity models for rapid linear estimation triggered by specific processes; and receives control commands from the cloud and drives intelligent tensioning equipment, jacks, and other field devices to execute them. The cloud platform is responsible for running high-fidelity digital twin models and physical simulation calculations, training and correcting data-driven sub-models, performing multi-scale prediction model fusion and uncertainty assessment, realizing intelligent deviation diagnosis based on sensitivity analysis and solving multi-objective optimization algorithms, and managing and maintaining a knowledge base containing cases and parameters.

[0079] The hardware components may include: a perception layer (such as point cloud scanners, total stations, displacement gauges, strain gauges, temperature sensors, etc.), a network and edge computing layer (data acquisition gateways, edge servers), an execution layer (intelligent tensioning equipment, programmable hydraulic jacks), and a cloud center (high-performance computing servers, data storage arrays). Various software functional modules (such as data acquisition, model building, dynamic calibration, hybrid prediction, diagnostic control, etc.) interact and integrate through a unified data bus and application programming interface (API).

[0080] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any modifications to the technical solutions described in the foregoing embodiments, or equivalent substitutions of some or all of the technical features thereof, within the spirit and principles of the present invention, do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting and controlling the alignment of cantilever bridges constructed using the short-line method based on digital twins, characterized in that, Includes the following steps: S1. Acquire geometric data, physical property data, construction behavior data, and engineering rule data of the precast beam segments using the short-line method; S2. Based on the data, establish an initial digital twin model that integrates the geometric model, structural mechanics model, and construction procedure rules; S3. Deploy a multi-source sensor network during construction to collect real-time data, and use a data assimilation algorithm to dynamically calibrate the parameters of the initial digital twin model to achieve synchronous evolution between the model and the actual bridge; S4. Construct and run a multi-scale hybrid prediction model, and output the predicted value and uncertainty range of the alignment throughout the construction process based on the digital twin model; S5. Determine whether the predicted value has an out-of-limit deviation based on the target alignment and adaptive threshold; S6. When an out-of-limit deviation exists, perform intelligent diagnosis to identify the cause of the deviation, automatically generate and optimize the control scheme, and convert the optimized scheme into executable instructions to be issued to the on-site construction equipment; S7. Collect the result data after the instruction is executed, and feed the result data back to update the digital twin model and the hybrid prediction model to form a closed-loop control.

2. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 1, characterized in that, In step S1, acquiring geometric data includes: performing three-dimensional laser scanning on each prefabricated segment to obtain point cloud data, denoising, simplifying and extracting features from the point cloud data, calculating the geometric parameters of the segment end face and the assembly reference surface, matching them with the design model, and generating a three-dimensional error field to guide assembly.

3. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 1 or 2, characterized in that, In step S1, the construction behavior data includes event-driven process sequence, tensioning sequence and loss value, hanging basket advancement and locking status, and maintenance system; the engineering rule data includes control point target alignment, pre-camber value, deviation threshold and risk level set according to specifications.

4. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 1 or 2, characterized in that, In step S2, the initial digital twin model couples the segmental geometric model, the structural mechanics finite element model considering time-varying effects, and the executable logic encoding the construction behavior data and engineering rule data under a unified spatiotemporal reference.

5. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 1 or 2, characterized in that, In step S3, the data collected by the multi-source sensor network includes: geometric displacement data obtained by total station, displacement gauge, and inclinometer; mechanical data obtained by strain gauge, cable force gauge, and jack pressure sensor; environmental data obtained by temperature and humidity sensor and anemometer; and construction management data reflecting the status of tensioning equipment and the age of concrete.

6. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 1 or 2, characterized in that, Step S3 further includes: preprocessing the collected real-time data by cleaning, spatiotemporal alignment, anomaly identification, and credibility assessment; the data assimilation algorithm uses extended Kalman filtering, unscented Kalman filtering, particle filtering, or Bayesian update methods to perform online calibration of material parameters, boundary stiffness, and prestress loss.

7. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 1 or 2, characterized in that, In step S4, the multi-scale hybrid prediction model includes: Embedded entity models are used for rapid incremental analysis driven by process events; A high-fidelity physical simulation sub-model is used for detailed time history calculations for each construction stage; A data-driven sub-model is used to learn and correct the historical prediction residuals of the physical simulation sub-model. The hybrid prediction model fuses the outputs of the three types of models using a weighted or Bayesian approach to obtain the final linear prediction value with confidence intervals.

8. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 7, characterized in that, The data-driven sub-model employs a long short-term memory network, gradient boosting tree, or Gaussian process regression algorithm. Its input features include concrete age, structural temperature gradient, measured prestressing tension values, historical alignment deviations, and construction equipment status data.

9. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 1, 2, or 8, characterized in that, In step S6, the intelligent diagnosis identifies the main cause of deviation based on sensitivity analysis or variable importance assessment; the automatically generated control scheme includes one or more combinations of adjusting the subsequent prestressing tension or sequence, adjusting the temporary support elevation, controlling the thickness of the inter-segment cementing layer, implementing local temperature control, and correcting the hanging basket forward movement parameters; after the generated scheme is evaluated by a multi-objective optimization algorithm, the optimal scheme is converted into operation instructions that can directly control the tensioning system, jacks, or hanging basket.

10. The method for predicting and controlling the alignment of cantilever bridges using the short-line method based on digital twins according to claim 1, 2, or 8, characterized in that, In step S7, the returned data is used to update the geometric calibration parameters and mechanical parameters of the digital twin model, then train the data-driven sub-model, and the entire process data of diagnosis, decision-making and execution of this regulation is precipitated into a reusable knowledge base.

Images