Method and system for establishing a digital twin model of steel structure construction based on BIM

Abstract

The application relates to the technical field of building construction informatization, and provides a steel structure construction digital twin model establishment method and system based on BIM, which comprises the following steps: acquiring a design model containing component geometric information and an identity mark and a manufacturing data package containing the identity mark, geometric deviation and material attributes; establishing a data mapping relationship between the design model and the manufacturing data package based on the identity mark; constructing an initial geometric state of the digital twin model based on the design model, correcting the initial geometric state and updating model parameters of the digital twin model according to the manufacturing data package, and generating a digital twin model fused with manufacturing end data. The application solves the problem of initial state distortion of the model by pre-injecting real manufacturing end data into the model initialization link, and improves the initial accuracy and dynamic mapping accuracy of the digital twin model.

Description

Technical Field

[0001] This application relates to the technical field of information technology and digital twin data processing, specifically to a method and system for establishing a BIM-based digital twin model for steel structure construction. Background Technology

[0002] Currently, in the field of steel structure construction, digital twin technology has been applied to the mechanical performance evaluation and safety management of the construction process. Existing technologies typically establish a digital twin model that integrates multi-source information and links it with on-site sensor networks to acquire monitoring data such as structural deformation and environmental parameters, thereby achieving digital representation and model correction of the physical entity's state.

[0003] Existing digital twin model construction methods mostly focus on data feedback during the construction phase, failing to effectively connect the data flow between design and manufacturing. Specifically, there is a lack of effective data flow channels between the structured information such as geometric topology and material properties in the BIM model at the design end and the actual processing deviations and component identification data at the manufacturing end. This results in inherent information deficiencies in the twin model during initialization, often requiring a long period of sensor data training to converge to the true state. This severely affects the digital twin model's ability to accurately map physical entities and the timeliness of early warning.

[0004] Therefore, there is an urgent need for a method and system for establishing digital twin models of steel structure construction based on BIM to solve the above-mentioned technical problems. Summary of the Invention

[0005] This application provides a method and system for establishing a digital twin model of steel structure construction based on BIM, which solves the above-mentioned technical problems.

[0006] To achieve the above objectives, this application employs the following technical solution: Firstly, this application provides a method for establishing a BIM-based digital twin model for steel structure construction, including: Obtain the design model of the steel structure, which is built based on BIM and includes the geometric information and identification of the components; Obtain the manufacturing data package of the component, the manufacturing data package containing the component's identification, geometric deviations, and material properties; Based on their identities, the geometric deviations and material properties in the manufacturing data package are mapped to the geometric information and material parameters of the corresponding components in the design model, establishing a data mapping relationship between the measured data from the manufacturing end and the design model; the initial geometric state of the digital twin model is constructed based on the design model; and based on the data mapping relationship, the initial geometric state is corrected and the model parameters of the digital twin model are updated according to the manufacturing data package, generating a digital twin model that integrates the manufacturing end data. The process of acquiring the manufacturing data package includes: parsing the parametric BIM model, extracting the component's geometric parameters and process constraints, generating processing instructions to drive the processing equipment to execute component processing; after the component processing is completed, collecting the component's actual geometric deviation data and material property data; binding the actual geometric deviation data, the material property data, and the component's identity identifier to form the manufacturing data package.

[0007] As one implementation method, updating the model parameters of the digital twin model includes: The geometric deviations of the component are superimposed onto the initial geometric state of the digital twin model to correct the initial geometric state of the component; The material properties are updated in the material constitutive relation definition of the digital twin model.

[0008] As one implementation, the method further includes: During construction, a set of response parameters of the steel structure is collected, and the set of response parameters is semantically fused with the design model and the manufacturing data package under a unified spatiotemporal reference. Based on the fusion result, the parameters of the digital twin model are updated and the state is corrected.

[0009] As one implementation method, updating the parameters and correcting the state of the digital twin model based on the fusion results includes: Extract structural deformation time series data from the response parameter set, and calculate the deviation sequence between the measured structural deformation value and the simulated deformation value of the digital twin model at the corresponding time. Statistical analysis is performed on the deviation sequence, and when the statistical characteristics of the deviation sequence exceed a preset tolerance threshold, a model parameter correction command is triggered. The material property parameters, boundary condition parameters, or component geometric parameters in the digital twin model are adjusted in reverse using an optimization algorithm, so that the deviation between the simulated response value output by the adjusted digital twin model and the measured value of the structural deformation meets the convergence criterion.

[0010] As one implementation method, the statistical analysis of the deviation sequence is based on the dynamic data fusion deviation index, which is calculated by weighting the weight coefficients of each monitoring point, the measured deviation values, and the time synchronization attenuation coefficients. The objective function of the optimization algorithm is constructed based on the deviation between the measured response values ​​and the simulated response values ​​of each monitoring point, the regularization constraints of the parameter space domain, and the parameter constraint terms based on the manufacturing end data.

[0011] As one implementation, the method further includes: Construction risk prediction and decision feedback are performed based on the corrected digital twin model, including: The mechanical state output of the corrected digital twin model and the real-time collected structural deformation time series data are input into the construction anomaly identification model to identify anomalies where the response mode deviation exceeds the threshold. The abnormal working condition characteristics and environmental parameters of the identified abnormal working conditions are input into the mechanical performance risk prediction model to predict the probability and location information of structural damage or instability during a specified construction stage in the future. The abnormal working condition characteristics, the risk probability, and the implementation location information are input into the construction process optimization decision model to generate construction adjustment instructions.

[0012] Secondly, this application provides a BIM-based digital twin model creation system for steel structure construction, including: The design model acquisition module is used to acquire the design model of the steel structure. The design model is built based on BIM and contains the geometric information and identification of the components. A data packet acquisition module is used to acquire the manufacturing data packet of the component, wherein the manufacturing data packet contains the component's identification, geometric deviation, and material properties; wherein the process of acquiring the manufacturing data packet includes: parsing the parametric BIM model, extracting the component's geometric parameters and process constraints, generating processing instructions to drive the processing equipment to execute component processing; after the component processing is completed, collecting the component's actual geometric deviation data and material property data; binding the actual geometric deviation data, the material property data, and the component's identification to form the manufacturing data packet; The model building module is used to map the geometric deviations and material properties in the manufacturing data package to the geometric information and material parameters of the corresponding components in the design model based on their identities, thereby establishing a data mapping relationship between the measured data from the manufacturing end and the design model; constructing the initial geometric state of the digital twin model based on the design model; and correcting the initial geometric state and updating the model parameters of the digital twin model according to the manufacturing data package based on the data mapping relationship, thereby generating a digital twin model that integrates the manufacturing end data.

[0013] Thirdly, this application provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method as described in any one of the first aspects.

[0014] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any one of the first aspects.

[0015] This application provides a method and system for establishing a BIM-based digital twin model for steel structure construction. By mapping a manufacturing data package containing component identification, geometric deviations, and material properties to the design model, and correcting the initial geometric state and updating model parameters based on the manufacturing data package during the digital twin model initialization phase, it breaks down the data barriers between the design and manufacturing ends in traditional technologies. Its principle lies in directly compensating for the idealized assumptions of the design model using measured data from the manufacturing end, enabling the twin model to integrate manufacturing data from its inception. This solves the problem of initial model distortion caused by the lack of real manufacturing data in existing technologies, significantly improving the initial accuracy of the digital twin model and shortening the sensor data acclimatization time required for the model to converge to its true state. By collecting response parameter sets during construction and semantically fusing them with the design model and manufacturing data package under a unified spatiotemporal reference, the digital twin model's parameters are updated and its state corrected based on the fusion results. This mechanism solves the data silo problem through source alignment and closed-loop correction of multi-source heterogeneous data, enabling the digital twin model to continuously track changes in the physical construction site's state. In particular, during the model calibration process, by constructing an optimization objective function that includes constraints based on manufacturing data parameters, it is ensured that parameter adjustments do not deviate from the manufacturing reality base, avoiding model distortion caused by blind calibration, thereby improving the timeliness and accuracy of construction mechanical performance evaluation. Attached Figure Description

[0016] The present application will be further described below with reference to the accompanying drawings and embodiments.

[0017] Figure 1 The diagram illustrates a flowchart of a BIM-based digital twin model creation method for steel structure construction provided in this application.

[0018] Figure 2 The diagram shows a module schematic of a BIM-based digital twin model creation system for steel structure construction provided in this application. Detailed Implementation

[0019] To facilitate understanding of this application, a more complete description will be provided below with reference to the accompanying drawings. Preferred embodiments of this application are shown in the drawings. However, this application can be implemented in many other different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of this application.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0021] The technical solutions of the embodiments of this application and how the technical solutions of the embodiments of this application solve the above-mentioned technical problems will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the various embodiments or technical features described below can be arbitrarily combined to form new embodiments, and the same or similar concepts or processes may not be described again in some embodiments. Obviously, the described embodiments are some embodiments of the embodiments of this application, but not all embodiments.

[0022] Existing digital twin model construction methods mostly focus on data feedback during the construction phase. For example, Chinese invention patent application CN120974791A discloses a method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins. This method establishes a digital twin model that integrates structural design information, construction schedule, and material physical properties. A multi-dimensional sensor network is deployed at the construction site to collect response parameters in real time. Dynamic data fusion algorithms and anomaly identification algorithms are used to update and correct the parameters of the digital twin model. An intelligent analysis model is then invoked to identify construction anomalies, predict risks, and make process optimization decisions, thus constructing a data-driven closed-loop management mechanism for construction. However, the data source for this approach is limited to the sensor network at the construction site. The initialization of the digital twin model is directly based on design information, without involving the acquisition and feedback of actual manufacturing deviations and material property data. This results in an inherent information gap in the model during initialization due to the difference between idealized design assumptions and actual manufacturing errors.

[0023] For example, Chinese invention patent application CN121809280A discloses a method for dynamic updating and optimization of a digital twin model of a tall steel structure construction process. This method involves deploying multi-source data acquisition equipment at the construction site, reconstructing an initial digital twin model based on point cloud scanning and BIM technology, and completing multi-dimensional data deviation analysis through a three-step process of deviation classification, impact scope definition, and correction priority ranking. This triggers hierarchical model updates and generates model parameter correction schemes and construction process adjustment parameters using a construction process adaptation algorithm. However, in this scheme, point cloud scanning and deviation calibration are both performed at the construction site, and the data source is still limited to on-site measurements during the construction phase. It does not involve data integration from the design to the manufacturing end, and the actual processing deviations and material property data from the manufacturing end are not pre-injected into the initialization stage of the digital twin model. Furthermore, its model update mechanism only corrects parameters based on on-site measured data, without establishing a constraint mechanism for the model correction process during the construction phase using manufacturing end data, posing a risk that parameter correction may deviate from manufacturing accuracy.

[0024] Furthermore, based on the model initialization and integration of manufacturing data, it is also necessary to continuously constrain the adjustment range of model parameters using manufacturing data during the construction calibration process, so as to avoid the parameter calibration deviating from the actual manufacturing state of the physical entity and causing the calibrated model to become distorted.

[0025] This application presents a method and system for establishing digital twin models of steel structure construction based on BIM. This method injects actual geometric deviations and material properties from the manufacturing process into the initialization phase of the digital twin model, significantly improving the initial accuracy and dynamic mapping accuracy of the model. The method will be described first, followed by a description of the system.

[0026] Example 1: This embodiment provides a method for establishing a BIM-based digital twin model for steel structure construction. This method corrects the initial state of the digital twin model by introducing real data from the manufacturing process before construction, thereby solving the problem of discrepancies between the initial state of the model and the physical entity in existing technologies.

[0027] Specifically, such as Figure 1 As shown, the method includes the following steps: Step S101: Obtain the design model of the steel structure. The design model is built based on BIM and includes the geometric information and identification of the components.

[0028] The design model is typically a parametric model built using BIM software during the design phase, containing information such as the theoretical geometric dimensions, spatial location, and topological relationships of components. The identifier is a unique code assigned to each component, such as a GlobalID based on the IFC standard or a component number generated according to project rules. This identifier is used to uniquely identify components throughout their entire lifecycle and is the foundation for linking the design model with subsequent manufacturing data. It should be understood that the design model contains not only geometric information but also semantic information such as material design grades and connection details.

[0029] Step S102: Obtain the manufacturing data package of the component, which includes the component's identification, geometric deviations, and material properties.

[0030] The manufacturing data package is a digital record generated after the steel structure component is manufactured in the factory. It reflects the actual physical state of the component at the time of manufacture, rather than the ideal state during the design phase. Geometric deviations refer to the differences between the actual geometric parameters of the component after manufacturing and the theoretical geometric parameters in the design model. Examples include length deviations, cross-sectional dimension deviations, hole offsets, and weld shrinkage obtained through 3D laser scanning or total station measurements. Material properties refer to the actual mechanical performance parameters of the component obtained through material re-inspection or non-destructive testing, such as measured yield strength, tensile strength, elastic modulus, and elongation. These parameters often differ from the standard values ​​in the design specifications due to factors such as smelting batches and rolling processes. The manufacturing data package also includes the same identification identifier as in step S101 to ensure that the measured data can be accurately traced back to the corresponding component entity. This step, by collecting objective measured data from the manufacturing end, provides a true physical benchmark for subsequent correction of the digital twin model, avoiding the inherent errors caused by relying solely on theoretical design values.

[0031] Step S103: Based on the identity identifiers of the two, the geometric deviations and material properties in the manufacturing data package are mapped to the geometric information and material parameters of the corresponding components in the design model, thereby establishing a data mapping relationship between the measured data at the manufacturing end and the design model.

[0032] Specifically, since design models and manufacturing data packages typically originate from different software systems or data platforms (such as the BIM platform on the design side and the MES system on the manufacturing side), they often exhibit heterogeneity in data structure and storage format. This step utilizes identity identifiers as data anchors to match and map the two. For example, database mapping can be used, with the identity identifier as the primary key, to attach the geometric deviation and material attribute fields in the manufacturing data package to the attribute set of the corresponding component in the design model; or a separate mapping index table can be established to enable bidirectional queries.

[0033] The establishment of data mapping relationships essentially breaks down the data barriers between the design domain and the manufacturing domain, enabling the previously isolated theoretical design information and measured manufacturing information to achieve semantic alignment in a unified digital space. Specifically, through this data mapping relationship, step S104, when constructing a digital twin model, can accurately locate the geometric information and material parameters of each component in the design model, corresponding to the measured manufacturing data, based on the identification identifier. This allows for precise data retrieval on a component-by-component basis in subsequent geometric correction and parameter update operations.

[0034] Step S104: Construct the initial geometric state of the digital twin model based on the design model; and based on the data mapping relationship, correct the initial geometric state and update the model parameters of the digital twin model according to the manufacturing data package to generate a digital twin model that integrates manufacturing data.

[0035] Specifically, when constructing a digital twin model, an initial geometric state is generated based on the design model. This state represents the theoretical form of the component under ideal design conditions.

[0036] Based on the data mapping relationship established in step S103, the geometric deviations and material properties corresponding to each component are read from the manufacturing data package, and the initial geometric state is substantially corrected. The correction process includes two aspects: First, geometric correction, which involves superimposing or mapping the geometric deviations in the manufacturing data package onto the initial geometric state. For example, adjusting the coordinates of the finite element mesh nodes to reflect the actual length changes or hole offsets, so that the model geometry approximates the actual processed shape of the physical entity. Second, parameter updating, which involves replacing the standard material parameters in the design model with the measured material properties in the manufacturing data package, updating the model parameters of the digital twin model (such as the definition of material constitutive relations), so that the mechanical calculation basis of the model reflects the true material property discreteness.

[0037] The digital twin model generated after the above corrections is a digital twin model that integrates manufacturing data. In this embodiment, "integrating manufacturing data" means that at the moment of initialization (i.e., before the first process begins), the geometric shape and mechanical property parameters of the digital twin model already include the actual deviations and performance fluctuations generated in the manufacturing process, rather than merely a digital copy of the design drawings. In this way, this embodiment eliminates the initial model distortion caused by the difference between idealized design assumptions and actual manufacturing errors from the source, improving the mapping accuracy of the digital twin model in the early stages of application. This makes subsequent state assessments and data analyses based on the model more reliable, and also significantly shortens the cycle required for the model to converge to the true state, which relies on long-term monitoring data in traditional methods.

[0038] The technical advantage of this embodiment lies in the fact that existing technologies typically use the design model directly as the initial state of the digital twin model, or only perform calibration through sensor data during the construction phase. This results in deviations in the initial state of the model, requiring extensive monitoring data to converge to the true state. In contrast, this embodiment injects manufacturing data packages before the initialization phase, ensuring that the digital twin model is highly consistent with the physical entity from the outset. This eliminates initial model errors caused by processing errors and material property variations, significantly improving the mapping accuracy and reliability of the model in the early stages of construction.

[0039] In one embodiment, correcting the initial geometry and updating material property parameters based on the manufacturing data package includes: (Based on the data mapping relationship,) the geometric deviation of the component is superimposed on the initial geometric state of the digital twin model to correct the initial geometric state of the component; (Based on the data mapping relationship,) the material properties are updated in the material constitutive relation definition of the digital twin model.

[0040] The superposition of geometric deviations is not a simple numerical addition or subtraction, but a coordinate transformation based on the spatial topological relationship of the components.

[0041] For example, for steel beam components, the geometric deviations recorded in the manufacturing data package might include length deviations (e.g., +2mm), section height deviations (e.g., -0.5mm), and bolt hole position deviations (e.g., hole offset 1mm). In the finite element mesh of the digital twin model, these deviation values ​​are superimposed onto the corresponding mesh node coordinates. Specifically, length deviations are achieved by adjusting the relative distance between the nodes at both ends of the beam element, section deviations are achieved by correcting the section definition parameters, and hole position deviations are achieved by adjusting the local coordinates of the corresponding nodes on the connecting plate. Through this refined geometric correction, the initial geometric state of the digital twin model is no longer an idealized design form, but rather a physical entity that truly reflects the manufacturing errors, thereby eliminating the initial geometric errors caused by manufacturing tolerances.

[0042] Updating material properties directly impacts the core mechanical calculations of the digital twin model. Design models typically employ standard material parameters specified in regulations; for example, the standard yield strength of Q345 steel is 345 MPa. However, the material properties of actual components exhibit dispersion due to factors such as smelting batches and rolling processes. The material verification attributes included in the manufacturing data package provide actual measured values; for example, the measured yield strength of a certain batch of steel is 358 MPa, and the elastic modulus is 206 GPa. The system updates these measured values ​​into the material constitutive relation definition of the digital twin model, replacing the original standard values.

[0043] It should be understood that updating the constitutive relation of materials is not limited to yield strength and elastic modulus, but can also include parameters such as tensile strength, elongation, and impact toughness, depending on the computational needs of the model. By updating material properties, the digital twin model can use real material parameters in mechanical calculations, avoiding computational biases caused by the discreteness of material properties.

[0044] Therefore, by accurately locating the geometric information and material parameters of each component in the design model through data mapping relationships, the geometric deviations are directly superimposed on the initial geometric state and the material properties are updated to the material constitutive relation definition. This achieves component-by-component fine-grained correction of the digital twin model in both geometric shape and mechanical properties, avoiding model correction distortion caused by data misalignment, and enabling the model to have a mechanical behavior basis that is highly consistent with the physical entity from the initialization stage.

[0045] Furthermore, the process of acquiring and generating data packets is a prerequisite for implementing the above modifications. This process includes: The parametric BIM model is parsed to extract the geometric parameters and process constraints of the components, and processing instructions are generated to drive the processing equipment to execute the component processing. After the component processing is completed, the actual geometric deviation data and material property data of the component are collected. The actual geometric deviation data, the material property data and the component identification are bound together to form the manufacturing data package.

[0046] This process establishes a data chain from the design end to the manufacturing end. First, at the manufacturing end, a parametric BIM model from the design end is received via a standardized data interface. When parsing this model, the geometric parameters of the components (such as length, cross-sectional dimensions, and hole coordinates) and process constraints (such as weld bevel forms and cutting path planning) are extracted. This information is then transformed into specific processing instructions, such as the cutting path code for a CNC cutting machine and the welding trajectory parameters for a welding robot, thereby driving the processing equipment to perform automated component processing. This step ensures the accurate realization of design intent at the manufacturing end.

[0047] Secondly, after the components are manufactured, the manufacturing end collects actual geometric deviation data of the components using precision measuring equipment. For example, a 3D laser scanner is used to perform a full-range scan of the component to obtain point cloud data of the component surface. By comparing this data with the theoretical geometry of the design model, the geometric deviation values ​​of each part are calculated. Simultaneously, the manufacturing end acquires the material property data of the components, which typically comes from material inspection reports, recording the measured mechanical properties of that batch of steel. It should be understood that the methods for collecting geometric deviation data are not limited to 3D laser scanning; total station measurements, industrial photogrammetry, etc., can also be used. Similarly, the sources of material property data are not limited to inspection reports; they can also include online non-destructive testing data.

[0048] Finally, the collected actual geometric deviation data and material property data are bound to the component's identification identifier. The identifier plays a crucial indexing role here, associating the measured data from the manufacturing end with the unique component code in the design model, ensuring that data can be accurately transmitted back to the corresponding component. For example, the component's identification identifier can be a one-dimensional barcode, a QR code, or an RFID tag, read by a scanning device during processing and automatically associated during data collection. After binding, a complete manufacturing data package is formed, which is transmitted back to the digital twin platform through a data interface, providing a real data source for the model's initialization and correction.

[0049] The technical advantages of this embodiment are as follows: By finely superimposing geometric deviations and updating material properties through actual measurements, high-fidelity correction of the initial state of the digital twin model is achieved, making the model highly consistent with the physical entity in terms of geometric shape and mechanical properties. Simultaneously, by analyzing the parametric BIM model to drive processing, collecting measured data, and binding identity identifiers, a complete process is established between the design and manufacturing ends, ensuring the accuracy and traceability of manufacturing data packages and providing data support for the manufacturing authenticity of the digital twin model.

[0050] In one embodiment, during the steel structure construction process, as the construction progresses, the stress state, boundary conditions, and environmental factors of the structure are constantly changing. The digital twin model needs to perceive these changes in real time and make dynamic adjustments to maintain a high degree of consistency with the physical entity.

[0051] Specifically, the method in this embodiment further includes: during the construction process, collecting a set of response parameters of the steel structure, semantically fusing the set of response parameters with the design model and the manufacturing data package under a unified spatiotemporal reference, and updating the parameters and correcting the state of the digital twin model based on the fusion result.

[0052] As an example, the set of response parameters is acquired in real time through a multi-dimensional sensor monitoring network deployed at the construction site. This monitoring network includes displacement sensors (such as GNSS receivers and total station prisms) for capturing overall or local deformation of the structure, vibrating wire strain gauges or fiber optic grating sensors for monitoring the stress and strain state of key components, and thermocouples and anemometers for recording environmental load conditions. These sensors operate continuously at preset sampling frequencies, forming a comprehensive set of response parameters that includes time-series data on structural deformation, stress and strain data, environmental temperature fluctuation data, and wind load variation data. It should be understood that the selection, deployment location, and sampling frequency of sensors are not fixed but are configured specifically according to the specific form of the steel structure (such as large-span trusses and high-rise frames), key risk points at the construction stage, and the monitoring plan, as long as they can reflect the mechanical response of the structure and the environmental conditions during construction.

[0053] Because the design model, manufacturing data package, and on-site monitoring data originate from different business systems and physical spaces, there is significant data heterogeneity among them. For example, the design model is usually based on a global design coordinate system, the manufacturing data package may be based on a factory processing measurement coordinate system, and the on-site monitoring data is often based on an independent construction local coordinate system. In terms of data format, BIM models are mostly in IFC or RVT format, manufacturing data may be structured database records, and sensor data are time-series signals. Therefore, the "semantic fusion under a unified spatiotemporal reference" in this embodiment is not a simple data splicing, but includes three levels of in-depth processing operations: First, format conversion, that is, unifying various heterogeneous data into a standard internal data structure that can be recognized by the digital twin platform; second, coordinate unification, that is, accurately mapping the data under the factory measurement coordinate system and the on-site local coordinate system to the global design coordinate system through a pre-calibrated coordinate transformation matrix, eliminating spatial reference differences; and finally, semantic association, that is, using the component's identity identifier as the core index, combined with spatial location information, accurately matching specific monitoring data streams, manufacturing attribute data, and specific component entities in the design model. Through the above three-layer fusion operation, the alignment of multi-source heterogeneous data on the basis of homogeneity, homomorphism and synonymy is realized, providing high-quality data input for subsequent model correction.

[0054] Furthermore, based on the fusion results, the digital twin model undergoes parameter updates and state correction, including: Extract structural deformation time series data from the response parameter set, and calculate the deviation sequence between the measured structural deformation value and the simulated deformation value of the digital twin model at the corresponding time. Statistical analysis is performed on the deviation sequence, and when the statistical characteristics of the deviation sequence exceed a preset tolerance threshold, a model parameter correction command is triggered. The material property parameters, boundary condition parameters, or component geometric parameters in the digital twin model are adjusted in reverse using an optimization algorithm, so that the deviation between the simulated response value output by the adjusted digital twin model and the measured value of the structural deformation meets the convergence criterion.

[0055] In this process, the instantaneous deviation value at a single moment or a single measuring point is not relied upon as the basis for correction. Instead, the deviation sequence and its statistical characteristics are used as the judgment benchmark. Specifically, continuous deformation data of the monitoring points within a time window are extracted from the fused data stream and compared point by point with the simulated values ​​output synchronously by the model to form a deviation time series. Subsequently, statistical analysis is performed on this series to extract its mean, variance, trend term, and other characteristic quantities. The preset tolerance threshold is a dynamic or static threshold pre-set according to the engineering accuracy requirements, structural safety level, and relevant acceptance specifications.

[0056] A systematic deviation between the model and the entity is determined only when the statistical characteristics of the deviation sequence (such as mean deviation or variance dispersion) consistently exceed their corresponding thresholds, thereby triggering a correction command. This sequence statistics-based judgment mechanism can effectively filter out random errors caused by sensor noise, instantaneous environmental interference, or data transmission jitter, avoid frequent model oscillations due to false triggering, and ensure the robustness of the correction process.

[0057] When the calibration command is triggered, the built-in optimization algorithm is activated to adjust the parameters in reverse. It is important to emphasize that this adjustment process is incremental, not starting from zero. The optimization algorithm uses material properties (such as elastic modulus and yield strength), boundary conditions (such as support stiffness and semi-rigid coefficient of connection nodes), or component geometric parameters as variables to be optimized, and iteratively seeks the best within the parameter space. In each iteration, the algorithm re-runs the mechanical calculations of the digital twin model based on the current parameter combination, generates new simulated response values, and calculates their deviation from the measured values. The iterative process continues until the deviation meets a preset convergence criterion, such as the mean deviation approaching zero and the variance stabilizing within the allowable range. In this way, the digital twin model can continuously absorb on-site measured information during construction, dynamically correcting model prediction deviations caused by factors such as changes in construction loads, accumulated environmental effects, or the propagation of installation errors. This achieves an evolution from static initial high fidelity to dynamic process high fidelity, ensuring that the model always accurately maps the actual state of the physical construction site.

[0058] Regarding the specific types of the aforementioned optimization algorithms, this embodiment can employ the following numerical optimization methods: When the parameter dimension is low and the objective function is approximately quadratic, the Levenberg-Marquardt algorithm is preferred, as it combines the stability of gradient descent with the fast convergence of the Gauss-Newton method; when the parameter dimension is high or explicit constraints exist, the Sequential Quadratic Programming (SQP) algorithm is used, which approximates the global optimum by constructing and solving a quadratic approximate subproblem in each iteration; when the objective function exhibits strong nonlinearity or multimodal characteristics, the Particle Swarm Optimization (PSO) algorithm can be used for global search. In practical engineering, a coarse-fine two-stage strategy can also be adopted: first, PSO is used to quickly locate the approximate optimal region globally, and then Sequential Quadratic Programming is used for local refinement.

[0059] In a preferred embodiment, the statistical analysis of the deviation sequence is based on a dynamic data fusion deviation index, which is calculated by weighting the weight coefficients of each monitoring point, the measured deviation values, and the time synchronization attenuation coefficients. The objective function of the optimization algorithm is constructed based on the deviation between the measured response values ​​and the simulated response values ​​of each monitoring point, the regularization constraints of the parameter space domain, and the parameter constraint terms based on the manufacturing end data.

[0060] The dynamic data fusion deviation index is a comprehensive evaluation indicator used to quantify the overall deviation between the digital twin model and the physical entity. The weight coefficients of each monitoring point are determined based on its spatial importance and structural sensitivity. For example, for long-span steel structures, the weight coefficient of monitoring points at mid-span is usually higher than that of monitoring points near supports. The time synchronization attenuation coefficient is used to handle minor misalignments between monitoring and simulated data on the time axis. When the timestamps of measured and simulated data are not perfectly aligned, the attenuation coefficient reduces the weight of that data point in the fusion index, thereby improving the robustness of the algorithm. The dynamic data fusion deviation index, calculated through weighted averages, more accurately reflects the overall deviation level of the model and avoids interference from individual abnormal data points in the judgment.

[0061] The objective function of the optimization algorithm constructs a multi-objective optimization framework. Its core term is the deviation between the measured and simulated response values ​​at each monitoring point, aiming to minimize the difference between the model output and the measured data. Regularization constraints in the parameter space domain prevent overfitting or oscillations within the parameter space, ensuring the physical rationality of parameter adjustments. Parameter constraints based on manufacturing data utilize the manufacturing data package returned in Example 1 to limit the parameter adjustment range. For example, the adjustment range of the material's elastic modulus should be limited to a reasonable fluctuation range of the material's verified properties, and the adjustment range of the component's geometric parameters should be limited to the range of measured geometric deviations. This constraint ensures that the model calibration process does not deviate from manufacturing realism, realizing the continuous value extension of manufacturing data during the construction phase.

[0062] By applying the two models described above, problems such as asynchronous data interference, parameter overfitting, and deviation of correction results from the manufacturing reality base encountered during the calibration of digital twin models for steel structure construction are avoided. The specific composition and physical meaning of these two models are briefly explained below.

[0063] Regarding the calculation of the dynamic data fusion deviation index: This index is not a simple average error, but a multi-dimensional evaluation indicator that comprehensively considers spatial sensitivity, temporal synchronization, and environmental cumulative effects. Specifically, the dynamic data fusion deviation index is composed of two parts: an instantaneous deviation component and a temporal cumulative deviation component. The formula for calculating the dynamic data fusion deviation index is:

[0064] In the formula, for The dynamic fusion deviation index of multi-source data at any given time; This represents the total number of monitoring points. For the first Each monitoring point is weighted based on its spatial location and structural sensitivity. For the first monitoring points The deviation between the measured values ​​of structural deformation at any given time and the simulated values ​​of the digital twin model; For the first Time synchronization attenuation coefficient of each monitoring point; To unify the spatiotemporal reference and synchronize time; It is a positive number indicating stability; Correct the gain coefficient for geometric deviations at the manufacturing end; This is a correction function for environmental parameters coupled with ambient temperature and wind load. Deviation timing transfer function driven by component identification coding at the manufacturing end; Sensitivity coefficient for boundary conditions; To design the residual function after the coordinate system is transformed from the field monitoring coordinate system; This marks the start time of the current construction phase. It is the integral variable.

[0065] The formula consists of two terms, representing instantaneous state deviation and time-series cumulative deviation, respectively. The first term is a weighted normalization term for instantaneous deviation. Its design mechanism is based on the fact that different parts of the steel structure contribute differently to the overall safety. For example, the deformation sensitivity of the mid-span node of a large-span truss is significantly higher than that of the support node. Therefore, a weighting coefficient is introduced. Apply differentiated weighting. The specific values ​​can be determined through finite element perturbation analysis of the digital twin model. Meanwhile, considering that there is often a slight misalignment between the timestamps of the measured data collected by sensors at the construction site and the timestamps of the simulation output from the digital twin model, direct comparison would introduce spurious errors; therefore, an exponential decay factor is introduced. When the time deviation increases, the fusion weight of that data point is automatically reduced, thereby improving the algorithm's robustness to asynchronous data. In the denominator... It is used to normalize the weighted summation result, eliminate the influence of changes in the number of monitoring points on the magnitude of the index, and make the deviation index of monitoring networks of different sizes comparable.

[0066] The second term is the cumulative integral term of environmental and manufacturing deviations. Its design mechanism is that deviations in steel structure construction often have cumulative and delayed characteristics. For example, the thermal expansion and contraction effect caused by environmental temperature fluctuations, the dynamic response caused by wind loads, and the transmission of geometric deviations at the manufacturing end along the component installation sequence cannot be fully captured by instantaneous deviations. Temperature and wind load are coupled into an environmental correction function to reflect the combined effects of the environment; The evolution path of manufacturing errors during construction is traced by constructing a time-series record of geometric deviations associated with component identification codes in the manufacturing data package.

[0067] denominator As a coordinate alignment confidence suppression factor, when the transformation residual between the design coordinate system and the field monitoring coordinate system... A larger value indicates poor coordinate alignment quality. This factor automatically suppresses the contribution of the integral term, preventing the deviation exponent from being artificially exaggerated due to coordinate mapping errors. In practical applications, when... When the value is in the range [0, 0.3), it is considered a normal mapping state; when it is in the range [0.3, 0.7), it is considered a warning state and local parameter fine-tuning is initiated; when it is greater than or equal to 0.7, a global model parameter correction command is triggered. The above thresholds are recommended values ​​based on construction experience of large-span steel structure stadiums. In actual projects, each threshold can be calibrated according to the safety level, structural type, and statistical distribution of historical deviation data of the specific project. The specific mathematical expressions of the three core sub-functions in the above integral term are defined as follows. The environmental parameter correction function φ(τ) is defined as: , where ΔT(τ) is the temperature deviation value relative to the reference temperature at the corresponding time, v(τ) is the wind speed value at the corresponding time, and α1 and α2 are the weighting coefficients of temperature deviation and wind load, respectively, which are determined by statistical regression of the influence of temperature and wind speed on structural deformation in historical construction data.

[0068] The deviation timing transfer function P(τ) is calculated as follows: for each component installed up to that time, first calculate the attenuation factor. The function is defined as follows: where k is the deviation transmission attenuation coefficient, and tj is the installation time of the component. The geometric deviation amplitude d of the component is then multiplied by the attenuation factor to obtain the deviation transmission contribution value of the component. P(τ) is the sum of the deviation transmission contributions of all installed components. This function describes the transmission effect of manufacturing deviations of early-installed components in subsequent construction processes using an exponential attenuation form. The coordinate transformation residual function G(τ) is defined as the root mean square value of the positional deviation of each control point between the design coordinate system and the field monitoring coordinate system. When the control points are evenly distributed and the measurement accuracy is high, the value of G(τ) is small, and the contribution of the integral term is fully utilized. The time synchronization attenuation coefficient L is determined based on the statistical distribution of the sensor sampling period and data transmission delay, and is not restricted here.

[0069] Regarding the objective function for back-optimization of model parameters, to ensure that the corrected parameters match the measured response in the field, this embodiment constructs a composite objective function that includes manufacturing constraints. Its calculation formula is as follows:

[0070] In the formula, The parameter vector to be corrected The inverse optimization objective function; This represents the total number of monitoring points. For the first Measured structural response values ​​at each monitoring point; For the first The simulated response values ​​output by the digital twin model for each monitoring point; For the first Standard deviation of measurements at each monitoring point; The Tikhonov regularization coefficient; For the parameter space domain; For the parameter vector in the spatial domain gradient at; Constraint weights for material re-inspection properties and geometric deviation data at the manufacturing end; These are parameter regularization terms constructed based on data transmitted back from the manufacturing end.

[0071] The objective function consists of three terms, each corresponding to a specific technical problem-solving mechanism. The first term is the weighted sum of squared residuals, which aims to minimize the difference between the model output and the measured data. This is achieved by dividing by... This involves normalizing the residuals at each monitoring point to improve measurement accuracy. The smaller monitoring points have a higher weight in the optimization process to avoid diluting the correction contribution of high-precision data with low-precision data. The second term is the Tikhonov regularization term, which addresses the ill-posed inverse problem of structural parameter inversion. By penalizing the parameter gradient, it constrains the parameter field to change smoothly in space, preventing physically unreasonable parameter abrupt changes between adjacent units (such as drastic jumps in elastic modulus), and ensuring the mechanical rationality of the correction results. It can be determined by the L-curve method or the generalized cross-validation method to balance fitting accuracy and smoothness.

[0072] Third item The physical significance lies in the fact that the digital twin model has already corrected its initial state based on the manufacturing data package during the initialization phase. Therefore, the parameter correction in the subsequent construction phase should not deviate from this manufacturing authenticity base without limit. The physical constraints on the parameters are defined, and the corrected parameter vector... This item incurs a penalty value when the value deviates from the range defined by the manufacturing data package, forcing the optimization search to proceed within a reasonable range. For example, if the manufacturing data package records the measured yield strength of a component as 360 MPa, then... The corrected yield strength will be limited to a reasonable fluctuation range of the measured value (e.g., ±5%), rather than being allowed to deviate to ranges contradicting the measured value, such as 300MPa or 400MPa, in order to fit the field noise. Constraint weights The value is determined based on the data quality of the manufacturing data packet: when the data integrity rate at the manufacturing end is high and the measurement accuracy is good, A larger value is chosen to strengthen the constraint of manufacturing authenticity; when manufacturing data is missing or has low accuracy. Reduce the parameter size appropriately to allow for a wider range of parameter searches.

[0073] Specifically, R(θ) adopts a quadratic penalty function form with a dead zone. For each parameter to be corrected, the deviation ratio is first calculated. Where p is the correction value of the parameter, q is the measured value of the parameter in the generated data packet, and d is the allowable fluctuation range of the parameter. When r is less than or equal to 1, the penalty value is zero; when r is greater than 1, the penalty value is the square of (r-1). R(θ) is the sum of the penalty values ​​of all parameters.

[0074] This approach ensures both ample room for parameter fine-tuning near a realistic baseline and effective prevention of excessive deviation. Allowable fluctuation range. The measurement uncertainty and engineering safety margin of each parameter are set separately. For example, the yield strength can be taken as ±5% of the measured value, and the elastic modulus can be taken as ±3%. By introducing this item, this embodiment effectively prevents overfitting during the model calibration process, ensuring that the digital twin model is always anchored to the actual manufacturing state of the physical entity throughout its entire life cycle, thereby improving the reliability of the model prediction and its practical engineering value.

[0075] The technical advantages of this embodiment are as follows: By using semantic fusion under a unified spatiotemporal reference, it solves the technical problems of inconsistent formats, different coordinate references, and broken semantic associations among multi-source heterogeneous data, achieving alignment of design data, manufacturing data, and monitoring data. Through statistical analysis and reverse optimization correction of deviation sequences, a closed-loop mechanism of monitoring-fusion-correction is constructed, enabling the digital twin model to continuously track changes in the geometric, mechanical, and environmental states of the physical construction site, elevating the assessment of construction mechanical performance from discrete and lagging staged judgments to continuous and dynamic precise mapping. Simultaneously, by introducing parameter constraints based on manufacturing data, it ensures that the model correction process is always based on the actual manufacturing state of the physical entity, avoiding model distortion caused by blind correction, and further improving the robustness and reliability of the digital twin system.

[0076] As a preferred implementation method, after achieving dynamic correction of the digital twin model, intelligent analysis is performed using the corrected high-fidelity model, thereby realizing the transformation from passive monitoring to active control.

[0077] Specifically, the method in this embodiment further includes: The mechanical state output of the corrected digital twin model and the real-time collected structural deformation time series data are input into the construction anomaly identification model to identify anomalies where the response mode deviation exceeds the threshold. The abnormal working condition characteristics and environmental parameters of the identified abnormal working conditions are input into the mechanical performance risk prediction model to predict the probability and location information of structural damage or instability during a specified construction stage in the future. The abnormal working condition characteristics, the risk probability, and the implementation location information are input into the construction process optimization decision model to generate construction adjustment instructions.

[0078] In this embodiment, risk prediction and decision feedback are not simply black-box AI predictions, but are concretized into a three-level progressive structured process: anomaly identification, probability prediction, and process optimization. As the first level, the updated mechanical state output of the digital twin model and the real-time collected structural deformation time-series data are input into the construction anomaly identification model to identify anomaly conditions where the response pattern deviation exceeds a threshold. Specifically, the core of the construction anomaly identification model lies in the calculation of pattern similarity, rather than simple threshold comparison. This model has a pre-built library of normal construction response patterns trained based on historical normal construction data. To quantify the difference between the current state and the normal pattern, the following pattern similarity function is used for calculation:

[0079] In the formula, The similarity between the current construction response mode and the baseline mode in the normal construction response mode library; The length of the time window; The length of the structural feature; The field function of the current construction response mode is generated by fusing the mechanical state output of the digital twin model with real-time monitoring data; This is the baseline function for the normal construction response mode library; This is a correction factor for environmental disturbances. This is the coupling parameter between ambient temperature fluctuation data and wind load variation data. The formula extends the cosine similarity in vector space to a continuous spatiotemporal field. Anomalies in steel structure construction often manifest as distortions in the spatiotemporal distribution of the response (such as changes in stress wave propagation paths and abnormal deformation cloud map shapes), rather than simply exceeding limits at a single point. This is achieved through a spatiotemporal double inner product. This allows for the capture of deviations at the morphological level. Furthermore, considering the environmental sensitivity of steel structures, even if the structure is normal, the response pattern will exhibit a systematic shift when the ambient temperature or wind load deviates from the conditions recorded in the baseline model library. Environmental disturbance correction factor. This is used to compensate for such unstructured biases: when environmental biases are small. Approaching 1; when environmental deviations are large, The similarity value should be appropriately amplified to prevent false alarms caused by environmental interference. Specifically, H(e) is defined as a piecewise linear function: when e≤e0, H(e)=1; when e>e0, H(e)=1+b·(e-e0), where e0 is the baseline environmental condition value (the statistical mean of environmental parameters in the normal construction response model library), and b is the sensitivity coefficient. The environmental coupling parameter e is composed of a weighted combination of temperature deviation and wind load deviation, i.e. When actual environmental conditions are close to the baseline, no correction is needed. The greater the deviation, the larger the correction factor to compensate for the impact of environmental changes on the structural response. The sensitivity coefficient is determined statistically based on the correlation between environmental parameters and structural response in historical data, with a typical range of 0.05 to 0.5. When the calculated... When the value falls below a preset threshold (e.g., 0.6), the current response mode is determined to have a significant deviation, identified as an abnormal working condition, and characteristic information such as the abnormality type, location, and deviation magnitude is output. Regarding the construction process of the construction abnormal working condition identification model: the normal construction response mode library is generated based on historical normal construction data. The training data includes structural deformation time-series data, stress-strain data, environmental parameters, and corresponding construction procedure information. By performing spatiotemporal interpolation and statistical averaging on monitoring data under multiple sets of normal construction conditions, a benchmark function G(x,t) is constructed, and multiple benchmark modes under different construction stages and environmental conditions are stored to adapt to changes in working conditions.

[0080] As the second level, the identified abnormal working condition characteristics and environmental parameters are input into the mechanical performance risk prediction model to predict the probability and location of structural damage or instability during a specified future construction phase. This model employs a Logistic regression framework but introduces a crucial confidence correction mechanism to address the risk of extrapolation from small samples in construction scenarios. Its calculation formula is as follows:

[0081] In the formula, For the future Time structure in position The probability of damage or instability occurring at the location; For a linear predictor in a machine learning regression model, where These are the feature weight coefficients. For feature mapping function, For bias terms; This is a confidence correction function trained based on historical construction accident data; This represents the training sample coverage. In this formula, The standard Logistic function maps a linear combination of multidimensional risk factors to probability values ​​in the interval [0,1]. The abnormal operating condition characteristics (including anomaly type, location, and deviation magnitude) and environmental parameters (including ambient temperature and wind load) are respectively mapped by their corresponding feature mapping functions. Transform into a linear predictor The input feature vector, and the feature mapping functions map different types of input data into numerical features that can be processed by the regression model, and the feature weight coefficients. It reflects the degree to which each input feature contributes to the risk probability. Its core innovation lies in the confidence correction factor. In actual engineering projects, extreme risk events are often rare samples, when the current construction conditions lack representativeness in historical training data (i.e., When the value is low, conventional models are prone to making overconfident and incorrect predictions. Coverage is evaluated by calculating the Mahalanobis distance between the current feature vector and the training sample set. When the vector falls within the extrapolation region, the predicted probability value is automatically reduced, forcing decision-makers to recognize the uncertainty of the current prediction and thus avoiding blind reliance on model results. Specifically, K(ρ) is defined as an exponential saturation function: , where ρ is the reciprocal of the Mahalanobis distance from the current feature vector to the center of the training sample set (a larger ρ value indicates better representativeness of the current condition in the training data), and σ is a parameter controlling the decay rate, determined based on the distribution of eigenvalues ​​of the covariance matrix of the training sample set. When ρ is large, K(ρ) approaches 1, and the prediction probability is not corrected; when ρ is small (i.e., in the extrapolation region), K(ρ) gradually decreases, reducing the prediction probability. A value within the range of [0.5, 0.75] is considered high-risk and requires close monitoring; a value greater than or equal to 0.75 is considered extremely high-risk and requires immediate intervention; when... A value below 0.5 is considered low-risk or normal; maintain the regular monitoring frequency without special intervention. Regarding the training process of the mechanical performance risk prediction model: feature weight coefficients... and bias terms Based on historical construction accident data and normal construction data, the maximum likelihood estimation method is used for parameter estimation. To address the data imbalance problem caused by the scarcity of extreme risk event samples, Focal Loss is used instead of the standard cross-entropy loss function during training. The training weight of minority class samples is increased by a modulation factor. At the same time, Bootstrap resampling is used to expand the limited accident samples to improve the model's generalization ability under small sample conditions.

[0082] As the third level, abnormal working condition characteristics, predicted risk probabilities, and location information are input into the construction process optimization decision model to generate construction adjustment instructions. This model is essentially a constrained multi-objective optimization problem, aiming to balance safety, cost, and schedule. Its decision function is defined as:

[0083] The constraints are: , .

[0084] In this formula, The decision variable vector includes component hoisting sequence, temporary support arrangement, welding process parameters, etc. This represents the feasible solution space. Normalized weights for safety risks, construction costs, and construction delays, respectively, and satisfying the following conditions: ; , , These correspond to the risk probability, construction cost, and construction delay of the decision variables, respectively; superscript and Represents the extreme values ​​within the feasible region; This is the total planned construction period.

[0085] In this decision function, the risk probability It is directly used as input to the safety risk term in the objective function; the abnormal operating condition features are used to determine the inequality constraints. The constraint boundaries impose corresponding safety restrictions on the decision-making scheme for specific anomaly types; the location information is used to limit the decision variables. The formula's effective area ensures that generated construction adjustment instructions accurately target the specific location where the risk occurs. It employs a min-max normalized weighted summation framework to avoid direct comparisons between risk (dimensionless), cost (monetary), and schedule (time). Weights It can be dynamically adjusted according to the construction stage, for example, increasing the speed during the critical hoisting stage. Safety must be the top priority. Constraints. and This ensures the feasibility of the generated decision-making scheme. Among these, inequality constraints... This includes the rated capacity of the lifting equipment (to prevent overloading), site space limitations, and material strength limits (derived from measured data in the manufacturing data package to prevent stress exceeding limits in decision-making); equality constraints. This ensures the continuity of the process (procedural logic) and the mechanical balance of the nodes (the structure must be stable under each construction step). By solving this optimization problem, specific, executable, and physically safe construction adjustment instructions can be automatically generated, such as "pause hoisting and add temporary support at node N-03" or "adjust the welding sequence to reduce residual stress," realizing intelligent reverse control of the construction site by the digital twin model. Regarding the solution method of the construction process optimization decision model: Since the decision variable x contains both discrete variables (such as the component hoisting sequence) and continuous variables (such as welding process parameters), this embodiment uses a mixed integer nonlinear programming solver for solving the problem. Specifically, a combination strategy of using the branch and bound method to handle discrete variables and the interior point method to handle continuous variables can be used; in large-scale problems, a genetic algorithm can also be used to encode discrete and continuous variables into chromosomes for global search.

[0086] The specific mathematical expression of constraints is shown below: Inequality constraints This indicates the lifting capacity constraint, where For the actual lifting load under the corresponding decision variables, This refers to the crane's rated lifting capacity under current operating conditions. Represents stress constraint, where For the maximum stress of the critical component, To generate the measured yield strength recorded in the data packet, For safety factor. Equality constraint. This indicates the force balance constraint at the nodes, ensuring that the structural nodes meet the static equilibrium conditions under each construction step.

[0087] It should be understood that the specific algorithm implementations of the above three models are only illustrative examples. In other implementations, anomaly identification can also use convolutional neural networks to extract spatiotemporal features, risk prediction can use Bayesian networks to handle uncertainty, and decision optimization can use genetic algorithms or reinforcement learning, mainly following the three-level progressive logic of identification-prediction-decision.

[0088] Example 2: See Figure 2 This embodiment provides a BIM-based digital twin model creation system for steel structure construction, including: The design model acquisition module is used to acquire the design model of the steel structure. The design model is built based on BIM and contains the geometric information and identification of the components. A data packet acquisition module is used to acquire the manufacturing data packet of the component, wherein the manufacturing data packet contains the component's identification, geometric deviation, and material properties; wherein the process of acquiring the manufacturing data packet includes: parsing the parametric BIM model, extracting the component's geometric parameters and process constraints, generating processing instructions to drive the processing equipment to execute component processing; after the component processing is completed, collecting the component's actual geometric deviation data and material property data; binding the actual geometric deviation data, the material property data, and the component's identification to form the manufacturing data packet; The model building module is used to map the geometric deviations and material properties in the manufacturing data package to the geometric information and material parameters of the corresponding components in the design model based on their identities, thereby establishing a data mapping relationship between the measured data from the manufacturing end and the design model; constructing the initial geometric state of the digital twin model based on the design model; and correcting the initial geometric state and updating the model parameters of the digital twin model according to the manufacturing data package based on the data mapping relationship, thereby generating a digital twin model that integrates the manufacturing end data.

[0089] As a preferred implementation, updating the model parameters of the digital twin model includes: The geometric deviations of the component are superimposed onto the initial geometric state of the digital twin model to correct the initial geometric state of the component; The material properties are updated in the material constitutive relation definition of the digital twin model.

[0090] In a preferred embodiment, the system further includes: The data fusion and correction module is used to collect a set of response parameters of the steel structure during construction, and to perform semantic fusion of the set of response parameters with the design model and the manufacturing data package under a unified spatiotemporal reference. Based on the fusion result, the digital twin model is updated with parameters and its state is corrected.

[0091] As a preferred implementation, the step of updating parameters and correcting the state of the digital twin model based on the fusion results includes: Extract structural deformation time series data from the response parameter set, and calculate the deviation sequence between the measured structural deformation value and the simulated deformation value of the digital twin model at the corresponding time. Statistical analysis is performed on the deviation sequence, and when the statistical characteristics of the deviation sequence exceed a preset tolerance threshold, a model parameter correction command is triggered. The material property parameters, boundary condition parameters, or component geometric parameters in the digital twin model are adjusted in reverse using an optimization algorithm, so that the deviation between the simulated response value output by the adjusted digital twin model and the measured value of the structural deformation meets the convergence criterion.

[0092] In a preferred embodiment, the statistical analysis of the deviation sequence is based on a dynamic data fusion deviation index, which is calculated by weighting the weight coefficients of each monitoring point, the measured deviation values, and the time synchronization attenuation coefficients. The objective function of the optimization algorithm is constructed based on the deviation between the measured response values ​​and the simulated response values ​​of each monitoring point, the regularization constraints of the parameter space domain, and the parameter constraint terms based on the manufacturing end data.

[0093] In a preferred embodiment, the system further includes: The prediction and feedback module is used for construction risk prediction and decision feedback based on the corrected digital twin model. The construction risk prediction and decision feedback includes: The mechanical state output of the corrected digital twin model and the real-time collected structural deformation time series data are input into the construction anomaly identification model to identify anomalies where the response mode deviation exceeds the threshold. The abnormal working condition characteristics and environmental parameters of the identified abnormal working conditions are input into the mechanical performance risk prediction model to predict the probability and location information of structural damage or instability during a specified construction stage in the future. The abnormal working condition characteristics, the risk probability, and the implementation location information are input into the construction process optimization decision model to generate construction adjustment instructions.

[0094] Example 3 This embodiment provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described in any of the above embodiments.

[0095] Example 4 This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any one of the above embodiments.

[0096] This application describes the invention from the perspectives of purpose, effectiveness, progress, and novelty, and it meets the functional enhancement and use requirements emphasized by the Patent Law. The above description and drawings are only preferred embodiments of this application and are not intended to limit this application. Therefore, all structures, devices, features, etc. that are similar to or identical to those of this application, i.e., all equivalent substitutions or modifications made in accordance with the scope of this patent application, shall fall within the scope of protection of this patent application.

Claims

1. A method for establishing a BIM-based digital twin model for steel structure construction, characterized in that, include: Obtain the design model of the steel structure, which is built based on BIM and includes the geometric information and identification of the components; Obtain the manufacturing data package of the component, the manufacturing data package containing the component's identification, geometric deviations, and material properties; Based on their identities, the geometric deviations and material properties in the manufacturing data package are mapped to the geometric information and material parameters of the corresponding components in the design model, establishing a data mapping relationship between the measured data at the manufacturing end and the design model; and the initial geometric state of the digital twin model is constructed based on the design model. Based on the data mapping relationship, the initial geometric state is corrected and the model parameters of the digital twin model are updated according to the manufacturing data package to generate a digital twin model that integrates manufacturing data. The process of acquiring the manufacturing data package includes: parsing the parametric BIM model, extracting the component's geometric parameters and process constraints, generating processing instructions to drive the processing equipment to execute component processing; after the component processing is completed, collecting the component's actual geometric deviation data and material property data; binding the actual geometric deviation data, the material property data, and the component's identity identifier to form the manufacturing data package.

2. The method according to claim 1, characterized in that, Updating the model parameters of the digital twin model includes: The geometric deviations of the component are superimposed onto the initial geometric state of the digital twin model to correct the initial geometric state of the component; The material properties are updated in the material constitutive relation definition of the digital twin model.

3. The method according to claim 1, characterized in that, The method further includes: During construction, a set of response parameters of the steel structure is collected, and the set of response parameters is semantically fused with the design model and the manufacturing data package under a unified spatiotemporal reference. Based on the fusion result, the parameters of the digital twin model are updated and the state is corrected.

4. The method according to claim 3, characterized in that, The parameter update and state correction of the digital twin model based on the fusion results include: Extract structural deformation time series data from the response parameter set, and calculate the deviation sequence between the measured structural deformation values ​​and the simulated deformation values ​​of the digital twin model at the corresponding time. Statistical analysis is performed on the deviation sequence, and when the statistical characteristics of the deviation sequence exceed a preset tolerance threshold, a model parameter correction command is triggered. The material property parameters, boundary condition parameters, or component geometric parameters in the digital twin model are adjusted in reverse using an optimization algorithm, so that the deviation between the simulated response value output by the adjusted digital twin model and the measured value of the structural deformation meets the convergence criterion.

5. The method according to claim 4, characterized in that, The statistical analysis of the deviation sequence is based on the dynamic data fusion deviation index, which is calculated by weighting the weight coefficients of each monitoring point, the measured deviation values, and the time synchronization decay coefficient. The objective function of the optimization algorithm is constructed based on the deviation between the measured response values ​​and the simulated response values ​​of each monitoring point, the regularization constraints of the parameter space domain, and the parameter constraint terms based on the manufacturing end data.

6. The method according to claim 3, characterized in that, The method further includes: Construction risk prediction and decision feedback are performed based on the corrected digital twin model, including: The mechanical state output of the corrected digital twin model and the real-time collected structural deformation time series data are input into the construction anomaly identification model to identify anomalies where the response mode deviation exceeds the threshold. The abnormal working condition characteristics and environmental parameters of the identified abnormal working conditions are input into the mechanical performance risk prediction model to predict the probability and location information of structural damage or instability during a specified construction stage in the future. The abnormal working condition characteristics, the risk probability, and the implementation location information are input into the construction process optimization decision model to generate construction adjustment instructions.

7. A BIM-based digital twin model creation system for steel structure construction, characterized in that, include: The design model acquisition module is used to acquire the design model of the steel structure. The design model is built based on BIM and contains the geometric information and identification of the components. A data packet acquisition module is used to acquire the manufacturing data packet of the component, wherein the manufacturing data packet contains the component's identification, geometric deviation, and material properties; wherein the process of acquiring the manufacturing data packet includes: parsing the parametric BIM model, extracting the component's geometric parameters and process constraints, generating processing instructions to drive the processing equipment to execute component processing; after the component processing is completed, collecting the component's actual geometric deviation data and material property data; binding the actual geometric deviation data, the material property data, and the component's identification to form the manufacturing data packet; The model building module is used to map the geometric deviations and material properties in the manufacturing data package to the geometric information and material parameters of the corresponding components in the design model based on their identities, thereby establishing a data mapping relationship between the measured data from the manufacturing end and the design model; constructing the initial geometric state of the digital twin model based on the design model; and correcting the initial geometric state and updating the model parameters of the digital twin model according to the manufacturing data package based on the data mapping relationship, thereby generating a digital twin model that integrates the manufacturing end data.

8. The system according to claim 7, characterized in that, The system also includes: The data fusion and correction module is used to collect a set of response parameters of the steel structure during construction, and to perform semantic fusion of the set of response parameters with the design model and the manufacturing data package under a unified spatiotemporal reference. Based on the fusion result, the digital twin model is updated with parameters and its state is corrected.

9. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images