Steel structure building construction whole process mechanical property evaluation method based on digital twinning

Abstract

The invention relates to the technical field of building construction monitoring, and discloses a method for evaluating mechanical properties of a steel structure building construction whole process based on digital twinning. The method comprises the steps of establishing a digital twin model fusing multi-source information, and performing real-time linkage with a sensor network arranged on site. Collected data such as deformation, temperature and wind speed are processed through dynamic fusion and an anomaly recognition algorithm, model parameters are continuously corrected, and real-time dynamic high-fidelity mapping of the mechanical state in the construction process is achieved. And based on the updated model, the intelligent analysis module performs cooperative calculation, autonomously identifies construction abnormity, quantitatively predicts potential risks, generates a process optimization decision instruction and feeds back the process optimization decision instruction to a site. According to the method, a closed-loop regulation and control mechanism from data perception, model analysis to decision execution is constructed, and accurate online evaluation of construction mechanical properties and active prediction control of safety risks are realized.

Description

Technical Field

[0001] This invention relates to the field of building construction monitoring technology, specifically a method for evaluating the mechanical performance of steel structure buildings throughout the entire construction process based on digital twins. Background Technology

[0002] In the traditional construction of tall steel structure buildings, the assessment of structural mechanical performance mainly relies on finite element simulation based on design drawings and phased manual on-site monitoring. Conventional techniques typically employ pre-construction simulations or offline data comparisons after key milestones. The monitoring data is often isolated and static, making it difficult to continuously reflect the true structural state during the dynamic construction process. Existing digital twin technology in building construction often focuses on visualizing geometric progress or displaying static models. Its mechanical model parameters are usually fixed throughout the construction period, lacking the ability to be updated in real-time and adaptively based on the actual on-site environment and structural response.

[0003] The shortcomings of existing technical solutions lie in the significant lag and one-sidedness of structural safety assessments during construction. Because the model and the physical entity on site fail to achieve real-time synchronization of their mechanical states, the identification of abnormal conditions relies on human experience and post-event analysis, failing to provide early warnings of construction risks. Furthermore, an effective closed loop is not formed between monitoring data and construction decisions. Even if potential risks are identified, there is a lack of quantitative decision support methods for dynamically optimizing construction processes based on real-time mechanical states; construction control still relies primarily on manual judgment and predetermined plans. The purpose of this invention is to solve the problem of real-time dynamic mapping between digital twin models and the mechanical states of physical entities during construction, and to achieve closed-loop proactive control of construction based on real-time mechanical performance assessment. Summary of the Invention

[0004] The purpose of this invention is to provide a method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins, the method comprising:

[0006] A digital twin model of the construction process of a tall steel structure is established. The digital twin model integrates structural design information, construction schedule and material physical properties. The digital twin model is dynamically mapped in real time to the geometric, mechanical and environmental conditions of the actual construction site.

[0007] A multi-dimensional sensor monitoring network is deployed at the construction site to collect the set of response parameters of the steel structure in real time during the construction process. The set of response parameters includes structural deformation time series data, ambient temperature fluctuation data, and on-site wind speed change data.

[0008] The set of response parameters is synchronously input into the digital twin model. Based on the dynamic data fusion algorithm and the abnormal state identification algorithm, the parameters of the digital twin model are updated and the state is corrected to generate a visual diagnostic result of the construction process state and a quantitative assessment index of safety risks.

[0009] The pre-integrated intelligent analysis model is invoked to perform collaborative analysis and processing on the updated digital twin model and the real-time collected response parameter set. The collaborative analysis and processing includes autonomous identification of abnormal construction conditions, quantitative prediction of potential mechanical performance risks, and decision support for construction process optimization.

[0010] The decision support based on the optimized construction process generates construction instructions that are fed back to the construction site, thus constructing a closed-loop management mechanism that integrates data-driven approaches, model verification, and construction control.

[0011] Preferably, the establishment of a digital twin model of the construction process of the tall steel structure includes:

[0012] A three-dimensional geometric model of the tall steel structure is constructed based on the structural design information. The three-dimensional geometric model includes the spatial coordinates of all structural component nodes, component cross-sectional dimensions, and connection details between components.

[0013] Material constitutive relations are embedded on the basis of the three-dimensional geometric model. The material constitutive relations are based on the material physical properties defined, including the elastic modulus, yield strength, ultimate strength of steel, and constitutive model parameters.

[0014] The construction schedule is transformed into a load application sequence and a boundary condition change sequence in the time dimension. The load application sequence includes the application time and location of construction machinery loads, temporary support loads and self-weight loads of installed components. The boundary condition change sequence includes the temporal information of the evolution of structural constraint state with the construction stage.

[0015] By integrating the three-dimensional geometric model, the material constitutive relation, the load application sequence, and the boundary condition change sequence, a construction process digital twin model with time-varying characteristics is generated. The construction process digital twin model can simulate the structural geometry and mechanical state at any construction moment.

[0016] Preferably, the step of updating parameters and correcting the state of the digital twin model based on dynamic data fusion algorithm and abnormal state identification algorithm to generate visualized diagnostic results of the construction process state and quantitative assessment indicators of safety risks includes:

[0017] Extract structural deformation time series data from the response parameter set, and calculate the difference sequence between the measured structural deformation value and the simulated deformation value of the digital twin model at the corresponding time.

[0018] Statistical analysis is performed on the difference sequence to extract the mean, variance, and trend characteristics of the difference sequence. When the mean or variance of the difference sequence exceeds a preset tolerance threshold, a model parameter correction command is triggered.

[0019] According to the model parameter correction instructions, the material constitutive relation parameters or boundary condition parameters in the digital twin model are adjusted in reverse based on the optimization algorithm, so that the difference between the deformation simulation value output by the adjusted digital twin model and the measured deformation value of the structure meets the convergence criterion.

[0020] Based on the corrected digital twin model, the stress distribution, displacement response and overall stability safety factor of the key sections of the structure under the current construction stage are calculated, and a visual diagnostic result including mechanical state cloud map and safety level label is generated.

[0021] By combining ambient temperature fluctuation data and on-site wind speed change data, the structural dynamic characteristics are corrected. Based on the corrected dynamic characteristics and the overall stability safety factor, the failure probability of the structure in subsequent construction steps is calculated, and a quantitative safety risk index is generated.

[0022] Preferably, the step of calling a pre-integrated intelligent analysis model to perform collaborative analysis and processing on the updated digital twin model and the set of response parameters collected in real time includes:

[0023] 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. The construction anomaly identification model identifies anomalies where the response mode deviation exceeds a threshold by comparing the mechanical response mode with a preset normal construction response mode library. The anomalies include local buckling signs, abnormal node connections, or overload.

[0024] The identified abnormal working condition characteristics, current ambient temperature fluctuation data, and on-site wind speed change data are input into the mechanical performance risk prediction model. The mechanical performance risk prediction model is trained and generated based on historical accident data and mechanical simulation data, and is used to predict the probability and possible location of structural damage or instability in a specified construction stage in the future.

[0025] The abnormal working conditions, predicted risk probabilities, and location information are input into the construction process optimization decision model. The construction process optimization decision model is based on a multi-objective optimization algorithm and generates decision suggestions such as adjusting the construction sequence, reinforcing temporary supports, or changing the hoisting scheme to reduce the risk probabilities.

[0026] Preferably, the method further includes a preliminary analysis of the structural dynamic response and fatigue life based on a digital twin model before construction, including:

[0027] An equivalent wind load time history is applied to the digital twin model, the equivalent wind load time history being generated based on wind tunnel test data or standard wind spectra of the target construction site;

[0028] Perform wind-induced dynamic time history analysis to obtain the stress time history curves of key structural components under the equivalent wind load time history.

[0029] The cycle counting algorithm is applied to the stress time history curve to obtain the number of cycles corresponding to each stress amplitude, thus forming the probability density distribution of the stress amplitude;

[0030] Based on the probability density distribution of the stress amplitude and the fatigue strength curve of the material, the cumulative damage theory is used to calculate the wind-induced fatigue damage and estimated fatigue life of the key components under the design wind speed during the construction period.

[0031] Preferably, the method further includes assessing the structural dynamic response and vulnerability under seismic loading based on a digital twin model, including:

[0032] A refined three-dimensional finite element model of the tall steel structure was established as the core mechanical analysis model of the digital twin model;

[0033] Modal analysis was performed on the refined three-dimensional finite element model to obtain the natural frequencies, mode shapes, and participation coefficients of the structure.

[0034] Multiple ground motion time histories with different intensity characteristics were selected as inputs, and elastic time history analysis under frequent earthquakes and elastoplastic time history analysis under rare earthquakes were performed on the refined three-dimensional finite element model respectively.

[0035] Based on the results of the elastic time history analysis and the elastoplastic time history analysis, the overall dynamic response parameters and component damage indices of the structure under different intensity ground motions are extracted. The overall dynamic response parameters include inter-story drift angle and vertex acceleration, and the component damage indices include plastic hinge development state and cross-sectional yielding degree.

[0036] A two-parameter system with seismic ground motion intensity index as the abscissa and structural damage index as the ordinate was established, and seismic vulnerability curves of the structure under different damage states were obtained based on statistical analysis and fitting.

[0037] Preferably, the method further includes parameter sensitivity analysis and structural optimization based on the seismic vulnerability analysis results, including:

[0038] A set of key parameters affecting the seismic performance of the structure is selected. The set of key parameters includes the geometric cross-sectional dimensions of the components, the material property parameters of the steel, and the construction details of the node connections.

[0039] In the refined three-dimensional finite element model, one or more parameters in the set of key parameters are perturbed and adjusted, and the modal analysis, elastic time history analysis and elastoplastic time history analysis are re-executed;

[0040] By comparing and analyzing the changes in the seismic vulnerability curves of the structure before and after parameter adjustment, the influence of each key parameter on the seismic vulnerability of the structure is quantitatively assessed, and the key parameters most sensitive to the adjustment of the vulnerability curve are identified.

[0041] Based on the results of the parameter sensitivity analysis, weak links and potential failure modes of the structure under special seismic conditions are identified.

[0042] A seismic performance improvement strategy is proposed, which focuses on adjusting the cross-sectional configuration of sensitive components, enhancing the energy dissipation structure of key nodes, and optimizing the stiffness distribution of the overall structure, thereby generating a structural optimization design scheme that takes into account both seismic safety and economy.

[0043] Preferably, the construction and training of the mechanical performance risk prediction model includes:

[0044] Historical steel structure construction accident case data, laboratory component test data, and a large amount of high-fidelity mechanical simulation data are collected to form a training dataset. Each data point in the training dataset contains input features and output labels.

[0045] The input features include the geometric morphology description of the structure at the construction stage corresponding to the data used for model training, load conditions, identified abnormal working conditions, environmental conditions, and key mechanical indicators calculated by the digital twin model.

[0046] The output label indicates whether the structure suffers a specific type of damage or instability in subsequent construction steps under the construction stage and working conditions, and the corresponding risk level.

[0047] The machine learning model, which is a deep neural network or a gradient boosting decision tree, is trained using the training dataset under supervision. The training objective is to minimize the error between the risk level predicted by the model and the true label.

[0048] The trained machine learning model is deployed as the mechanical performance risk prediction model, which is used to receive real-time input features online and output quantified risk probability and location prediction.

[0049] Preferably, the closed-loop management mechanism for data-driven, model-validation, and construction control includes:

[0050] The decision support for optimizing the construction process is transformed into specific construction adjustment instructions, which include adjusting the hoisting sequence of components, modifying the temporary support layout scheme, or changing the welding process parameters.

[0051] The construction adjustment instructions are executed at the construction site, and at the same time, the new set of response parameters after the instructions are collected through the multi-dimensional sensor monitoring network.

[0052] The new set of response parameters is input into the digital twin model to verify the improvement of the structural mechanical performance after the adjustment command is executed, and to calculate a new quantitative assessment index of safety risks.

[0053] By comparing the quantitative assessment indicators of safety risks before and after the adjustment, the effectiveness of the construction adjustment instructions can be evaluated.

[0054] If the new quantitative assessment indicators for security risks still fail to meet the expected goals, a new round of collaborative analysis and decision generation will be initiated until the risk indicators meet the requirements, forming a complete closed loop of assessment, decision-making, execution, and verification.

[0055] Preferably, the method further includes a comprehensive evaluation and report generation of the mechanical properties throughout the construction process, including:

[0056] The results of fatigue life prediction based on wind load before construction, vulnerability assessment based on seismic action, risk assessment based on real-time monitoring during construction, and performance verification results after adjustment based on optimization decision-making are integrated.

[0057] Based on the predetermined evaluation index system, the integrated results are weighted and comprehensively scored. The evaluation index system covers multiple dimensions such as structural safety, construction feasibility, economy and durability.

[0058] Based on the weighted comprehensive score, a comprehensive evaluation level of the mechanical performance of the steel structure building throughout the entire construction process is generated;

[0059] All analysis process data, intermediate results, final evaluation level, and key decision recommendations are summarized to form a structured mechanical performance evaluation report for the entire construction process.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] Utilizing a multi-dimensional sensor network deployed at the construction site, the system continuously collects real-time data on structural deformation, ambient temperature, and wind speed changes. Through a dynamic data fusion algorithm, it synchronously integrates and calibrates multi-source heterogeneous data with the geometric topology, material properties, and boundary conditions of the digital twin model. This process, combined with anomaly identification algorithms, proactively identifies noise and biases in the measurement data and dynamically corrects the model parameters. This allows the digital twin model to break free from reliance on idealized assumptions made during the design phase, forming a dynamic mirror image that consistently maintains the same mechanical state as the physical construction site. This elevates the understanding of construction mechanical performance from discrete, lagging, phased assessments to continuous, real-time, and precise mapping, improving the timeliness and reliability of structural safety monitoring.

[0062] Based on a high-fidelity twin model with real-time correction, an integrated intelligent analysis model is invoked to perform in-depth collaborative analysis of structural response data. This analysis process can not only autonomously identify and locate abnormal mechanical behaviors during construction, but also quantitatively predict potential performance risks such as instability and overload based on real-time status and construction progress. The conclusions generated by the analysis are directly translated into specific construction process adjustment decisions and automatically fed back to the construction site to guide execution. This mechanism constructs an automated closed loop of "perception-analysis-decision-control," transforming traditional, passive construction management that relies on human experience and fixed processes into a proactive control mode driven by real-time data and models, possessing predictive and adaptive capabilities, thereby improving the risk management capabilities and overall scientific nature of the construction process. Attached Figure Description

[0063] Figure 1 This is a schematic diagram illustrating the working principle of the method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins, as described in this invention.

[0064] Figure 2 A flowchart for establishing a digital twin model of the construction process of tall steel structures;

[0065] Figure 3 A flowchart for collaborative analysis and processing by intelligent analysis models;

[0066] Figure 4 The curves showing the changes in training loss and validation accuracy for a deep neural network model;

[0067] Figure 5 A heat map showing the distribution of safety risk levels at each construction stage. Detailed Implementation

[0068] 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.

[0069] Please see Figure 1 This invention provides a method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins. The method includes: establishing a digital twin model of the construction process of a tall steel structure, which integrates structural design information, construction schedule, and material physical properties, and dynamically maps this model to the geometric, mechanical, and environmental conditions of the actual construction site in real time. A multi-dimensional sensor monitoring network is deployed at the construction site to collect a set of response parameters of the steel structure during construction in real time. This set includes time-series data of structural deformation, ambient temperature fluctuations, and on-site wind speed changes. These real-time collected response parameters are synchronously input into the digital twin model, and the model is updated and its state corrected based on dynamic data fusion algorithms and anomaly identification algorithms, thereby generating visualized diagnostic results of the construction process status and quantitative assessment indicators of safety risks. A pre-integrated intelligent analysis model is invoked to collaboratively analyze the updated digital twin model and the real-time collected response parameters. This process includes autonomous identification of abnormal construction conditions, quantitative prediction of potential mechanical performance risks, and decision support for construction process optimization. Decision support based on construction process optimization generates specific construction instructions and feeds them back to the construction site, thereby building a closed-loop management mechanism that integrates data-driven approaches, model validation, and construction control.

[0070] Example 1: See Figure 2 A three-dimensional geometric model of the tall steel structure is constructed based on the structural design information. This model includes the spatial coordinates of all structural member nodes, member cross-sectional dimensions, and details of the connections between members. Material constitutive relations are embedded into the three-dimensional geometric model. These relations are based on material physical properties, including the elastic modulus, yield strength, ultimate strength of the steel, and constitutive model parameters. The construction schedule is transformed into a time-dimensional load application sequence and boundary condition change sequence. The load application sequence includes the application time and location of construction machinery loads, temporary support loads, and the self-weight loads of installed members. The boundary condition change sequence includes the temporal information of the evolution of structural constraint states as construction progresses. By integrating the three-dimensional geometric model, material constitutive relations, load application sequence, and boundary condition change sequence, a time-varying digital twin model of the construction process is generated. This digital twin model can simulate the structural geometry and mechanical state at any given construction moment.

[0071] In practical implementation, establishing a digital twin model of the construction process of the towering steel structure is a core step. The construction of the digital twin model begins with structural design information, which typically includes complete construction drawings, calculation sheets, and component lists. In a construction example of a 200-meter-high lattice-structured TV tower, the structural design information was parsed into a digitized database of component spatial relationships. This database precisely records the three-dimensional spatial coordinates of over 5,000 major nodes, with the accuracy of node coordinates controlled to the millimeter level. Based on the structural design information, a three-dimensional geometric model of the towering steel structure is constructed in a computer-aided design platform. This three-dimensional geometric model not only reproduces the geometric shapes of all main columns, crossbars, and diagonal braces, but also defines in detail the connection details between components. These connection details include the arrangement of high-strength bolt groups, the beveling form of welds, and the dimensions and thickness of gusset plates, ensuring that the three-dimensional geometric model is completely consistent with the solid structure at the geometric level.

[0072] In practical implementation, the three-dimensional geometric model serves as the carrier of physical properties. Material physical properties are defined as key parameters of the model, directly derived from the material certificates and re-inspection reports of the steel used in the steel structure project. For example, when the main structure uses Q345B steel, the material physical properties explicitly define the steel's elastic modulus as 2.06 × 10⁵ MPa, yield strength as 345 MPa, and ultimate strength as 470 to 630 MPa. Based on the three-dimensional geometric model, these material physical properties are used to embed the material constitutive relation, which defines the stress-strain response behavior of the steel. For some complex nodes requiring nonlinear analysis, the material constitutive relation may employ a multi-segmented model including strengthening segments, whose mathematical expression can be formally described as:

[0073]

[0074] in; Represents engineering stress, Represents engineering strain. Represents the elastic modulus. Represents yield strength. Represents yield strain. Represents the tangent modulus. Represents the ultimate strain.

[0075] In some embodiments, the static model needs to be combined with the dynamic construction process. The construction schedule is provided in the form of a network diagram or Gantt chart, and the construction schedule is transformed into a series of discrete events in the time dimension. The transformation process resolves the construction schedule into a load application sequence and a boundary condition change sequence. The load application sequence explicitly specifies the new load information under each construction step. For example, in the event of "lifting the 20th section on day 15", the load application sequence records the self-weight load of all components in that section, the concentrated load of the lifting machinery acting on the structure, and any possible construction live loads, and precisely specifies the position and direction of these loads on the three-dimensional geometric model. The boundary condition change sequence synchronously records the evolution of the structural constraint state. The boundary condition change sequence may be described as follows: after the event of "installing the 10th floor temporary support", the horizontal displacement constraint of the structure at the corresponding elevation is activated; after the event of "completing the welding of the 5th section and removing the lower temporary support", the vertical constraint at the original support point is released.

[0076] In some embodiments, discrete model components are integrated into a unified computational entity. The integration process inputs the 3D geometric model, material constitutive relations, load application sequences, and boundary condition change sequences into the finite element analysis kernel. Based on the construction timeline, the finite element analysis kernel sequentially calls the load application sequences and boundary condition change sequences, progressively applying loads and changing constraints on the 3D geometric model, thereby numerically simulating the continuous construction process of the structure. This ultimately generates a time-varying digital twin model of the construction process, capable of retrospectively or prospectively simulating the state at any specified construction moment.

[0077] Example 2: See Figure 3Based on dynamic data fusion algorithms and abnormal state identification algorithms, the digital twin model is updated and its state is corrected, generating visualized diagnostic results of the construction process state and quantitative assessment indicators of safety risks. Structural deformation time-series data is extracted from the response parameter set, and the difference sequence between the measured structural deformation value and the simulated deformation value of the digital twin model at the corresponding time is calculated. Statistical analysis is performed on this difference sequence to extract its mean, variance, and trend characteristics. When the mean or variance of the difference sequence exceeds a preset tolerance threshold, a model parameter correction command is triggered. According to the model parameter correction command, the material constitutive relation parameters or boundary condition parameters in the digital twin model are adjusted in reverse using optimization algorithms, so that the difference between the simulated deformation value output by the adjusted digital twin model and the measured structural deformation value meets the convergence criterion. Based on the corrected digital twin model, the stress distribution, displacement response, and overall stability safety factor of the key sections of the structure under the current construction stage are calculated, generating visualized diagnostic results including a mechanical state cloud map and safety level labels. By combining ambient temperature fluctuation data and on-site wind speed change data, the structural dynamic characteristics are corrected. Based on the corrected dynamic characteristics and the overall stability safety factor, the failure probability of the structure in subsequent construction steps is calculated, and a quantitative safety risk index is generated.

[0078] The updated digital twin model and the real-time collected response parameter set are collaboratively analyzed using a pre-integrated intelligent analysis model. The mechanical state output of the updated digital twin model and the real-time collected structural deformation time-series data are input into the construction anomaly identification model. This model identifies anomalies where the response pattern deviates from a pre-set normal construction response pattern library, including signs of local buckling, abnormal node connections, or excessive loads. The identified anomaly characteristics, current ambient temperature fluctuations, and on-site wind speed changes are input into the mechanical performance risk prediction model. This model, trained based on historical accident data and mechanical simulation data, predicts the probability and possible location of structural damage or instability during a specified future construction phase. Finally, the anomalies, predicted risk probabilities, and location information are input into the construction process optimization decision model. This model, based on a multi-objective optimization algorithm, generates recommendations to adjust the construction sequence, reinforce temporary supports, or change the hoisting scheme to reduce the risk probability.

[0079] In practical implementation, updating parameters and correcting the state of the digital twin model based on dynamic data fusion algorithms and abnormal state identification algorithms is a key process. A set of response parameters is acquired in real time from a multi-dimensional sensor monitoring network deployed at the construction site. This set includes time-series data on structural deformation, ambient temperature fluctuations, and on-site wind speed changes. The time-series data on structural deformation is continuously collected by high-precision tilt sensors and total stations deployed at key structural nodes. For example, in the segmented lifting construction of a large-span steel truss, the response parameter set is updated hourly, containing three-dimensional displacement and rotation data from twenty monitoring points. The time-series data on structural deformation is extracted from the response parameter set. The measured values ​​of the structural deformation time-series data are compared point-by-point with the simulated deformation values ​​output by the digital twin model at the corresponding construction simulation time. A set of difference sequences is calculated, where each element of the difference sequence represents the deviation between the measured and simulated values ​​at a specific monitoring point at a specific time.

[0080] In practice, the difference sequence needs to be systematically analyzed to determine the model's reliability. Statistical analysis is performed on the difference sequence, including calculating its arithmetic mean and unbiased variance, and using linear regression analysis to determine the trend of the difference sequence as construction progresses. Preset tolerance thresholds are established based on engineering accuracy requirements and specifications; for example, the tolerance threshold for the main structure's displacement might be set at 30% of the design allowable value. When either the arithmetic mean or the unbiased variance of the difference sequence exceeds the preset tolerance threshold, the system automatically triggers a model parameter correction command. This command is a logical signal indicating that certain input parameters in the digital twin model may have deviations and require adjustment.

[0081] In some embodiments, executing model parameter calibration instructions involves the application of an optimization algorithm. Based on the model parameter calibration instructions, the system initiates a built-in optimization algorithm with the objective function of minimizing the overall difference between the simulated deformation values ​​and the measured structural deformation values ​​at all monitoring points. The optimization algorithm adjusts the parameters to be calibrated in the digital twin model in reverse. These parameters may include the elastic modulus or yield strength of steel in the material constitutive relation parameters, and may also include the equivalent spring stiffness of temporary supports in the boundary condition parameters. The adjustment process is iterative; each iteration runs the calculation of the digital twin model, generating new simulated deformation values ​​and comparing them with the measured structural deformation values, until the newly generated difference sequence meets a preset convergence criterion. The convergence criterion may be that the root mean square error of the difference sequence is less than one millimeter.

[0082] In some embodiments, the corrected digital twin model is used to generate advanced analysis results. Based on the corrected digital twin model, structural mechanics analysis is performed at the current construction stage, calculating the stress distribution cloud map, displacement response vector, and overall stability safety factor of key structural sections. The calculation of the overall stability safety factor is typically completed through eigenvalue buckling analysis. These calculation results are integrated to generate a visual diagnostic result containing a mechanical state cloud map and safety level labels. The visual diagnostic result is displayed through a 3D graphical interface, where stress levels are represented by color gradients, and safety level labels are marked next to key components in the form of "Safe," "Caution," or "Danger." Combining ambient temperature fluctuation data and on-site wind speed change data in the response parameter set, the material elastic modulus in the digital twin model is temperature-corrected, and the structural mass matrix and stiffness matrix are wind-induced dynamic characteristic corrections are performed. Based on the corrected dynamic characteristics and the calculated overall stability safety factor, the first-order second-moment method in reliability theory is used to calculate the failure probability of overall instability of the structure in the subsequent three construction steps. The failure probability is output as a value between zero and one, forming a quantified safety risk indicator.

[0083] In practical implementation, the next step is to call upon the pre-integrated intelligent analysis model for collaborative analysis. The mechanical state calculated and output by the updated digital twin model, such as stress distribution and displacement field, is input into the construction anomaly identification model along with the real-time collected structural deformation time-series data. The construction anomaly identification model internally stores a preset normal construction response pattern library, which is generated by training with a large amount of historical normal construction simulation data and monitoring data. The construction anomaly identification model uses a pattern matching algorithm to compare the currently input mechanical response pattern with all patterns in the normal construction response pattern library and calculate the pattern similarity. When the response pattern deviation of a certain monitoring point or a group of monitoring points exceeds a set threshold, the construction anomaly identification model identifies the anomaly. The identified anomaly may be classified as local buckling signs, abnormal node connections, or overload, and its possible location and time window are marked.

[0084] In practical implementation, the identified abnormal working conditions need to be further assessed for their long-term impact. The abnormal working condition characteristics identified by the construction anomaly identification model, current ambient temperature fluctuation data, and on-site wind speed change data are collectively used as input feature vectors and fed into the mechanical performance risk prediction model. The mechanical performance risk prediction model is a prediction module built based on a machine learning regression algorithm, trained and generated using historical accident data and mechanical simulation data. After receiving the input feature vector, the mechanical performance risk prediction model outputs one or more predicted values. These predicted values ​​represent the probability that the structure will experience damage or instability at a specific location within a specified future construction phase, such as within the next five construction steps. The probability value is a quantified numerical value.

[0085] Example 3: Pre-analysis of structural dynamic response and fatigue life based on a digital twin model before construction. An equivalent wind load time history is applied to the digital twin model, generated based on wind tunnel test data or standard wind spectra of the target construction site. Wind-induced dynamic time history analysis is performed to obtain stress time history curves of key structural components under the equivalent wind load time history. A cycle counting algorithm is applied to the stress time history curves to statistically obtain the number of cycles corresponding to each stress amplitude, forming a probability density distribution of stress amplitude. Based on the probability density distribution of stress amplitude and the fatigue strength curve of the material, cumulative damage theory is used to calculate the wind-induced fatigue damage and estimated fatigue life of key components at the design wind speed during construction.

[0086] Structural dynamic response and vulnerability assessment under seismic loading were conducted using a digital twin model. A refined three-dimensional finite element model of the tall steel structure was established as the core mechanical analysis model of the digital twin model. Modal analysis was performed on the refined three-dimensional finite element model to obtain the structure's natural frequencies, mode shapes, and participation coefficients. Multiple seismic ground motion time histories with different intensity characteristics were selected as inputs, and elastic time history analysis under frequent earthquakes and elastoplastic time history analysis under rare earthquakes were performed on the refined three-dimensional finite element model. Based on the results of the elastic and elastoplastic time history analyses, the overall dynamic response parameters and component damage indices of the structure under different intensities of seismic ground motion were extracted. The overall dynamic response parameters included inter-story drift angle and vertex acceleration, while the component damage indices included plastic hinge development state and section yielding degree. A two-parameter system was established with the seismic ground motion intensity index as the abscissa and the structural damage index as the ordinate. Based on statistical analysis, the seismic vulnerability curves of the structure under different damage states were obtained.

[0087] In practice, pre-construction analysis of structural dynamic response and fatigue life based on digital twin models is a crucial component of the evaluation process. The digital twin model is used to apply equivalent wind load time histories before construction simulation. These time histories are generated based on wind tunnel test data or standard wind spectra from the target construction site. For example, for a 300-meter-high television tower steel structure construction project located in a coastal area, the generation of equivalent wind load time histories relies on site-specific wind tunnel test data. This data provides samples of fluctuating wind pressure coefficients and time histories at various heights and wind directions. Based on the wind tunnel test data, equivalent wind load time histories acting on the nodes of the digital twin model are synthesized using random vibration theory. The total duration of these equivalent wind load time histories covers all wind conditions expected during construction, and their time steps are matched to the basic period of the structure to ensure analytical accuracy.

[0088] In practice, wind-induced dynamic time history analysis is performed to obtain the dynamic response of the structure. In the digital twin model, the generated equivalent wind load time history is used as the dynamic input to perform nonlinear time history analysis on the structure, considering the effects of geometric and material nonlinearity. Through wind-induced dynamic time history analysis, stress-time history curves of key structural components under equivalent wind load time history are obtained. These curves record the complete history of normal and shear stresses at critical sections of the components over time. For example, analyzing a typical section of a tower's outer frame column, the stress-time history curve shows that the stress fluctuates around the average stress level, with the amplitude of the fluctuation reflecting the pulsating characteristics of the wind load.

[0089] In some embodiments, stress time history data needs to be processed to assess fatigue damage. A cycle counting algorithm, employing the rainflow counting method, is applied to the obtained stress time history curves. This rainflow counting method identifies and statistically analyzes complete stress cycles from complex stress time histories. The number of cycles corresponding to each stress amplitude is statistically obtained, thus forming the probability density distribution of the stress amplitude of the component during the construction period. This probability density distribution describes the statistical regularity of the frequency of occurrence of different stress amplitudes. Based on the probability density distribution of the stress amplitude and the fatigue strength curve of the material, the cumulative damage theory is used to calculate the wind-induced fatigue damage and estimated fatigue life of the key component under the design wind speed during the construction period. The calculation process involves accumulating the damage of all stress cycles, and the cumulative damage degree... This can be expressed as:

[0090]

[0091] in: This indicates the cumulative fatigue damage caused by wind load. This indicates the total number of stress amplitude levels. Indicates the first The number of cycles in which the stress amplitude occurs during the time history. Indicates corresponding to the first The stress amplitude and the number of damaging cycles determined based on the material fatigue strength curve. When the cumulative damage... When the value reaches 1, it indicates that the fatigue life has been exhausted.

[0092] In practical implementation, assessing the structural dynamic response and vulnerability under seismic loading based on a digital twin model is another key analysis. A refined three-dimensional finite element model of the tall steel structure is established as the core mechanical analysis model of the digital twin model. This refined three-dimensional finite element model employs a hybrid modeling approach using shell and beam elements to accurately simulate the stress behavior of the nodal domains. Modal analysis is performed on the refined three-dimensional finite element model. The modal analysis extracts the first twenty natural frequencies, mode shapes, and mass participation factors of the structure under elastic conditions. The natural frequencies reflect the structure's inherent dynamic characteristics, and the mode shapes describe the deformation patterns of the structure at each order.

[0093] In some embodiments, the structure needs to be tested under earthquake inputs of varying intensities. Multiple earthquake time histories with different intensities are selected as inputs, drawn from an earthquake database, with peak ground acceleration covering levels from frequent to rare earthquakes. Elastic time history analysis under frequent earthquakes and elastoplastic time history analysis under rare earthquakes are performed on the refined three-dimensional finite element model. In the elastic time history analysis, the structural constitutive model remains elastic, and the elastic dynamic response of the structure is calculated. In the elastoplastic time history analysis, plasticity is introduced into the material constitutive model to simulate the nonlinear energy dissipation behavior of the structure under strong earthquakes. Based on the results of the elastic and elastoplastic time history analyses, the overall dynamic response parameters and component damage indices of the structure under earthquakes of different intensities are extracted. The overall dynamic response parameters include the inter-story drift angles of each floor and the peak ground acceleration at the structural apex. The component damage indices include the development state of plastic hinges, the yield strength of key sections, and cumulative plastic energy dissipation.

[0094] Example 4: Parameter Sensitivity Analysis and Structural Optimization Based on Seismic Vulnerability Analysis Results. A set of key parameters affecting the seismic performance of the structure is selected, including the geometric cross-sectional dimensions of components, material properties of steel, and detailed parameters of node connections. In a refined three-dimensional finite element model, one or more parameters in the key parameter set are perturbed and adjusted, and modal analysis, elastic time history analysis, and elastoplastic time history analysis are re-executed. The changes in the seismic vulnerability curve of the structure before and after parameter adjustment are compared and analyzed, the influence of each key parameter on the seismic vulnerability of the structure is quantitatively assessed, and the key parameters most sensitive to the adjustment of the vulnerability curve are identified. Based on the parameter sensitivity analysis results, weak links and potential failure modes of the structure under special seismic conditions are identified. A seismic performance improvement strategy is proposed, focusing on adjusting the cross-sectional configuration of sensitive components, enhancing the energy-dissipating structure of key nodes, and optimizing the stiffness distribution of the overall structure, generating a structural optimization design scheme that balances seismic safety and economy.

[0095] The construction and training of the mechanical performance risk prediction model involves specific steps. Historical steel structure construction accident case data, laboratory component test data, and a large amount of high-fidelity mechanical simulation data are collected to form a training dataset. Each data point in the training dataset contains input features and output labels. Input features include the geometric description of the structure at the construction stage corresponding to the data used for model training, load conditions, identified abnormal working conditions, environmental conditions, and key mechanical indicators calculated by the digital twin model. The output label indicates whether the structure will experience a specific type of damage or instability in subsequent construction steps under the stated construction stage and working conditions, and the corresponding risk level. The machine learning model is trained under supervised supervision using the training dataset. The machine learning model is a deep neural network or gradient boosting decision tree, and the training objective is to minimize the error between the model's predicted risk level and the true label. The trained machine learning model is then deployed as a mechanical performance risk prediction model, used to receive real-time input features online and output quantified risk probabilities and location predictions.

[0096] In practice, parameter sensitivity analysis and structural optimization based on seismic vulnerability analysis results is a process of in-depth design. Selecting the set of key parameters affecting the seismic performance of the structure is the first step in parameter sensitivity analysis. The selection of the key parameter set is based on engineering experience and preliminary analysis. The key parameter set typically includes the geometric dimensions of the components, the material properties of the steel, and the structural details of the joint connections. For example, for a steel-concrete composite column-beam joint with a stiffening ring plate, the key parameter set might specifically include the steel tube wall thickness, the width and thickness of the ring plate, the core concrete strength grade, the beam flange thickness, and the spacing of the stiffening ribs in the joint area. These parameters are identified as variables that significantly affect the seismic performance of the joint.

[0097] In some embodiments, the set of key parameters is adjusted in a refined three-dimensional finite element model to observe its effects. In the refined three-dimensional finite element model, one or more parameters in the set of key parameters are subjected to planned perturbation adjustments, typically within a certain range, such as varying the steel pipe wall thickness within ±20% of the design value. For each adjusted parameter combination, a complete modal analysis, elastic time history analysis, and elastoplastic time history analysis are re-executed. The changes in the seismic vulnerability curve of the structure before and after parameter adjustment are compared and analyzed. By calculating the offset of key points on the vulnerability curve, the influence of each key parameter on the seismic vulnerability of the structure is quantitatively assessed, thereby identifying the key parameters most sensitive to vulnerability curve control. Referring to Table 1, a simplified parameter sensitivity analysis result can be presented in tabular form, quantifying the impact of different parameter changes on the median value of the spectral acceleration corresponding to the "severe damage" limit state.

[0098] Table 1: Sensitivity Analysis of Key Parameters to Structural Seismic Vulnerability

[0099] Parameter name Parameter baseline value Parameter variation range "Severely damaged" state spectrum acceleration median rate of change Steel pipe wall thickness 20 mm +10% +3.2% Steel pipe wall thickness 20 mm -10% -4.8% Ring plate thickness 30 mm +10% +1.5% Ring plate thickness 30 mm -10% -2.1% Node domain stiffening rib spacing 300 mm +10% -1.8% Node domain stiffening rib spacing 300 mm -10% +2.0%

[0100] In practical implementation, the results of parametric sensitivity analysis can guide design optimization. Based on these results, weak points and potential failure modes of the structure under special seismic conditions can be identified. For example, the analysis may show that under rare earthquakes, key node areas on certain floors are regions where plastic hinges concentrate and accumulate the greatest damage, constituting potential weak points. Seismic performance improvement strategies are proposed, focusing on adjusting the cross-sectional configuration of sensitive components, enhancing energy-dissipating structures at key nodes, and optimizing the overall structural stiffness distribution. Specific measures of these strategies may include: increasing the wall thickness of steel pipes sensitive to vulnerability based on the sensitivity analysis results; adding energy-dissipating steel plates or viscous dampers to identified weak node areas; and adjusting the column cross-sectional dimensions of adjacent floors to smooth out abrupt changes in stiffness.

[0101] In practice, the construction and training of the mechanical performance risk prediction model is an independent and crucial modular process. The construction of the mechanical performance risk prediction model begins with data collection and preparation, including historical steel structure construction accident case data, laboratory component test data, and a large amount of high-fidelity mechanical simulation data, forming a training dataset. Each data sample in the training dataset contains input features and output labels. The input features are multi-dimensional vectors, including the geometric description of the structure at the construction stage corresponding to the data used for model training, load conditions, identified abnormal working conditions, environmental conditions, and key mechanical indicators calculated by the digital twin model. The output labels are defined based on expert knowledge or high-fidelity simulation results. The output label indicates whether the structure will experience a specific type of damage or instability in subsequent construction steps under the stated construction stage and working conditions, and the corresponding risk level. The risk level can be represented by discrete values, such as levels 1 to 5, to indicate the severity.

[0102] In some embodiments, the machine learning model is supervisedly trained using a prepared training dataset. The machine learning model may employ a deep neural network algorithm, and the training objective is to enable the model to accurately predict risk. The training process aims to minimize the error between the risk level predicted by the model and the true label in the training dataset. This objective is achieved by optimizing a loss function, such as the cross-entropy loss function. To measure:

[0103]

[0104] in: This represents the cross-entropy loss value. This represents the total number of samples in a training batch. This represents the total number of risk level categories. It is an indicator function, when the sample The true label is the category The value is 1 if it is true, and 0 otherwise. Indicates the model predicts samples Category The probability. The model parameters are iteratively adjusted using the backpropagation algorithm to make the loss function... The value of is continuously reduced until the model's prediction accuracy meets the preset requirements. The machine learning model, which has been trained and tested on the validation set, is deployed as a mechanical performance risk prediction model. This model is integrated into the online system to receive real-time input feature vectors from the digital twin and monitoring system, and outputs quantified risk probabilities and possible risk location predictions, providing immediate support for construction decisions.

[0105] See Figure 4This figure presents the performance evolution of a deep neural network model during the training process in a risk level prediction task. The figure uses the training epoch as the horizontal axis and simultaneously shows the trends of training loss (cross-entropy, blue curve), validation loss (cross-entropy, red dashed line), and validation set prediction accuracy (green curve, right vertical axis): The training loss continuously decreases with increasing epochs, reflecting the model's increasing fit to the training data; the validation loss generally decreases and then fluctuates slightly, with the deviation from the training loss gradually widening, suggesting a potential overfitting tendency in later stages; the validation set accuracy gradually increases and stabilizes as training progresses, eventually maintaining a high level above 90%, indicating that the model has good risk level prediction ability for unseen data. Fluctuations in validation accuracy during training can be attributed to local differences in the distribution of the validation set data or phased adjustments in model parameter updates.

[0106] Example 5: Constructing a closed-loop management mechanism for data-driven, model-validated, and construction-controlled operations. Decision support for construction process optimization is translated into specific construction adjustment instructions, including adjusting the component hoisting sequence, modifying temporary support layouts, or changing welding process parameters. These instructions are executed on-site, while a multi-dimensional sensor network collects new response parameter sets after execution. The new response parameter sets are input into a digital twin model to verify the improvement in structural mechanical performance after executing the adjustment instructions and to calculate new quantitative safety risk assessment indicators. The effectiveness of the construction adjustment instructions is evaluated by comparing the quantitative safety risk assessment indicators before and after the adjustment. If the new quantitative safety risk assessment indicators still do not meet the expected goals, a new round of collaborative analysis and decision generation is initiated until the risk indicators meet the requirements, forming a complete closed loop of assessment, decision-making, execution, and verification.

[0107] A comprehensive assessment and report on the mechanical performance of the entire construction process is generated. This integrates pre-construction fatigue life predictions based on wind loads, vulnerability assessments based on seismic effects, risk assessments based on real-time monitoring during construction, and performance verification results after optimization decisions. Based on a predetermined assessment index system, a weighted comprehensive score is applied to each integrated result, covering multiple dimensions including structural safety, construction feasibility, economy, and durability. Based on the weighted comprehensive score, a comprehensive assessment level of the mechanical performance of the steel structure building throughout the construction process is generated. All analysis process data, intermediate results, final assessment level, and key decision recommendations are summarized to form a structured mechanical performance assessment report for the entire construction process.

[0108] In practical implementation, building a closed-loop management mechanism that integrates data-driven approaches, model validation, and construction control is the core of achieving dynamic control. The decision support generated by the construction process optimization decision model needs to be transformed into executable commands on-site. This means converting the construction process optimization decision support into specific construction adjustment instructions, which must be clear and unambiguous. For example, the decision support might be "reduce the dynamic response during hoisting," while the transformed specific construction adjustment instructions would include "adjust the component hoisting sequence: postpone the hoisting of the eastern unit of segment 30 until after segment 31 is completed," "modify the temporary support layout: add a temporary support at node J-15 with a stiffness of not less than 5000 kN / m," or "change welding process parameters: adjust the upper limit of the weld interpass temperature control for the current construction section from 120 degrees Celsius to 100 degrees Celsius." These construction adjustment instructions are directly issued to the construction site supervisor through the construction management system or mobile terminals.

[0109] In practice, the execution of construction instructions and the verification of their effects form a crucial closed-loop process. At the construction site, the construction teams strictly execute the received adjustment instructions, adjusting the hoisting sequence of components, modifying the temporary support layout, or changing welding process parameters. Simultaneously, a pre-deployed multi-dimensional sensor monitoring network continuously collects a set of new response parameters generated by the structure after the execution of the adjustment instructions. This new set of parameters includes temporal data on structural deformation under the new working conditions after the instructions are executed, environmental data, such as the stress and deformation data of the support points and adjacent components after the addition of temporary supports. The new set of response parameters is synchronously input into a digital twin model. Based on the updated site conditions, the digital twin model performs mechanical analysis to verify the improvement in the structural mechanical performance after the adjustment instructions are executed, and calculates a new set of quantitative assessment indicators for safety risks, such as a new stability safety factor and failure probability.

[0110] In some embodiments, the effectiveness of control measures needs to be evaluated through quantitative comparison. Comparing the quantitative assessment indicators of safety risks before and after adjustment is a direct evaluation method. For example, comparing the changes in stress levels, maximum displacement values, and overall instability probability of key sections of the structure under the same design load conditions before and after the addition of temporary supports. The effectiveness of construction adjustment instructions is evaluated based on this comparison. If the new quantitative assessment indicators of safety risks show a significant reduction in risk and achieve the expected goals, it indicates that this round of adjustments is effective, and the closed-loop process can be temporarily concluded. If the new quantitative assessment indicators of safety risks still do not achieve the expected goals, for example, if the reduction in failure probability does not meet the requirements, the system automatically or the engineer manually initiates a new round of collaborative analysis and decision generation. The new round of processes will, based on the latest model status and monitoring data, run the abnormal condition identification, risk prediction, and optimization decision model again to generate further construction adjustment instructions until the risk indicators meet the requirements, thus forming a complete closed loop of evaluation, decision-making, execution, and verification.

[0111] In some embodiments, a comprehensive assessment and report of the mechanical performance throughout the construction process provides a conclusive document for the project. This integrates pre-construction fatigue life predictions based on wind loads, vulnerability assessments based on seismic action, risk assessments based on real-time monitoring during construction, and adjusted performance verification results based on optimization decisions. These results, derived from analysis modules at different stages, are stored in a structured data format. A weighted comprehensive score is applied to the integrated results according to a predetermined evaluation index system. This system covers multiple dimensions, including structural safety, construction feasibility, economy, and durability, with each dimension containing several specific evaluation sub-items and assigned weights. Weighted comprehensive scoring is then performed. The calculation can be expressed as:

[0112]

[0113] in: This indicates the final weighted composite score. Indicates the total number of evaluation dimensions. Indicates the first The weighting coefficients of each evaluation dimension, Indicates the first The total number of evaluation items under each dimension Indicates the first The first dimension The weight of each item Indicates the first The first dimension The original scores for each sub-item are generated. Based on the weighted comprehensive score and compared with the preset scoring level standards, a comprehensive evaluation level of the mechanical performance of the entire steel structure construction process is generated, such as "excellent", "qualified" or "requires re-inspection".

[0114] See Figure 5In the quantitative analysis of safety risks in the mechanical performance assessment of steel structure construction throughout the entire process, heatmaps visually present the distribution of safety risk levels under various risk factors at different construction stages. Specifically, the graph uses construction stages (hoisting, welding, support installation, overall assembly, and final acceptance) as the vertical dimension and risk factors (structural instability, local buckling, node failure, overload, and environmental impact) as the horizontal dimension. It quantifies the risk level of each condition using color gradients (corresponding to risk levels of 0-10) and numerical labels. The overall assembly stage has a structural instability risk level of 9, and the hoisting stage has a load overload risk level of 9, both considered high-risk conditions. The final acceptance stage, however, has a load overload risk level of only 2, placing it in the low-risk range. This risk distribution data can serve as a key input for digital twin model parameter calibration and construction process optimization decisions. For example, for the high risk of overload during the hoisting stage, the hoisting scheme can be optimized through construction adjustment instructions. Subsequently, the risk improvement effect can be verified by combining new response parameters monitored by sensors, supporting the realization of closed-loop management of construction control.

[0115] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0116] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the mechanical performance of steel structure buildings throughout the entire construction process based on digital twins, characterized in that, The method includes: A digital twin model of the construction process of a tall steel structure is established. The digital twin model integrates structural design information, construction schedule and material physical properties. The digital twin model is dynamically mapped in real time to the geometric, mechanical and environmental conditions of the actual construction site. A multi-dimensional sensor monitoring network is deployed at the construction site to collect the set of response parameters of the steel structure in real time during the construction process. The set of response parameters includes structural deformation time series data, ambient temperature fluctuation data, and on-site wind speed change data. The set of response parameters is synchronously input into the digital twin model. Based on the dynamic data fusion algorithm and the abnormal state identification algorithm, the parameters of the digital twin model are updated and the state is corrected to generate a visual diagnostic result of the construction process state and a quantitative assessment index of safety risks. The pre-integrated intelligent analysis model is invoked to perform collaborative analysis and processing on the updated digital twin model and the real-time collected response parameter set. The collaborative analysis and processing includes autonomous identification of abnormal construction conditions, quantitative prediction of potential mechanical performance risks, and decision support for construction process optimization. The decision support based on the optimized construction process generates construction instructions that are fed back to the construction site, thus constructing a closed-loop management mechanism that integrates data-driven approaches, model verification, and construction control.

2. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins as described in claim 1, characterized in that, The establishment of a digital twin model of the construction process of the tall steel structure includes: A three-dimensional geometric model of the tall steel structure is constructed based on the structural design information. The three-dimensional geometric model includes the spatial coordinates of all structural component nodes, component cross-sectional dimensions, and connection details between components. Material constitutive relations are embedded on the basis of the three-dimensional geometric model. The material constitutive relations are based on the material physical properties defined, including the elastic modulus, yield strength, ultimate strength of steel, and constitutive model parameters. The construction schedule is transformed into a load application sequence and a boundary condition change sequence in the time dimension. The load application sequence includes the application time and location of construction machinery loads, temporary support loads and self-weight loads of installed components. The boundary condition change sequence includes the temporal information of the evolution of structural constraint state with the construction stage. By integrating the three-dimensional geometric model, the material constitutive relation, the load application sequence, and the boundary condition change sequence, a construction process digital twin model with time-varying characteristics is generated. The construction process digital twin model can simulate the structural geometry and mechanical state at any construction moment.

3. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins as described in claim 2, characterized in that, The method of updating parameters and correcting the state of the digital twin model based on dynamic data fusion algorithm and abnormal state identification algorithm generates visualized diagnostic results of the construction process state and quantitative assessment indicators of safety risks, including: Extract structural deformation time series data from the response parameter set, and calculate the difference 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 difference sequence to extract the mean, variance, and trend characteristics of the difference sequence. When the mean or variance of the difference sequence exceeds a preset tolerance threshold, a model parameter correction command is triggered. According to the model parameter correction instructions, the material constitutive relation parameters or boundary condition parameters in the digital twin model are adjusted in reverse based on the optimization algorithm, so that the difference between the deformation simulation value output by the adjusted digital twin model and the measured deformation value of the structure meets the convergence criterion. Based on the corrected digital twin model, the stress distribution, displacement response and overall stability safety factor of the key sections of the structure under the current construction stage are calculated, and a visual diagnostic result including mechanical state cloud map and safety level label is generated. By combining ambient temperature fluctuation data and on-site wind speed change data, the structural dynamic characteristics are corrected. Based on the corrected dynamic characteristics and the overall stability safety factor, the failure probability of the structure in subsequent construction steps is calculated, and a quantitative safety risk index is generated.

4. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins as described in claim 3, characterized in that, The process of calling a pre-integrated intelligent analysis model to perform collaborative analysis and processing on the updated digital twin model and the set of real-time collected response parameters includes: 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. The construction anomaly identification model identifies anomalies where the response mode deviation exceeds a threshold by comparing the mechanical response mode with a preset normal construction response mode library. The anomalies include local buckling signs, abnormal node connections, or overload. The identified abnormal working condition characteristics, current ambient temperature fluctuation data, and on-site wind speed change data are input into the mechanical performance risk prediction model. The mechanical performance risk prediction model is trained and generated based on historical accident data and mechanical simulation data, and is used to predict the probability and possible location of structural damage or instability in a specified construction stage in the future. The abnormal working conditions, predicted risk probabilities, and location information are input into the construction process optimization decision model. The construction process optimization decision model is based on a multi-objective optimization algorithm and generates decision suggestions such as adjusting the construction sequence, reinforcing temporary supports, or changing the hoisting scheme to reduce the risk probabilities.

5. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins as described in claim 1, characterized in that, The method also includes pre-analysis of structural dynamic response and fatigue life based on a digital twin model before construction, including: An equivalent wind load time history is applied to the digital twin model, the equivalent wind load time history being generated based on wind tunnel test data or standard wind spectra of the target construction site; Perform wind-induced dynamic time history analysis to obtain the stress time history curves of key structural components under the equivalent wind load time history. The cycle counting algorithm is applied to the stress time history curve to obtain the number of cycles corresponding to each stress amplitude, thus forming the probability density distribution of the stress amplitude; Based on the probability density distribution of the stress amplitude and the fatigue strength curve of the material, the cumulative damage theory is used to calculate the wind-induced fatigue damage and estimated fatigue life of the key components under the design wind speed during the construction period.

6. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins as described in claim 1, characterized in that, The method also includes assessing the structural dynamic response and vulnerability under seismic loading based on a digital twin model, including: A refined three-dimensional finite element model of the tall steel structure was established as the core mechanical analysis model of the digital twin model; Modal analysis was performed on the refined three-dimensional finite element model to obtain the natural frequencies, mode shapes, and participation coefficients of the structure. Multiple ground motion time histories with different intensity characteristics were selected as inputs, and elastic time history analysis under frequent earthquakes and elastoplastic time history analysis under rare earthquakes were performed on the refined three-dimensional finite element model respectively. Based on the results of the elastic time history analysis and the elastoplastic time history analysis, the overall dynamic response parameters and component damage indices of the structure under different intensity ground motions are extracted. The overall dynamic response parameters include inter-story drift angle and vertex acceleration, and the component damage indices include plastic hinge development state and cross-sectional yielding degree. A two-parameter system with seismic ground motion intensity index as the abscissa and structural damage index as the ordinate was established. Based on statistical analysis, the seismic vulnerability curves of the structure under different damage states were obtained by fitting.

7. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins as described in claim 6, characterized in that, The method also includes parameter sensitivity analysis and structural optimization based on seismic vulnerability analysis results, including: A set of key parameters affecting the seismic performance of the structure is selected. The set of key parameters includes the geometric cross-sectional dimensions of the components, the material property parameters of the steel, and the construction details of the node connections. In the refined three-dimensional finite element model, one or more parameters in the set of key parameters are perturbed and adjusted, and the modal analysis, elastic time history analysis and elastoplastic time history analysis are re-executed; By comparing and analyzing the changes in the seismic vulnerability curves of the structure before and after parameter adjustment, the influence of each key parameter on the seismic vulnerability of the structure is quantitatively assessed, and the key parameters most sensitive to the adjustment of the vulnerability curve are identified. Based on the results of the parameter sensitivity analysis, weak links and potential failure modes of the structure under special seismic conditions are identified. A seismic performance improvement strategy is proposed, which focuses on adjusting the cross-sectional configuration of sensitive components, enhancing the energy dissipation structure of key nodes, and optimizing the stiffness distribution of the overall structure, thereby generating a structural optimization design scheme that takes into account both seismic safety and economy.

8. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins according to claim 4, characterized in that, The construction and training of the mechanical performance risk prediction model includes: Historical steel structure construction accident case data, laboratory component test data, and a large amount of high-fidelity mechanical simulation data are collected to form a training dataset. Each data point in the training dataset contains input features and output labels. The input features include the geometric morphology description of the structure at the construction stage corresponding to the data used for model training, load conditions, identified abnormal working conditions, environmental conditions, and key mechanical indicators calculated by the digital twin model. The output label indicates whether the structure suffers a specific type of damage or instability in subsequent construction steps under the construction stage and working conditions, and the corresponding risk level. The machine learning model, which is a deep neural network or a gradient boosting decision tree, is trained using the training dataset under supervision. The training objective is to minimize the error between the risk level predicted by the model and the true label. The trained machine learning model is deployed as the mechanical performance risk prediction model, which is used to receive real-time input features online and output quantified risk probability and location prediction.

9. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins according to claim 1, characterized in that, The aforementioned closed-loop management mechanism for data-driven, model-validation, and construction control includes: The decision support for optimizing the construction process is transformed into specific construction adjustment instructions, which include adjusting the hoisting sequence of components, modifying the temporary support layout scheme, or changing the welding process parameters. The construction adjustment instructions are executed at the construction site, and at the same time, the new set of response parameters after the instructions are collected through the multi-dimensional sensor monitoring network. The new set of response parameters is input into the digital twin model to verify the improvement of the structural mechanical performance after the adjustment command is executed, and to calculate the new safety risk quantitative assessment index. By comparing the quantitative assessment indicators of safety risks before and after the adjustment, the effectiveness of the construction adjustment instructions can be evaluated. If the new quantitative assessment indicators for security risks still fail to meet the expected goals, a new round of collaborative analysis and decision generation will be initiated until the risk indicators meet the requirements, forming a complete closed loop of assessment, decision-making, execution, and verification.

10. The method for evaluating the mechanical performance of steel structure buildings throughout the construction process based on digital twins according to claim 6, characterized in that, The method also includes a comprehensive assessment and report generation of the mechanical properties throughout the construction process, including: The results of fatigue life prediction based on wind load before construction, vulnerability assessment based on seismic action, risk assessment based on real-time monitoring during construction, and performance verification results after adjustment based on optimization decision-making are integrated. Based on the predetermined evaluation index system, the integrated results are weighted and comprehensively scored. The evaluation index system covers multiple dimensions such as structural safety, construction feasibility, economy and durability. Based on the weighted comprehensive score, a comprehensive evaluation level of the mechanical performance of the steel structure building throughout the entire construction process is generated; All analysis process data, intermediate results, final evaluation level, and key decision recommendations are summarized to form a structured mechanical performance evaluation report for the entire construction process.

Images