A mobile form construction monitoring method and system based on finite element simulation

Abstract

The application discloses a kind of based on finite element simulation mobile mould frame construction monitoring method and system, it is related to bridge construction monitoring and structural health monitoring field;Solve the problem of model fixed in traditional mobile mould frame construction monitoring, early warning lag, difficult to dynamically reflect the real state of structure.The present application establishes digital structure module including calibratable parameter, carries out working condition simulation to identify theoretical high-risk area and optimizes sensor layout;Carry out preposition simulation output predicted value before each process, synchronously collect measured data and calculate standardized deviation sequence during construction;Based on time domain feature extraction and multi-index fusion, dynamic safety evaluation and early warning are carried out;Using historical deviation data, the model parameters are calibrated and updated online by bayesian variational inference, realize the closed loop of simulation, monitoring, early warning, model optimization, significantly improve the safety monitoring precision and real-time of mobile mould frame construction process.

Description

Technical Field

[0001] This invention belongs to the field of bridge construction monitoring and structural health monitoring, specifically a method and system for monitoring the construction of mobile formwork based on finite element simulation. Background Technology

[0002] Mobile formwork, as an indispensable large-scale temporary construction equipment in modern bridge engineering, plays a crucial role in the overall quality and progress of the project due to its structural safety and stability. Because its construction involves complex conditions such as high altitudes, large spans, and variable loads, the structural stress state exhibits significant time-varying and concealed characteristics. Traditional methods relying on experience and partial point-based monitoring are insufficient for a comprehensive and accurate assessment of its overall safety status. With the rapid development of computer simulation and IoT monitoring technologies, deeply integrating high-precision finite element simulation with real-time on-site monitoring data to achieve predictability, quantification, and controllability of the construction process has become an important direction for improving the digital construction and intelligent management of large temporary structures. Therefore, it is necessary to develop a monitoring method that deeply integrates simulation and field measurement, enabling dynamic updates and evaluation, to provide forward-looking early warnings and decision support for the safe construction of mobile formwork.

[0003] The following problems exist in the existing technology:

[0004] Traditional methods rely on fixed finite element model parameters, which cannot reflect uncertainties such as material properties and connection status during construction, resulting in inherent deviations between simulation prediction results and actual structural mechanical responses.

[0005] The placement of physical sensors mainly relies on engineers' experience and has not been optimized for the mechanical properties of specific structures, which can easily lead to insufficient monitoring of actual high-risk areas or waste of resources.

[0006] Traditional assessments are mostly conducted after an anomaly occurs, relying solely on a simple comparison between measured values ​​and fixed thresholds, lacking prior prediction and comprehensive dynamic analysis, resulting in low timeliness and accuracy of early warnings;

[0007] The lack of a feedback loop between the measured data generated by the monitoring system and the simulation model means that the model cannot use measured deviations to self-correct and improve accuracy throughout the entire construction period. Summary of the Invention

[0008] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a method and system for monitoring the construction of a moving formwork based on finite element simulation, which is used to solve the above-mentioned technical problem.

[0009] The first aspect of this invention provides a method for monitoring the construction of a moving formwork based on finite element simulation, comprising the following steps:

[0010] S1: Establish a parametric finite element model corresponding to the physical moving frame, wherein the parametric finite element model includes digital structural modules with calibrable parameters;

[0011] S2: Based on the digital structure module, the target construction load condition is simulated and analyzed. Based on the mechanical cloud map generated by the simulation analysis, the corresponding theoretical high-risk area in the physical moving formwork is identified, and the optimized layout scheme of the physical sensors is planned accordingly.

[0012] S3: Before each preset construction procedure begins, the procedure parameters are input into the parameterized finite element model for pre-simulation, and the mechanical response prediction dataset of the physical moving formwork under the preset construction procedure is output.

[0013] S4: When performing the preset construction procedure, obtain the measured response values ​​of the physical sensors installed according to the optimized layout scheme, and compare the measured response values ​​with the corresponding mechanical response prediction dataset in real time to generate a response deviation value time series.

[0014] S5: Based on the measured response value and response deviation value time series, the series is processed to extract time domain statistical features, and then a dynamic safety assessment is performed on the physical moving frame, and an early warning signal is output according to the assessment results;

[0015] S6: Use the response deviation value to perform reverse calibration on the calibrable parameters in the digital structure module to update the parameterized finite element model.

[0016] Preferably, step S1 includes the following steps:

[0017] Construct a three-dimensional geometric model corresponding to the physical moving formwork. The three-dimensional geometric model includes the geometric entities of the main beam system, guide beam system, support leg system, and key connection nodes.

[0018] The three-dimensional geometric model is meshed using finite element methods to generate a digital structural module composed of shell elements, beam elements, and rod elements.

[0019] The physical properties of interest in the impact mechanical response in the digital structure module are defined as a calibrable parameter set. , where each parameter Having a prior probability distribution ;

[0020] The caliable parameter set It includes at least one of the following parameter categories: material parameters, boundary condition parameters, and connection stiffness parameters; wherein, the material parameters include at least the equivalent elastic modulus, Poisson's ratio, and yield strength; the boundary condition parameters include at least the contact support stiffness between the outrigger and the pier, the shear stiffness of the pin connection, and the initial clearance between the connectors; and the connection stiffness parameters include at least the shear stiffness of the bolt connection and the coefficient of friction of the contact surface.

[0021] The digital structure module and its calibrable parameter set will be used. The structure formed by the collective elements is defined as the parametric finite element model.

[0022] Based on the initial values ​​of the calibrable parameter set, an initial verification simulation is performed on the digital structure module. The simulation results are compared with the design pre-camber or initial measured data. If the deviation is less than a preset threshold, the parameterized finite element model is determined to be reliable; otherwise, the initial parameter values ​​are adjusted and the verification simulation is performed again.

[0023] Preferably, step S2 includes the following steps:

[0024] Define at least one target construction load condition, and apply the corresponding load and boundary conditions to the digital structure module. Perform static finite element simulation calculations to obtain the stress and displacement response data of the digital structure module under the target construction load condition.

[0025] Based on the stress and displacement response data, stress field and displacement field distribution data of the digital structure module are generated, and further processed to generate Mises stress distribution cloud map and displacement distribution cloud map of the digital structure module under the target construction load condition.

[0026] Based on preset identification criteria, the system automatically analyzes the Mises stress distribution cloud map and displacement distribution cloud map to identify the corresponding theoretical high-risk areas in the physical moving formwork.

[0027] The preset identification criteria include high-stress area identification criteria and large deformation area identification criteria;

[0028] The high-stress zone identification criterion refers to identifying all Mises stress values ​​in the stress distribution contour map. satisfy The element that meets this condition is defined as the continuous structural part corresponding to the element on the physical moving formwork as a high-stress zone; stress threshold According to the material allowable of the physical moving formwork Compared with the preset importance coefficient To be determined jointly;

[0029] The criterion for identifying large deformation zones refers to identifying the absolute values ​​of all displacements in the displacement distribution contour map. satisfy For nodes that meet this condition, the corresponding continuous structural parts on the physical moving formwork are defined as large deformation zones; displacement threshold. Based on the span of the physical moving frame Compared with the preset ratio coefficient To be determined jointly;

[0030] Based on theoretical high-risk areas, an optimized deployment scheme for physical sensors is generated; wherein, the theoretical high-risk areas are a set of high-stress areas and large-deformation areas identified by preset identification criteria; the generation of the optimized deployment scheme is based on preset optimization criteria for coverage, sensitivity and redundancy.

[0031] Preferably, step S3 includes the following steps:

[0032] The continuous construction process is broken down into multiple sequentially executed pre-set construction procedures. Each pre-set construction procedure has a unique procedure code, a clear start and end state, load conditions, and structural configuration.

[0033] Before each preset construction procedure begins, the procedure parameters corresponding to that procedure are extracted. The procedure parameters include at least the load increment, the boundary condition change matrix, and the structural configuration update parameters.

[0034] The process parameters are input into the parameterized finite element model, and the mechanical state of the model at the end of the previous preset construction process is used as the initial condition to perform the pre-static finite element simulation calculation of the current process.

[0035] After the preliminary static simulation calculations are completed and the stress and displacement field distribution data are generated, all the data for installing the physical sensors, determined according to the optimized layout scheme, are acquired. The spatial coordinates of each planned measuring point are used; based on these spatial coordinates, the mechanical response at each planned measuring point is extracted from the stress and displacement field distribution data using interpolation methods; thus forming a mechanical response prediction dataset for this process that corresponds one-to-one with all planned measuring points. ,in, To plan the total number of measurement points, For the first Predicted values ​​for each planned measurement point ; The index of the current process indicates the number of steps. The preset construction procedures have a value range of 1. ,in This represents the total number of pre-set procedures within the construction period.

[0036] For each predicted value in the mechanical response prediction dataset Assess its uncertainty and provide a confidence interval, which is expressed as follows: ,in, The standard deviation of the uncertainty of the predicted value is... The coefficient is determined based on the target confidence level.

[0037] Preferably, step S4 includes the following steps:

[0038] At a frequency no lower than the preset sampling frequency The system synchronously acquires the output signals of all physical sensors installed according to the optimized deployment scheme using a data acquisition device. This synchronous acquisition is triggered by a unified clock on the data acquisition device. The synchronous acquisition process generates a series of discrete, equally timed sampling moments, denoted as... ,in , The start time of data collection The sampling interval is k, where k = 0, 1, 2, ... represents the sampling point number.

[0039] The raw physical sensor data collected is preprocessed, including at least outlier removal, digital filtering and noise reduction, engineering unit conversion, and temperature compensation, at each sampling time. Forming the measured response value vector ;

[0040] The measured response value vector The mechanical response prediction dataset after spatiotemporal alignment Real-time comparison is performed to obtain the residual between the measured values ​​and the predicted values ​​of each planned measurement point;

[0041] Based on the residuals between the measured and predicted values ​​of each planning measurement point, the response deviation value of each planning measurement point is obtained by calculating the standardized confidence bias index, where the i-th Each planned measurement point is located at Response deviation value at time The calculation formula is:

[0042]

[0043] in, and The first Each planned measurement point is at time [time] Measured and predicted values ​​after spatiotemporal alignment; For the first Measurement noise variance of physical sensors and their signal acquisition links installed at each planned measurement point; For parametric finite element model for the first Each planned measurement point is at time [time] Predicted value The variance of the forecast uncertainty;

[0044] Based on all planned measurement points at a series of discrete sampling times Calculated Generate a discrete time series for each measurement point, denoted as the response deviation value time series. .

[0045] Preferably, step S5 includes the following steps:

[0046] Define a sliding time window Through response deviation value time series In the sliding time window The calculations are performed internally to extract the time-domain statistical characteristics of each planned measurement point, including the window average deviation index. Deviation trend coefficient Among them, the window average deviation index , For measuring points At a historical moment The standardized confidence bias index, By analyzing the time series of response deviation values ​​within the time window Linear regression was performed on the data points to obtain the deviation trend coefficient. .

[0047] Based on the time-domain statistical characteristics of each planned measurement point and the measured response value, a comprehensive index for evaluating the overall safety status of the physical moving formwork is calculated, which includes the global comprehensive deviation index, the measured response safety index, the deviation significance index, and the trend hazard index.

[0048] Global comprehensive deviation index The calculation formula is:

[0049]

[0050] in, For the first Structural importance weights for each measurement point; It is the order of the norm;

[0051] Actual Response Safety Index ,in For the first Safety threshold for each measuring point;

[0052] Deviation significance index ,in This is the preset critical deviation threshold;

[0053] Trend Risk Index ,in, This is a set of key structural measurement points. As a trend risk indicator, This is the deviation trend coefficient;

[0054] Based on the measured response safety index, deviation significance index, and trend hazard index, a comprehensive risk index is calculated. ;

[0055] Comprehensive risk index It performs real-time matching with preset multi-level risk threshold ranges, triggers corresponding level of early warning signals based on the matching results, outputs early warning signals, and generates a structured early warning report that includes risk level, key anomaly measurement point location, potential cause analysis, and matching response plan.

[0056] Preferably, step S6 includes the following steps:

[0057] Obtain the calibrable parameter vector in the digital structure module As the object to be calibrated, each parameter is obtained simultaneously. Prior probability distribution ,in, ;

[0058] Collect monitoring data corresponding to multiple completed pre-set construction procedures from historical construction cycles to form a calibration dataset. ,in, For the first The measured response value vector of each process, For the corresponding mechanical response prediction dataset, For the response deviation value vector, For the first The planned measurement point is at the first Average deviation index of each process window , This represents the total number of processes that have been completed.

[0059] Within the Bayesian inference framework, the calibration dataset is utilized. Calculate the parameter vector posterior probability distribution Its calculation is performed by maximizing the lower bound of evidence. The problem is solved, and the lower bound of the evidence is determined. The expression is:

[0060]

[0061] in, This is a variational distribution used to approximate the posterior distribution; Let be the likelihood function, representing the likelihood given parameters. The observed response deviation vector The probability of; For parameters Prior probability distribution ;

[0062] The adjoint method is used to calculate the gradients required in the variational inference process. To address the high computational cost of the parameterized finite element model; to solve the high computational cost problem; and to find the maximizing lower bound of evidence using a gradient optimization algorithm. The problem is to obtain the optimal variational distribution. and average them As parameter calibration value Use the calibration value Update the corresponding parameters in the digital structure module to form a new version of the parameterized finite element model, and evaluate the prediction error of the calibrated parameterized finite element model on the validation dataset, and calculate the degree of reduction of its root mean square error relative to before calibration.

[0063] An online sequential update strategy is adopted; when new monitoring data is obtained... At that time, based on the current posterior distribution Likelihood with new data The posterior distribution of the parameters is updated recursively; the update rule is based on the idea of ​​exponentially weighted moving average, and is approximately expressed as: ,in, It is a forgetting factor.

[0064] A second aspect of the present invention provides a mobile formwork construction monitoring system based on finite element simulation, comprising the following modules:

[0065] Parametric Finite Element Model Establishment Module: Establishes a parametric finite element model corresponding to the physical moving mold frame. The parametric finite element model includes a digital structure module with calibrable parameters.

[0066] Simulation analysis and deployment planning module: Based on the digital structure module, the target construction load condition is simulated and analyzed. According to the mechanical cloud map generated by the simulation analysis, the corresponding theoretical high-risk area in the physical moving formwork is identified, and the optimized deployment scheme of the physical sensors is planned accordingly.

[0067] Pre-simulation and prediction module: Before the start of each preset construction procedure, the procedure parameters are input into the parameterized finite element model for pre-simulation, and the mechanical response prediction dataset of the physical moving formwork under the preset construction procedure is output.

[0068] Data acquisition and comparison module: When executing the preset construction procedure, it acquires the measured response values ​​of the physical sensors installed according to the optimized layout scheme, and compares the measured response values ​​with the corresponding mechanical response prediction dataset in real time to generate a response deviation value time series;

[0069] Safety assessment and early warning module: Based on the time series of the measured response value and response deviation value, the module processes the sequence to extract time-domain statistical features, and then performs a dynamic safety assessment of the physical moving frame, and outputs an early warning signal based on the assessment results;

[0070] Model reverse calibration module: Uses the response deviation value to reverse calibrate the calibrable parameters in the digital structure module to update the parameterized finite element model.

[0071] Compared with the prior art, the beneficial effects of the present invention are:

[0072] This invention establishes a parametric finite element model containing a calibrable parameter set, and performs reverse calibration and online updates based on monitoring data, enabling the model to continuously approximate the real mechanical properties of the physical structure, thus significantly improving the accuracy and reliability of simulation prediction throughout the entire construction cycle.

[0073] This invention identifies high-risk areas based on finite element simulation theory and generates optimized deployment schemes accordingly. It realizes the transformation of sensor deployment from "experience-driven" to "mechanical mechanism-driven", ensuring that the monitoring network accurately covers the most dangerous parts and greatly improving monitoring efficiency and effectiveness.

[0074] This invention provides a prediction benchmark through pre-simulation, generates a standardized deviation sequence through real-time comparison, and then integrates multi-dimensional indicators for dynamic security assessment and graded early warning. This method transforms passive alarm into proactive early warning and can comprehensively quantify overall risk, significantly improving the foresight and decision support capabilities of security monitoring.

[0075] This invention constructs a feedback loop of "monitoring-comparison-calibration" by reverse calibration of model parameters through response deviation values; this enables the system to have self-learning and self-evolution capabilities, and the model accuracy continues to improve with the construction process, forming a virtuous cycle of mutual reinforcement between monitoring and simulation.

[0076] This invention achieves the standardization and efficient integration of industry-education integration data by collecting and preprocessing data from diverse and heterogeneous data sources in real time.

[0077] This invention achieves real-time mapping and dynamic weight adjustment of relationships between industry and education entities by constructing an ontology model and a dynamically evolving knowledge graph of the core elements of industry-education integration.

[0078] This invention achieves a scientific assessment of the degree of matching between educational courses and industry needs through quantitative analysis of industry-education integration based on knowledge graphs.

[0079] This invention provides an intelligent recommendation service for course optimization schemes based on quantitative analysis through intelligent application and service modules, and continuously optimizes based on feedback data, forming a closed-loop optimization mechanism. Attached Figure Description

[0080] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0081] Figure 2 This is a schematic diagram of the module flow of the present invention. Detailed Implementation

[0082] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0083] Please see Figure 1 This invention relates to a method for monitoring the construction of a moving formwork based on finite element simulation, comprising the following steps:

[0084] S1: Establish a parametric finite element model corresponding to the physical moving frame, wherein the parametric finite element model includes digital structural modules with calibrable parameters;

[0085] S2: Based on the digital structure module, the target construction load condition is simulated and analyzed. Based on the mechanical cloud map generated by the simulation analysis, the corresponding theoretical high-risk area in the physical moving formwork is identified, and the optimized layout scheme of the physical sensors is planned accordingly.

[0086] S3: Before each preset construction procedure begins, the procedure parameters are input into the parameterized finite element model for pre-simulation, and the mechanical response prediction dataset of the physical moving formwork under the preset construction procedure is output.

[0087] S4: When performing the preset construction procedure, obtain the measured response values ​​of the physical sensors installed according to the optimized layout scheme, and compare the measured response values ​​with the corresponding mechanical response prediction dataset in real time to generate a response deviation value time series.

[0088] S5: Based on the measured response value and response deviation value time series, the series is processed to extract time domain statistical features, and then a dynamic safety assessment is performed on the physical moving frame, and an early warning signal is output according to the assessment results;

[0089] S6: Use the response deviation value to perform reverse calibration on the calibrable parameters in the digital structure module to update the parameterized finite element model.

[0090] Specifically, a parametric finite element model containing calibrable parameters is established based on the actual design of the physical mobile formwork. This model is used to simulate and analyze different construction load conditions, identifying theoretically high-risk areas and planning an optimized deployment scheme for physical sensors. Before each construction process begins, a pre-simulation is performed based on the model to generate a mechanical response prediction dataset. During construction, measured data from the sensors is acquired and compared in real time with the predicted values ​​from the mechanical response prediction dataset, generating a time series reflecting the deviation between the two. Statistical features are extracted based on this deviation series to conduct dynamic safety assessments and early warnings for the physical mobile formwork. Continuous monitoring of deviation data is used to reverse-calibrate key parameters in the parametric finite element model, updating the model to gradually approximate the actual structural state, thus forming a closed-loop intelligent monitoring process of "simulation prediction - real-time monitoring - assessment and early warning - model update".

[0091] In one embodiment of the present invention, step S1 includes the following steps:

[0092] Construct a three-dimensional geometric model corresponding to the physical moving formwork. The three-dimensional geometric model includes the geometric entities of the main beam system, guide beam system, support leg system, and key connection nodes.

[0093] The three-dimensional geometric model is meshed using finite element methods to generate a digital structural module composed of shell elements, beam elements, and rod elements.

[0094] The physical properties of interest in the impact mechanical response in the digital structure module are defined as a calibrable parameter set. , where each parameter Having a prior probability distribution ;

[0095] The caliable parameter set It includes at least one of the following parameter categories: material parameters, boundary condition parameters, and connection stiffness parameters; wherein, the material parameters include at least the equivalent elastic modulus, Poisson's ratio, and yield strength; the boundary condition parameters include at least the contact support stiffness between the outrigger and the pier, the shear stiffness of the pin connection, and the initial clearance between the connectors; and the connection stiffness parameters include at least the shear stiffness of the bolt connection and the coefficient of friction of the contact surface.

[0096] The digital structure module and its calibrable parameter set will be used. The structure formed by the collective elements is defined as the parametric finite element model.

[0097] Based on the initial values ​​of the calibrable parameter set, an initial verification simulation is performed on the digital structure module. The simulation results are compared with the design pre-camber or initial measured data. If the deviation is less than a preset threshold, the parameterized finite element model is determined to be reliable; otherwise, the initial parameter values ​​are adjusted and the verification simulation is performed again.

[0098] Specifically, based on detailed design drawings, computer-aided design (CAD) models, or 3D laser scanning point cloud data of the physical moving formwork, the 3D geometric model of the moving formwork is reconstructed with high precision using finite element analysis software (such as ANSYS, ABAQUS, MIDAS FEA, or ADINA) and its geometric modeling and preprocessing functions. This 3D geometric model fully covers all major load-bearing systems and force transmission paths, including the main beam system, guide beam system, leg system, and key connection nodes. The main beam system typically consists of two box-section main beams on the left and right, which are the main bending-bearing components. The guide beam system is installed at the front end of the main beam to provide temporary support when the formwork passes through holes. The leg system includes the front leg, middle leg, rear leg, and their associated shifting trolleys and support structures, which are responsible for transferring the superstructure load to the piers. Key connection nodes refer to critical force transmission parts such as the anchor points of the hangers (connecting the main beam and the formwork system), the connecting flanges between the legs and the main beam, and the splicing plates between the main beam segments. Their detailed geometric entities must be established.

[0099] Finite element meshing is performed on a three-dimensional geometric model, which involves discretizing it into a finite number of interconnected small elements to generate a finite element model that can be used for numerical mechanics calculations. This is called a digital structural module. The geometric parts in the digital structural module that have independent mechanical properties (such as material and cross-section) and can be clearly identified, and are used to bear and transfer loads, are defined as structural components, or simply components. For example, a complete box-section of a main beam or an independent connecting plate can be regarded as an independent component. Meshing requires a scientific selection of element types based on the mechanical properties and computational accuracy requirements of the main beam, guide beam, legs, and connecting components. At a minimum, shell elements, beam elements, and rod elements are essential. Shell elements are used to simulate box-type plates and connecting plates of the main beam and guide beam, efficiently and accurately calculating bending, shear, and membrane stresses in shell structures, making them the primary choice for simulating thin-walled structures. Beam elements are used to simulate secondary members such as horizontal and horizontal bracing, as well as stiffeners. Based on beam theory, they are suitable for rod structures where axial force and bending moment are the main internal forces. Rod elements are used to simulate components that only bear axial force, such as some hangers or precision-rolled threaded steel bars. During finite element mesh generation, local refinement is performed in areas with expected large stress gradients (such as near critical connection nodes or at abrupt changes in cross-section) to ensure sufficient computational accuracy in the mechanical contour maps of subsequent simulation analyses. The final generated set, containing all elements, nodes, and their connections, constitutes the digital structure module. Nodes are vertices or control points automatically generated during mesh generation to define the element geometry and connection relationships.

[0100] In a digital structural module, those physical properties that significantly affect the overall mechanical response but whose true values ​​are difficult to determine precisely during the design phase or change due to variations in construction, materials, or connection conditions are defined as calibrable parameters; these parameters constitute a set, denoted as the calibrable parameter set. .

[0101] Based on the Bayesian inference framework, for each parameter Assign a prior probability distribution The purpose is to provide a quantitative initial estimate of the uncertainty of each parameter based on existing knowledge in the field, in the absence of a large amount of measured data. The existing knowledge includes design specifications, material certificates, engineering experience, etc., which are usually derived from publicly available industry design specifications (such as the "General Specifications for Highway Bridge and Culvert Design"), product quality certificates (material certificates) provided by material manufacturers, and publicly available measured data of similar projects and engineering experience recorded in literature. For example, the prior probability distribution of the material parameter equivalent elastic modulus can be set based on the design value of the material grade, such as assuming that it follows a normal distribution with a mean of 210 GPa and a standard deviation of 5 GPa. The parameters in the calibrable parameter set must be selected from at least one of the following categories: material parameters, boundary condition parameters, and connection stiffness parameters. Material parameters include equivalent elastic modulus (comprehensively reflecting material constitutive properties, welding residual stress, etc.), Poisson's ratio, yield strength, and density. Boundary condition parameters include contact support stiffness between the outrigger and the pier pad (simulating non-ideal hinged or fixed connections), shear stiffness of pin connections, and initial clearance between connectors. Connection stiffness parameters include bolted connection shear stiffness and friction coefficient between contact surfaces. Material property parameters refer to the mechanical and physical properties of the main materials (such as steel) constituting the physical moving formwork structure. Their standard or design values ​​are derived from the product quality certificate (material certificate) provided by the material manufacturer and the industry standard corresponding to the material grade specified in the project design documents. The actual values ​​of boundary condition parameters and connection stiffness parameters are significantly affected by construction factors such as installation process and contact state, making them difficult to obtain accurately directly from the design documents. Therefore, when initially constructing the model, its parameter values ​​are set based on conventional engineering assumptions (such as simplifying the connection to an ideal hinge or fixed connection) or empirical formulas.

[0102] Digital structure modules and calibrable parameter sets The combination of physical properties and their uncertainties constitutes a complete, parameterized, and computable mechanical model, known as a parametric finite element model. The core characteristic of this parametric finite element model is that its mechanical response output (such as stress and displacement) explicitly depends on a calibrable parameter set. The values ​​of are taken; the set of elements in a parametric finite element model usually belongs to a specific component. After assigning initial values ​​to all calibrable parameters (usually the mean of their prior probability distribution), the baseline reliability of the initial parametric finite element model is verified.

[0103] In finite element analysis software, a gravity load is applied to the digital structure module, and the software's solver is invoked to perform static analysis. This static analysis refers to solving for the equilibrium state of the structure under static loads; its core is establishing and solving for the global stiffness matrix. Node displacement vector and node load vector The system control equations are as follows: Among them, the total stiffness matrix It is composed of the element stiffness characteristics, material constitutive relations, and boundary conditions of the digital structure module; load vector It is mainly composed of gravity loads. The solver solves the equation numerically, obtaining the displacement vectors of all nodes. ; where nodes are the vertices or connection points of elements in a finite element mesh, and are the basic calculation locations for physical quantities such as displacement and force.

[0104] After the static analysis is completed, the post-processing function of the finite element analysis software is used to extract the preset key control sections (such as mid-span, ...) on the main beam. Across Crossing, among them The displacement results for the nodes corresponding to the main beam span are given in the form of displacement components of each node in the global coordinate system. For any node... Its vertical deflection Defined as the vertical displacement component of the node (usually the Z-axis of the global coordinate system). The absolute value, that is: By obtaining the vertical deflection values ​​of a series of control section nodes on the main beam. This can be used to plot the deflection curve of the main beam, where the value with the largest absolute value is the maximum deflection value. : .

[0105] The calculated maximum deflection value The deflection distribution curve was compared with the benchmark. The benchmark was selected according to the following priority order: if the physical movable formwork has been assembled on site, the initial measured deflection value obtained by measuring instruments such as a total station should be used first. If the initial measured data is unavailable, the design pre-camber value provided by the design unit of the physical moving formwork shall be used. As a benchmark, a quantitative judgment threshold is set. This threshold represents the maximum allowable relative deviation between the calculated deflection and the benchmark. Its value is based on engineering experience and reasonable requirements for the accuracy of the initial model. For example, in the initial model validation stage, to balance accuracy requirements and model complexity, a threshold can often be set. This threshold can be adjusted according to the specific project's safety level, monitoring requirements, or contract specifications. (Definition reference value) Its value is or If satisfied If the trend of deflection along the beam length is consistent with the benchmark, then the initial parametric finite element model is considered reliable; if any condition is not met, the geometric model, boundary conditions, or initial parameter values ​​need to be checked and re-verified.

[0106] In one embodiment of the present invention, step S2 includes the following steps:

[0107] Define at least one target construction load condition, and apply the corresponding load and boundary conditions to the digital structure module. Perform static finite element simulation calculations to obtain the stress and displacement response data of the digital structure module under the target construction load condition.

[0108] Based on the stress and displacement response data, stress field and displacement field distribution data of the digital structure module are generated, and further processed to generate Mises stress distribution cloud map and displacement distribution cloud map of the digital structure module under the target construction load condition.

[0109] Based on preset identification criteria, the system automatically analyzes the Mises stress distribution cloud map and displacement distribution cloud map to identify the corresponding theoretical high-risk areas in the physical moving formwork.

[0110] The preset identification criteria include high-stress area identification criteria and large deformation area identification criteria;

[0111] The high-stress zone identification criterion refers to identifying all Mises stress values ​​in the stress distribution contour map. satisfy The element that meets this condition is defined as the continuous structural part corresponding to the element on the physical moving formwork as a high-stress zone; stress threshold According to the material allowable of the physical moving formwork Compared with the preset importance coefficient To be determined jointly;

[0112] The criterion for identifying large deformation zones refers to identifying the absolute values ​​of all displacements in the displacement distribution contour map. satisfy For nodes that meet this condition, the corresponding continuous structural parts on the physical moving formwork are defined as large deformation zones; displacement threshold. Based on the span of the physical moving frame Compared with the preset ratio coefficient To be determined jointly;

[0113] Based on theoretical high-risk areas, an optimized deployment scheme for physical sensors is generated; wherein, the theoretical high-risk areas are a set of high-stress areas and large-deformation areas identified by preset identification criteria; the generation of the optimized deployment scheme is based on preset optimization criteria for coverage, sensitivity and redundancy.

[0114] Specifically, in this implementation method, the load and working condition parameters include concrete unit weight, maximum design pouring volume, preload grading, and dynamic amplification factor, etc., and their specific values ​​and implementation methods are derived from the approved "Construction Organization Design" document and relevant construction technical specifications. Safety and judgment parameters include allowable stress of structural materials and allowable deformation limits (such as 1 / 1000 of the span), etc., which are derived from structural design calculations and mandatory industry design specifications (such as the "General Specifications for Highway Bridge and Culvert Design" (JTG D60)).

[0115] According to the construction organization design documents, all load states that are planned to occur and require close attention during the entire construction period for the physically movable formwork are summarized and defined as a series of target construction load conditions. These conditions should comprehensively cover the most unfavorable design conditions and possible asymmetric operation situations. Typical load conditions include, but are not limited to, standard full-load pouring conditions, asymmetric off-center load conditions, graded preloading test conditions, and formwork movement (through-hole) conditions.

[0116] The standard full-load pouring condition specifically simulates the state when the maximum designed concrete volume is poured. The concrete load is applied to the corresponding finite element elements of the main beam and bottom formwork in the form of distributed surface loads, based on the zonal arrangement of the formwork system (including bottom formwork, side formwork, and flange formwork). Specifically, the total weight of each pouring zone is calculated based on its volume and concrete density, and the surface pressure is calculated based on the support area of ​​the formwork in that zone, then applied to the corresponding shell element or surface in the digital structure module. The asymmetric eccentric loading condition specifically simulates lateral or longitudinal eccentric loads caused by the concrete pouring sequence (e.g., pouring one side of the web first) or equipment (e.g., pump truck) offset. The load is applied to the corresponding elements or nodes of the model in the form of asymmetrically distributed surface pressure or concentrated force, according to the eccentric load area and weight specified in the construction organization design. The graded preloading test condition specifically simulates the loading process of the preloading test before construction. It typically includes graded loading states of 60%, 100%, and 120% of the design load. The load is applied to the corresponding area of ​​the digital structure module in the form of equivalent nodal forces or uniformly distributed pressures, based on the stacking position and weight of the counterweights (or water bags) in the preloading test scheme. The formwork walking (through-hole) condition specifically simulates the longitudinal movement of the physical formwork to the next hole position after demolding. It primarily considers the formwork's self-weight, and may also consider the dynamic amplification factor caused by starting and braking.

[0117] In the digital structure module of the parametric finite element model, based on the description of each working condition, the corresponding loads (such as force, pressure, inertial force) and boundary conditions (such as the application and removal of leg constraints) are precisely applied to the corresponding nodes, element surfaces or geometric entities of the parametric finite element model through the preprocessing module of the finite element software.

[0118] The solver module of the finite element method software is invoked to perform static finite element simulation calculations for each target construction load condition. The solver automatically assembles the overall stiffness matrix based on the element properties (such as element type and geometry), material constitutive relations (such as elastic modulus and Poisson's ratio), and boundary conditions from the digital structure module. and load vector And solve the system of linear equations. To obtain the displacement field Based on displacement field The solver automatically calculates the strain field and stress tensor of all elements based on the geometric equations (strain-displacement relationship) and the physical equations (constitutive relations), and derives from this a scalar quantity used to evaluate the yield strength of the material, namely the Mises equivalent stress. In this process, element stresses are typically calculated first at numerical integration points within the element (i.e., Gaussian integration points, which are specific locations within the element set for accurate calculation of the element stiffness matrix and stress). After the calculation is complete, the solver generates a standard format binary or text result file (e.g., the .rst file in ANSYS software). This file fully stores the displacements of all nodes (i.e., the vertices or connection points of elements in the finite element mesh) of the entire model (i.e., the entire digital structure module), and the stresses of all elements (including each stress component and Mises equivalent stress). Response data, such as stress field and displacement field distribution data, refer to the set of mechanical response data stored in the results file, covering all spatial locations of the entire digital structure module; displacement field distribution data includes the three translational degrees of freedom displacement components of all nodes in the global coordinate system; stress field distribution data typically includes the six stress tensor components at all element integration points or nodes, as well as the Mises equivalent stress calculated therefrom.

[0119] Based on stress and displacement field distribution data, intuitive two-dimensional or three-dimensional color cloud maps are generated through the post-processing module of finite element software. Key cloud maps include Von Mises stress distribution cloud maps and displacement distribution cloud maps. The Von Mises stress distribution cloud map is generated based on the Von Mises stress calculation results of the parametric finite element model on all elements. The Von Mises equivalent stress comprehensively reflects the multi-directional stress state and can intuitively show the yielding trend and stress concentration degree of the physical moving formwork structure. The displacement distribution cloud map is generated based on the displacement calculation results of the parametric finite element model on all nodes, showing the overall deformation shape of the physical moving formwork structure under load, especially the vertical deflection and possible lateral displacement of the main beam.

[0120] The Mises stress distribution cloud map and displacement distribution cloud map are scanned and analyzed. Based on the preset algorithm and criteria, the corresponding theoretical high-risk areas on the physical moving formwork are identified. The identification process adopts multi-index fusion judgment, and the main criteria include the high stress area identification criterion and the large deformation area identification criterion.

[0121] The identification process for the high-stress zone identification criterion is as follows: set a stress threshold. This threshold is based on the allowable stress of the main structural material of the physical moving formwork. Multiply by a preset importance coefficient To determine, that is ;in, It is a safety factor less than 1 (e.g., 0.7) used to provide an early warning before the stress reaches the material's allowable stress (i.e., the design-allowed stress limit). By calling the finite element software's post-processing application programming interface (API), it iterates through all elements in the parametric finite element model and reads the average Mises stress value at each element's integration point or node from the result file generated by the static finite element simulation calculation. To satisfy all The elements are marked. In the three-dimensional physical space, by calculating the spatial distance between the centroids of the marked elements, all elements whose centroid distance is less than the set tolerance (e.g., 0.5 meters) are grouped into a cluster. The actual structural part corresponding to each cluster (such as a section of the lower flange of a main beam or a stiffening rib area) is defined as an independent high-stress zone.

[0122] The identification process for the large deformation zone identification criterion is as follows: set a displacement threshold. This threshold is based on the main span of the physical moving formwork. Multiplied by a preset scaling factor To determine, that is .in, This reflects the requirement for deformation sensitivity (e.g., taking 1 / 1000). By calling the finite element software's post-processing application programming interface (API), all nodes in the parametric finite element model are traversed, and the absolute value of the total displacement is read from the result file generated by the static finite element simulation calculation. Or the displacement components in the key direction (such as vertical). All components that satisfy... The nodes are marked. By calculating the spatial distance between the marked nodes, nodes whose distance is less than the set tolerance are grouped into a cluster, and the actual structural part corresponding to each cluster is defined as an independent large deformation zone.

[0123] Finally, all identified high-stress and large-deformation areas were compiled into a structured list of theoretically high-risk areas. This list details the type of each area, its spatial coordinate range, the component identification to which it belongs, and the theoretical stress or displacement extreme values ​​under the most unfavorable working conditions.

[0124] Based on the theoretical high-risk area list, and following the principles of monitoring and the feasibility of engineering implementation, a detailed plan for the deployment of physical sensors is formulated. The optimization criteria that must be strictly followed in the generation of the plan include coverage criteria, sensitivity criteria, redundancy criteria, and feasibility criteria.

[0125] The coverage criterion ensures that at least one sensor is deployed inside or on the boundary of each theoretically high-risk area listed in the inventory. For areas with a large spatial range (such as the high-stress area of ​​the lower flange of a long main beam), sensor arrays are deployed at certain intervals (e.g., 2-3 meters) to capture stress or strain gradients. The determination of this interval needs to be combined with the stress distribution reflected by the Mises stress distribution cloud map, and the sensor arrays should be appropriately densified in areas with drastic stress changes (i.e., areas with large stress gradients).

[0126] Sensitivity criteria dictate that the type and direction of the physical sensor must be specifically selected based on the type and direction of the mechanical quantity to be measured (including stress, strain, displacement, etc.) predicted by the simulation, in order to ensure that the critical response can be captured most effectively. For example, in high-stress areas, unidirectional sensors are placed with their sensitive axis aligned with the principal stress direction of the area shown in the simulation cloud map; in areas of large deformation or critical sections (such as mid-span or 1 / 4 span), displacement gauges or inclinometers are placed; and on critical force transmission path components (such as main hangers), axial force sensors or strain gauges are placed.

[0127] Redundancy criteria: For areas deemed to have extremely high risk or structural importance (such as critical welded joints between the web and flange of the main beam), consider deploying dual sensors for redundancy backup, or using sensors with different measurement principles for cross-validation to improve the reliability of monitoring data.

[0128] The feasibility criterion is that the measurement point locations (locations for deploying physical sensors) initially selected on the drawings need to be verified in conjunction with the actual conditions of the construction site (such as construction access, concrete pouring work space, curing facilities, etc.) to avoid areas that may be subject to collision, severe obstruction, or excessively high temperatures during construction, and to ensure that the physical sensors can survive and work stably throughout the entire construction period.

[0129] The final output of the optimized layout plan for physical sensors is a detailed technical document that can be directly used to guide on-site construction. It typically includes a measurement point layout diagram, a sensor list, and an installation instruction manual. The measurement point layout diagram, on the 2D design drawings or 3D BIM model of the moving formwork, clearly marks the installation location, unique number, and measurement direction of each sensor. The sensor list, in tabular form, lists the sensor model, range, accuracy, installation method (e.g., surface mount, welded mounting, pre-embedded), and corresponding data acquisition channel number for each measurement point, selected according to sensitivity criteria. The installation instruction manual provides specific installation steps, process requirements, precautions, and preliminary acceptance standards after installation, especially for critical or special sensors.

[0130] In one embodiment of the present invention, step S3 includes the following steps:

[0131] The continuous construction process is broken down into multiple sequentially executed pre-set construction procedures. Each pre-set construction procedure has a unique procedure code, a clear start and end state, load conditions, and structural configuration.

[0132] Before each preset construction procedure begins, the procedure parameters corresponding to that procedure are extracted. The procedure parameters include at least the load increment, the boundary condition change matrix, and the structural configuration update parameters.

[0133] The process parameters are input into the parameterized finite element model, and the mechanical state of the model at the end of the previous preset construction process is used as the initial condition to perform the pre-static finite element simulation calculation of the current process.

[0134] After the preliminary static simulation calculations are completed and the stress and displacement field distribution data are generated, all the data for installing the physical sensors, determined according to the optimized layout scheme, are acquired. The spatial coordinates of each planned measuring point are used; based on these spatial coordinates, the mechanical response at each planned measuring point is extracted from the stress and displacement field distribution data using interpolation methods; thus forming a mechanical response prediction dataset for this process that corresponds one-to-one with all planned measuring points. ,in, To plan the total number of measurement points, For the first Predicted values ​​for each planned measurement point ; The index of the current process indicates the number of steps. The preset construction procedures have a value range of 1. ,in This represents the total number of pre-set procedures within the construction period.

[0135] For each predicted value in the mechanical response prediction dataset Assess its uncertainty and provide a confidence interval, which is expressed as follows: ,in, The standard deviation of the uncertainty of the predicted value is... The coefficient is determined based on the target confidence level.

[0136] Specifically, based on the approved "Construction Organization Design" document, the entire continuous construction cycle of the physical moving formwork, from assembly, preloading, pouring to passing through the holes, is scientifically decomposed into a series of sequentially executed, discrete, pre-set construction procedures. Each pre-set construction procedure (hereinafter referred to as "procedure") must be clearly defined to ensure precise synchronization between simulation and construction. The definition of each procedure includes a unique procedure code, a clear start and end state, load conditions, and structural configuration. The unique procedure code is used to uniquely identify the procedure; for example, OP-101 indicates the standard symmetrical pouring of the first segment of the first hole. The clear start and end state can describe the structural configuration, support state, and load state of the formwork at the beginning of the procedure, as well as the target state at the end of the procedure; for example, the starting state of procedure OP-101 can be "the formwork is in place, all front, middle, and rear legs are supported and locked, the formwork system is cleaned, and the concrete pump truck is in place"; the ending state is "the symmetrical pouring of the first segment's design volume (e.g., 120m³) of concrete is completed, and the curing waiting period begins." Specifically, the load conditions quantify the new loads applied to the formwork during the execution of this process, including their magnitude, distribution, location, and possible changes over time. The loads mainly originate from construction activities, such as concrete pouring, equipment movement, and preloading counterweights. Structural configuration refers to whether the geometry and support state of the formwork itself changes during this process. The configuration remains unchanged for most pouring processes, but the configuration changes significantly during formwork movement (through-hole) processes (such as the release and transformation of leg constraints).

[0137] Assign a unique process index to each of the decomposed processes, denoted as . ; It is a positive integer, and its value range is 1. ,in The total number of pre-set procedures within the construction period; Index The physical sequence of operations is strictly incremental, i.e., the number of operations... Only after completion can the next process begin. +1; Index for each process It uniquely corresponds to a procedure code with engineering semantics (such as OP-101).

[0138] Based on the definition of the process, a typical process sequence can be identified; for example: OP-001: the initial state after the physical moving formwork is assembled (self-weight only); OP-002 to OP-005: graded preloading test conditions (such as loading to 60%, 100%, and 120% of the design load, and then unloading); OP-101 to OP-10n: concrete segmental and regional pouring process for the first span; OP-201: the physical moving formwork demolding and movement condition from the span to the next span.

[0139] In each pre-set construction procedure (denoted as the number 1) Before each process begins, process parameters corresponding to that process are extracted from the Construction Organization Design and related technical documents. These parameters are the input instructions that drive the parametric finite element model to perform finite element simulation calculations for the current process, including load increments. Boundary condition change matrix And structural configuration update parameters.

[0140] Load increment parameters This process is relative to the previous process (the first process). The additional load added to the state at the end of the process; for example, for the casting process, The physical essence of this load is the weight of concrete to be poured in this process. To accurately apply this load in the digital structure module of the parametric finite element model, the total weight of the concrete in this section needs to be converted into a load form recognizable by the finite element software. Specifically, the volume of each section is determined according to the pouring zoning diagram in the "Construction Organization Design". and concrete density (usually taken) ), calculate the total weight of the partition. Based on the support area of ​​the partition template Calculate the distributed surface pressure generated by it. For asymmetrical off-center loading conditions, the off-center loading area and weight need to be clearly defined according to the construction plan, and the asymmetrically distributed surface pressure or concentrated force needs to be calculated and applied.

[0141] Boundary condition change matrix This describes the changes in the support state of the mold frame boundary during this process. It is a logical matrix or set of parameters used to activate or deactivate corresponding constraints in the parametric finite element model. For example, at the start of the through-hole condition, the matrix... It includes instructions to release the vertical constraint of the middle support leg and activate the friction constraint of the displacement trolley roller shaft.

[0142] Structural configuration update parameters: If the process involves changes to the geometry of the mold frame itself (such as moving the guide beam forward or changing the support legs), then the parameters required to update the geometry of the digital structural module need to be provided, such as the movement distance. Rotation angle wait.

[0143] Extracted process parameter load increment Boundary condition change matrix The structural configuration update parameters are input into the verified and reliable parametric finite element model to perform the pre-process static finite element simulation calculation for the current process.

[0144] The pre-static finite element simulation calculation process is as follows:

[0145] The parameterized finite element model was used in the previous process (the first step). The mechanical state (including displacement field and stress field) at the end of the process is used as the initial condition for the simulation calculation of the current process; the first process (such as...) The initial conditions are zero stress and zero displacement.

[0146] according to Update the boundary conditions of the parametric finite element model in the finite element software; update the parameters according to the structural configuration; and update the geometric model of the digital structure module by the finite element software (if necessary).

[0147] Load increment In the form of surface pressure, nodal force, or body force (such as inertial force), it is precisely applied to the corresponding units or nodes of the digital structure module according to the construction plan.

[0148] Use the solver of the finite element method software to solve the static equilibrium equations: Among them, the total stiffness matrix The stiffness matrix is ​​assembled from all elements in the digital structure module according to their connection relationship. Its value depends on the element type, geometry, material properties (elastic modulus, Poisson's ratio) and current boundary conditions. It is the known displacement field vector from the previous step; It is the node load vector from the previous step; This is a newly added load vector.

[0149] The solver uses numerical methods (such as direct or iterative methods) to solve this static equilibrium equation, obtaining the displacement increment vector caused by the current load increment. This leads to the superposition of a new global nodal displacement field. Based on this displacement field The solver is based on the geometric equations in finite element theory (i.e., strain-displacement relationships). ,in The differential operator matrix (also known as the strain-displacement matrix) is used to calculate the strain field of all elements. Furthermore, through physical equations (i.e., material constitutive relations, such as Hooke's law for linear elasticity)... ,in (The elasticity matrix) is used to calculate the stress field of all elements. This represents the entire parametric finite element model in the process. The mechanical state under the following conditions. Among them, the Mises equivalent stress is used to evaluate the yield tendency of the material. It is also calculated at the same time.

[0150] After the finite element simulation calculation is completed, the displacement of all nodes and the stress of all elements in the entire digital structure module are obtained. Then, from this massive amount of data, the values ​​determined according to the optimized layout scheme are extracted. Predicted mechanical response values ​​at each planned measurement point (i.e., the spatial location where physical sensors are planned to be installed in the scheme).

[0151] Read each planned measurement point in the optimized deployment scheme precise three-dimensional spatial coordinates And the corresponding predicted mechanical quantities (such as displacement in a certain direction, normal stress, or Mises stress). Since the displacement results calculated by finite element simulation are directly stored on the mesh nodes, while the stress results are usually first calculated at the Gaussian integration points within the element, and then the nodal stress is obtained through averaging or interpolation. The coordinates of the planned measurement points... The locations of these nodes or integration points may not coincide, so interpolation methods are needed to obtain the predicted values ​​at the measurement points.

[0152] For nodal results such as displacement, shape function interpolation is used. The measurement points are then located. Unit Get all of this unit Displacement values ​​of each node (Displacement field obtained from finite element simulation calculation) ), combined with the shape function matrix defined in this unit (The shape function is a mathematical function that describes the displacement distribution within an element, and is determined by the element type.) This is achieved through formulas... The predicted displacement values ​​at the planned measuring points are calculated.

[0153] For element results such as stress (which are typically calculated more accurately at Gaussian integration points), stress smoothing and extrapolation techniques are employed. First, the elements are... The stress calculated at each Gaussian integration point The stress field is extrapolated to each node of the element using shape function interpolation or least squares fitting methods, thus forming a continuous nodal stress field. ; and then using the same shape functions Through formula The nodal stress is interpolated to the measurement point location, and finally extracted according to the stress components required at the measurement point.

[0154] For all After the above extraction operations are completed for each planned measurement point, the mechanical response prediction dataset corresponding to each planned measurement point under the current process is obtained, denoted as a vector form:

[0155]

[0156] in, Representing the Each planned measurement point is in the process Predicted mechanical response values ​​(units consistent with measured values, e.g., displacement is...) The stress is ),in , This represents the total number of planned measurement points.

[0157] Because the parametric finite element model itself has simplification assumptions, its calibrable parameter set Due to prior uncertainties and the potential for slight fluctuations in input loads (such as concrete volume), the predicted values... It is not an absolutely precise and certain value, but an estimate with uncertainty; the uncertainty of each predicted value must be quantitatively assessed.

[0158] Model parameter uncertainty ( Derived from calibrable parameter sets The prior probability distribution can be estimated using the first-order reliability method (FORM); specifically, the predicted value is first calculated. For each calibrable parameter Sensitivity coefficient Combine all sensitivity coefficients to form a sensitivity vector. Combined with parameters The prior covariance matrix (Its diagonal elements are the prior variances of each parameter) Let the off-diagonal elements represent the correlation between parameters (which can be initially set to zero). Then, the variance of the predicted value caused by the propagation of parameter uncertainty can be approximated as: .

[0159] Input load uncertainty ( The main consideration is the actual concrete density during construction. The actual pouring volume V and other parameters may deviate from the design values ​​in the "Construction Organization Design". Typically, based on material variability and construction control levels, a fluctuation range (e.g., ±2%) can be assumed, and the impact on the predicted value can be estimated using error propagation methods similar to sensitivity analysis. contribution variance .

[0160] Assuming that the uncertain sources are independent of each other, then the first... The total uncertainty variance of the predicted values ​​for all measurement points can be estimated as follows: ,in Used to cover other minor error sources (such as numerical calculation truncation error).

[0161] Choose the appropriate coefficient at a given target confidence level (e.g., 95%). (For the normal distribution assumption, =1.96). Then the first The measuring point at the ... The confidence interval for the predicted value of each process is expressed as follows: This confidence interval means that, considering model and input uncertainties, the true response value has a 95% probability of falling within this range.

[0162] In one embodiment of the present invention, step S4 includes the following steps:

[0163] At a frequency no lower than the preset sampling frequency The system synchronously acquires the output signals of all physical sensors installed according to the optimized deployment scheme using a data acquisition device. This synchronous acquisition is triggered by a unified clock on the data acquisition device. The synchronous acquisition process generates a series of discrete, equally timed sampling moments, denoted as... ,in , The start time of data collection The sampling interval is k, where k = 0, 1, 2, ... represents the sampling point number.

[0164] The raw physical sensor data collected is preprocessed, including at least outlier removal, digital filtering and noise reduction, engineering unit conversion, and temperature compensation, at each sampling time. Forming the measured response value vector ;

[0165] The measured response value vector The mechanical response prediction dataset after spatiotemporal alignment Real-time comparison is performed to obtain the residual between the measured values ​​and the predicted values ​​of each planned measurement point;

[0166] Based on the residuals between the measured and predicted values ​​of each planning measurement point, the response deviation value of each planning measurement point is obtained by calculating the standardized confidence bias index, where the i-th Each planned measurement point is located at Response deviation value at time The calculation formula is:

[0167]

[0168] in, and The first Each planned measurement point is at time [time] Measured and predicted values ​​after spatiotemporal alignment; For the first Measurement noise variance of physical sensors and their signal acquisition links installed at each planned measurement point; For parametric finite element model for the first Each planned measurement point is at time [time] Predicted value The variance of the forecast uncertainty;

[0169] Based on all planned measurement points at a series of discrete sampling times Calculated Generate a discrete time series for each measurement point, denoted as the response deviation value time series. .

[0170] Specifically, the physical sensors are installed on the construction site according to the measurement point layout diagram, sensor list, and installation instructions in the optimized deployment plan. Physical sensors include at least strain gauges, displacement gauges, inclinometers, and force sensors. Strain gauges are used to measure the strain (deformation per unit length) at a point on or inside a structure; common types include resistance strain gauges and fiber optic strain gauges. Displacement gauges are used to measure the relative displacement between two points or the absolute displacement of a point relative to a fixed reference point; common types include wire displacement gauges, laser displacement gauges, and GPS receivers. Inclinometers are used to measure the tilt angle of a structure or its components relative to a horizontal plane or initial position; common types include electrolyte inclinometers and MEMS inclinometers. Force sensors are used to measure the magnitude of forces or loads, such as boom tension and support reaction forces; common types include strain gauge force sensors and piezoelectric force sensors.

[0171] At the start of each preset construction procedure (denoted as the s-th procedure), the central control system automatically retrieves the corresponding data acquisition configuration file based on the procedure code and sends a start data acquisition command to the data acquisition system. The configuration file includes the sampling frequency, the channel activation list, and the acquisition duration. The sampling frequency... The settings are based on the dynamic characteristics of construction activities; for quasi-static processes (such as concrete pouring). It is usually set to 1-10Hz; for dynamic processes (such as mold frame movement start-up). The frequency of the sensor data acquisition should be set according to the frequency of structural vibrations that may be triggered, typically not less than 10 times the fundamental frequency of the structure and not less than 50Hz. The channel enable list is used to clearly specify which sensor channels are enabled for this process, reducing data volume. The acquisition duration is set according to the expected duration of the process, and can be set to continuously acquire data until the process ends.

[0172] After receiving the start data acquisition command, the data acquisition instrument uses the set... Sample all enabled channels at equal time intervals; at each sampling time ,in , For the start time, The sampling point number refers to the number of the sampling point during the data acquisition process, determined by the sampling frequency. A fixed series of discrete moments on the time axis This indicates the time point at which the physical sensor signal is measured, and each sampling time. All sensors at N measurement points perform a single measurement synchronously; a timing synchronization module is used to ensure all channels are in the same... The system continuously acquires data, ensuring time alignment. The acquired data is transmitted in real-time to the central processing server via wired Ethernet or wireless communication. The data acquisition unit is a multi-channel synchronous acquisition device responsible for converting conditioned analog signals into digital signals. Its key parameters include sampling frequency, resolution, and input range. Each channel corresponds to a sensor, and sampling across all channels is triggered by the same clock to ensure synchronization. The timing synchronization module uses GPS or Network Time Protocol (NTP) to provide a unified time reference for the entire data acquisition system, ensuring time synchronization accuracy between different acquisition devices is better than 1 millisecond.

[0173] The raw data collected from the physical sensors undergoes preprocessing. Based on physically reasonable ranges (e.g., sensor range) and statistical criteria (e.g., the Laida criterion: removing data points deviating from the mean by more than three standard deviations), obviously abnormal data points are identified and removed. Removed data points are filled using linear interpolation of the preceding and following valid data. A digital low-pass filter is used to remove high-frequency noise; the filter type and cutoff frequency are selected based on sensor characteristics and monitoring requirements. For strain and displacement signals, a cutoff frequency of [missing value] is typically used. A Butterworth low-pass filter; filtering is performed in the time domain, using bidirectional filtering to eliminate phase distortion. Voltage or digital readings are converted to engineering units based on the calibration coefficients of the physical sensor. ,in, and These are calibration coefficients (gain and zero offset), obtained from the sensor's factory calibration or field calibration. For temperature-sensitive sensors (such as strain gauges), data from the temperature sensor is also read simultaneously. Perform real-time temperature compensation: ,in, For temperature coefficient, This is a reference temperature.

[0174] Because there may be slight sampling time drift in each data acquisition channel, cubic spline interpolation is used to resample the data from all channels to a unified time series. Above, ensure all data are strictly aligned under the same timestamp. After preprocessing, each sampling time is obtained. The measured response vector: ,in This represents the total number of physical sensors used in the current process (i.e., the planned number of measurement points). Indicates the first Each measuring point is at Measured values ​​at time (units: dimensionless strain, displacement mm, angle °, force) wait).

[0175] Mechanical response prediction dataset This typically corresponds to a stable process state or a specific moment; spatiotemporal alignment is required for comparison with measured data; the spatiotemporally aligned mechanical response prediction dataset is... .

[0176] Predicted value The three-dimensional coordinates of the corresponding planned measurement points The coordinates are already included in the optimized deployment plan; when installing sensors on site, efforts should be made to ensure that their installation positions are aligned with the target location. If there is an installation error (within the allowable tolerance, such as ±10mm), then the physical sensor is considered to still correspond to the original planned measurement point, and the predicted value does not need to be adjusted.

[0177] For static processes (such as the stable state after pouring), the predicted value is a constant, compared with the measured data throughout the entire process period; that is, for any... , .

[0178] For dynamic processes (such as mold frame movement), the pre-simulation outputs time history predictions. The simulation time axis needs to be aligned with the actual measurement time axis, typically with the start time of the process as the time origin (t=0). After alignment, for the actual measurement time... The corresponding predicted values ​​are obtained from the simulation time history through linear interpolation. .

[0179] The measured response value vector Aligned prediction vector Subtracting each point one by one yields the residual vector: ,in, , .

[0180] Measurement noise variance Characterizing the first The comprehensive random measurement error of a physical sensor and its data acquisition chain (including the sensor itself, connecting cables, signal conditioner, analog-to-digital converter, etc.). Before construction, static calibration tests are performed on the installed physical sensors. Data is collected for a period of time (e.g., 1 hour) during periods of no load change (e.g., nighttime), and the variance of the data for that period is calculated as... The variance can be estimated using the accuracy specifications provided by the sensor manufacturer, such as assuming the error follows a uniform distribution. .Should The value is considered a constant throughout the construction period.

[0181] Model prediction uncertainty variance The uncertainty stems from the inherent approximation of the parametric finite element model and the uncertainty of the input parameters. Its specific composition and calculation methods have been explained in the preceding steps, and mainly include: the variance of model parameter uncertainty. , from calibrable parameter set The prior distribution is calculated through error propagation; the variance of the input load uncertainty is... The variance is calculated through error propagation based on the deviation between the actual construction load (such as concrete unit weight and volume) and the design value; other error variances are also calculated. This is used to cover secondary factors such as numerical discretization errors. Therefore, the variance of the total model prediction uncertainty is: For static processes, It is a constant; for dynamic processes, this value is a time-varying sequence. After being aligned with the actual measured time, the result was obtained. .

[0182] To uniformly assess the significance of deviations at each measuring point and eliminate the influence of differences in dimensions and uncertainties, the Standardized Confidence Bias Index (NCDI) is calculated. Each measuring point at time... The NCDI is defined as:

[0183]

[0184] In the formula, the denominator It is the total standard uncertainty of the difference between the measured value and the predicted value, that is, the expected deviation range after comprehensively considering the measurement error and the model prediction error. This indicates the multiple of the observed deviation relative to the total standard uncertainty.

[0185] like This indicates that the actual deviation is within the expected uncertainty range and is considered normal fluctuation; if This indicates that the deviation is slightly greater than expected and requires attention; if This indicates that the deviation is significantly greater than expected, suggesting possible anomalies or model mismatch.

[0186] For each measuring point In the process Calculate the data at each sampling time point throughout the entire data collection period. of The time series of response deviation values ​​at this measuring point is formed. , ,in For process The total number of sampling points.

[0187] all The NCDI time series of each measurement point constitutes the complete comparison result of this process, which is stored in the form of a two-dimensional array:

[0188]

[0189] Real time The corresponding measured values, predicted values, residuals, and other data are transmitted to the safety assessment module and the database. Database records include fields such as timestamp, process code, measurement point number, measured value, predicted value, and NCDI value.

[0190] An anomaly handling mechanism is established. Specifically, sensor fault detection manifests as follows: if data from a certain channel continuously exceeds its range, signal is lost, or variance is abnormally low (potentially indicating sensor failure), the data from that channel is automatically marked as invalid, and the measurement point is temporarily excluded from subsequent calculations. Simultaneously, an alarm is triggered to notify maintenance personnel. Communication interruption buffering manifests as follows: if communication is temporarily interrupted, the data acquisition instrument locally caches the data, retransmits it upon resumption, and adds the corresponding timestamp. Predictive data missing handling manifests as follows: if a predicted value is missing at a certain moment, the comparison at that moment is paused, and the NCDI value from the previous valid moment is used for interpolation to fill the gap, while a flag is recorded.

[0191] In one embodiment of the present invention, step S5 includes the following steps:

[0192] Define a sliding time window Through response deviation value time series In the sliding time window The calculations are performed internally to extract the time-domain statistical characteristics of each planned measurement point, including the window average deviation index. Deviation trend coefficient Among them, the window average deviation index , For measuring points At a historical moment The standardized confidence bias index, By analyzing the time series of response deviation values ​​within the time window Linear regression was performed on the data points to obtain the deviation trend coefficient. .

[0193] Based on the time-domain statistical characteristics of each planned measurement point and the measured response value, a comprehensive index for evaluating the overall safety status of the physical moving formwork is calculated, which includes the global comprehensive deviation index, the measured response safety index, the deviation significance index, and the trend hazard index.

[0194] Global comprehensive deviation index The calculation formula is:

[0195]

[0196] in, For the first Structural importance weights for each measurement point; It is the order of the norm;

[0197] Actual Response Safety Index ,in For the first Safety threshold for each measuring point;

[0198] Deviation significance index ,in This is the preset critical deviation threshold;

[0199] Trend Risk Index ,in, This is a set of key structural measurement points. As a trend risk indicator, This is the deviation trend coefficient;

[0200] Based on the measured response safety index, deviation significance index, and trend hazard index, a comprehensive risk index is calculated. ;

[0201] Comprehensive risk index It performs real-time matching with preset multi-level risk threshold ranges, triggers corresponding level of early warning signals based on the matching results, outputs early warning signals, and generates a structured early warning report that includes risk level, key anomaly measurement point location, potential cause analysis, and matching response plan.

[0202] Specifically, define a sliding time window. This refers to a fixed-time data segment used for local feature calculation, typically measured in units of sampling points; for example, if the sampling frequency... If the window length is set to correspond to 60 seconds of physical time, then... Each sampling point. The calculation within the sliding time window is performed continuously; for the current sampling moment... (Corresponding sequence index) Its corresponding sliding window contains the most recent Data points, i.e., sequence index from arrive subsequence of Within a continuous sliding time window, for each planned measurement point... Two key time-domain statistical features were extracted: the window average deviation index and the deviation trend coefficient.

[0203] For planning measurement points At the current moment (Corresponding index) ) window average deviation index Defined as the arithmetic mean of all NCDI values ​​for the planned measurement point within its sliding time window: ,in, For measuring points At a historical moment The standardized confidence bias index. Reflects the measuring point In recent times (window) Within the corresponding time period, this represents the average degree to which the measured response deviates from the model prediction. The larger the value, the more significant the deviation; if the value is consistently greater than 1, it indicates a systematic bias at that location that exceeds the expected uncertainty.

[0204] To quantify the direction and rate of change of the Standardized Confidence Deviation Index (NCDI) within a sliding time window, the deviation trend coefficient is calculated. For the NCDI data subsequence within the sliding time window Indexed by time As the independent variable, Perform univariate linear regression with the variable as the dependent variable: ;in, The intercept term obtained from this univariate linear regression model; This is the desired trend coefficient (slope), representing the rate at which the NCDI value changes with the sampling point index. If This indicates that the NCDI of the planned measurement point is trending upwards during the sliding time window, and the safety status may be deteriorating; if If the NCDI is basically stable with no significant trend, then it means that the NCDI is stable and there is no significant trend. If the NCDI is trending downwards, it indicates that the condition may be recovering or improving. The unit is the change in NCDI per sampling point, and its absolute value reflects the strength of the trend.

[0205] Based on the measured response values ​​of all N planned measurement points and extracted time-domain statistical features and The calculation of comprehensive indicators across four dimensions includes global comprehensive deviation indicators, measured response safety indicators, deviation significance indicators, and trend hazard indicators, in order to comprehensively assess the overall safety status of the physical moving formwork.

[0206] Global comprehensive deviation index Used to comprehensively assess the overall severity of deviations at all measurement points (i.e., the portion where NCDI>1), it is a quantitative representation of the overall deviation: ,in, For the first The structural importance weight of each planning measurement point; Let be the norm order, usually taken as (Euclidean norm), to balance the contributions of each measurement point, The larger the value, the higher the sensitivity to the maximum deviation. It is a non-negative scalar, with a value of 0 indicating that the recent average deviation of all measuring points is within the normal uncertainty (NCDI≤1); the larger the value, the more significant and severe the overall deviation.

[0207] The specific calculation method is as follows: Based on the component properties defined in the parametric finite element model, including their material properties (such as steel grade), cross-sectional shape (such as box girder, I-beam), and their position and role in the overall structural force transmission path, and referring to the general principles of component importance classification in structural design codes, a component importance classification and basic score table is formulated, which also serves as a fixed component of this method; for example, the main box girder plates and key hangers of the main beam are classified as core load-bearing components, and assigned... =5.0; Secondary beams and similar components are classified as general load-bearing members, and are assigned... =3.0; Non-load-bearing attachments are assigned =1.0. This classification table is a fixed component of this method. Based on the theoretical high-risk area list, the risk level of the area where the planned measurement point is located is determined; according to the judgment result (high, medium, low) of the list, the corresponding risk correction coefficient is assigned. For example: high-risk areas =1.5, medium-risk area =1.2, low-risk area =1.0. Calculate the unnormalized weights for each planning measurement point. Normalize all N planning measurement points to obtain the final weights: .

[0208] Actual Response Safety Index Directly assess whether the measured physical quantities (stress, displacement, etc.) are close to or exceed their safety thresholds: ,in, No. The safety threshold for the physical quantities measured at each measuring point is set based on the design specifications, structural design calculations, and construction monitoring plan adopted by the project (e.g., for stress measuring points). Allowable stress of the material For displacement measurement points, (The allowable deformation limit is L / 1000 according to the design specifications). It is the most dangerous ratio of the measured value to its safety threshold among all measuring points at the current moment; This indicates that all measured values ​​are within the safe range; This indicates that at least one measurement point has reached or exceeded the safety threshold, posing an immediate risk.

[0209] Deviation significance index Overall deviation The severity is mapped to the [0,1] interval: ,in, The preset critical deviation threshold; when At that time, the overall deviation was considered to have reached a significant level. The saturation level is 1.0; this threshold is determined based on historical data, engineering experience, or through an initial security assessment. It quantifies the significance of the current overall deviation status, with 0 indicating no significant deviation and 1 indicating that the deviation has reached a critical level.

[0210] Critical deviation threshold The specific method for determining the values ​​is as follows: after the physical movable formwork is assembled on-site, the graded preloading test (e.g., loading to 60%, 100%, and 120% of the design load) is used as the data calibration source as a construction quality control step; the global comprehensive deviation index under each level of stable load is calculated. This yields a sample set. The upper statistical limit of this sample set is taken as the threshold. ,in, and These are the mean and standard deviation of the sample set, respectively.

[0211] Trend Risk Index Focusing on the worsening trend of deviations in critical structural components (i.e., the components identified as having the highest risk level or being the most critical in the theoretical high-risk area list, such as the lower edge of the main beam mid-span and key welded joints), early warning of potential risks is provided: ,in, A set of measuring points for key structural components; For set Number of measuring points in key structural components. Trend hazard index. Calculations are performed only on key measuring points; Only consider the upward trend ( Ignoring stable or improving trends; It is an amplification factor when the average deviation The larger the value (greater than 1), the greater the amplification factor, which gives a higher degree of danger to the upward trend based on high deviation. The higher the value, the greater the risk of both high deviation and deterioration trend in key areas.

[0212] Based on the measured response safety index, deviation significance index, and trend hazard index, a comprehensive risk index is calculated. The calculation formula is: ,in, , , The corresponding indicators , The weighting index (a non-negative real number) is usually set to... The highest priority is given to the immediate risk of actual measured values ​​exceeding limits, followed by significant overall deviations, and lastly, the potential risk of a deteriorating trend. use( The form of ) ensures that when the trend risk is 0, the multiplier is 1, which does not affect the total risk. It is a comprehensive scalar value; the larger the value, the higher the overall safety risk level of the physical moving formwork.

[0213] , , The value is determined using a fixed set of empirical values: =2.0, =1.0, =0.5. This fixed empirical value is set strictly according to the engineering safety response priority of "immediate risk > systemic deviation > potential trend". =2.0 assigns the highest nonlinear weight (square term) to the measured response safety index (which directly reflects whether stress and displacement exceed the safety threshold), ensuring that any over-limit signal can be rapidly amplified and dominate the risk assessment. =1.0 allows the bias significance index (which reflects the overall confidence level of the model) to have a linear effect, thus reasonably representing its importance. =0.5 gives the trend risk index (warning of potential deterioration trends) a relatively small weight (square root term), enabling it to effectively avoid frequent false alarms caused by short-term normal fluctuations in data while providing early risk warnings. This set of values ​​is a balanced solution derived after extensive simulation analysis and verification through actual engineering cases, ensuring that the comprehensive risk index responds sensitively and reliably to various risks.

[0214] Based on the project's safety level requirements, a series of progressively increasing risk thresholds are preset. , , , The comprehensive risk index The value range is divided into multiple consecutive risk level intervals, and a corresponding warning signal is defined for each risk level interval. For example: Safe (Green) Pay attention (blue). Warning (yellow) Alarm (orange) Danger (red) ,in, , , , The threshold is an incremental one, determined through engineering experience, historical data, or risk assessment.

[0215] At each sampling time , calculate It performs real-time matching with risk level ranges; based on the matching results, it automatically triggers the corresponding level of early warning signal; the early warning signal can be sent to site management personnel, supervising engineers and project leaders in real time through various means such as monitoring software interface (color flashing, pop-up window), sound and light alarm, mobile phone text message or application push.

[0216] When a yellow alert or higher is triggered, a structured alert report is automatically generated to assist in decision-making. The report includes a risk overview, indicator contribution analysis, location of key anomalies, potential cause analysis, and a corresponding contingency plan. The alert report is automatically saved to the database in a standard document format (such as PDF or HTML) and pushed to the terminals of relevant responsible personnel. The risk overview includes the trigger time. Risk level, comprehensive risk index Values, etc. The contribution analysis of the indicators specifically involves listing the specific values ​​of the measured response safety index, deviation significance index, and trend hazard index, and identifying the main sources of risk (whether it is the measured value exceeding the limit, large overall deviation, or a deteriorating trend). The location of key anomaly measurement points specifically involves listing the measured values ​​exceeding the limit for the measured response safety index (…). The measurement points and their exceedance rates; for the significance index of deviation and the trend risk index, list... and The highest-risk key measuring points, along with their associated components and locations, are identified. Potential cause analysis, based on the location and type of abnormal measuring points, provides possible engineering causes, such as excessive mid-span displacement of the main beam, possibly due to over-pouring of concrete or uneven support of the outriggers. Matching emergency response plans are provided according to risk level and potential cause classification. For example, a yellow alert requires increased attention, verification of relevant measuring point data, and preparation of emergency plans; an orange alert requires suspending the current work, inspecting equipment and loads, and conducting on-site investigations based on the report's location; a red hazard requires immediate evacuation of personnel, cessation of all work, and activation of the highest-level emergency plan.

[0217] In one embodiment of the present invention, step S6 includes the following steps:

[0218] Obtain the calibrable parameter vector in the digital structure module As the object to be calibrated, each parameter is obtained simultaneously. Prior probability distribution ,in, ;

[0219] Collect monitoring data corresponding to multiple completed pre-set construction procedures from historical construction cycles to form a calibration dataset. ,in, For the first The measured response value vector of each process, For the corresponding mechanical response prediction dataset, For the response deviation value vector, For the first The planned measurement point is at the first Average deviation index of each process window , This represents the total number of processes that have been completed.

[0220] Within the Bayesian inference framework, the calibration dataset is utilized. Calculate the parameter vector posterior probability distribution Its calculation is performed by maximizing the lower bound of evidence. The problem is solved, and the lower bound of the evidence is determined. The expression is:

[0221]

[0222] in, This is a variational distribution used to approximate the posterior distribution; Let be the likelihood function, representing the likelihood given parameters. The observed response deviation vector The probability of; For parameters Prior probability distribution ;

[0223] The adjoint method is used to calculate the gradients required in the variational inference process. To address the high computational cost of the parameterized finite element model; to solve the high computational cost problem; and to find the maximizing lower bound of evidence using a gradient optimization algorithm. The problem is to obtain the optimal variational distribution. and average them As parameter calibration value Use the calibration value Update the corresponding parameters in the digital structure module to form a new version of the parameterized finite element model, and evaluate the prediction error of the calibrated parameterized finite element model on the validation dataset, and calculate the degree of reduction of its root mean square error relative to before calibration.

[0224] An online sequential update strategy is adopted; when new monitoring data is obtained... At that time, based on the current posterior distribution Likelihood with new data The posterior distribution of the parameters is updated recursively; the update rule is based on the idea of ​​exponentially weighted moving average, and is approximately expressed as: ,in, It is a forgetting factor.

[0225] Specifically, the construction period has been completed. A pre-set construction procedure ( (Indicates the number of completed operations). For each completed operation... The collected data includes measured response value vectors. ,in The total number of planned measurement points; the corresponding parametric finite element model prediction dataset. ; and the calculated response deviation vector ,in The average deviation index is usually taken as the window average during the stable phase of the process or at a specific moment. All The data from each process step is combined to form the calibration dataset. : .

[0226] Let the vector of calibrable parameters to be calibrated in the digital structure module be... ,in The number of parameters; these parameters are selected from the material parameter category, boundary condition parameter category, and connection stiffness parameter category, and may specifically include equivalent elastic modulus, boundary support stiffness, connection friction coefficient, etc.

[0227] In the Bayesian framework, based on the calibration dataset Update parameters The understanding of this, namely, the calculation of the posterior distribution of the parameters. This calculation requires a prior distribution. and likelihood function Prior distribution This indicates that the calibration dataset was obtained. Previously, parameters were based on design specifications, material certificates, and engineering experience. Quantitative estimation of uncertainty; assuming that the parameters are independent, the prior distribution can be expressed as the product of the prior distributions of the parameters: ,in An index for calibrable parameters; for example, equivalent elastic modulus. The prior can be set as a normal distribution. .

[0228] Likelihood function Indicates that under given parameters Below, a dataset is generated and calibrated from a parametric finite element model. The observed deviation value The probability density of consistent results; to construct the likelihood function, first define the standardized residuals. Theoretically, if the parametric finite element model is completely accurate and the parameters are correct, then It should follow a standard normal distribution with a mean of 0 and a variance of 1. To account for the simplification error of the parametric finite element model and the measurement noise of the solid sensor, we assume... They are independent and follow a pattern with a mean of 0 and a variance of . The normal distribution, where A value ≥1 is used to accommodate additional unmodeled errors arising from factors not considered in the simplifying assumptions of the parametric finite element model. The likelihood function can then be modeled as: ,in The mean is variance is The normal probability density function.

[0229] According to Bayes' theorem, the posterior distribution The calculation formula is: ,in For marginal likelihood (or evidence), in high-dimensional parameter spaces, marginal likelihood is often difficult to calculate directly. Therefore, in practical applications, numerical approximation methods (such as variational inference) are used to find an easily tractable variational distribution. To approximate the true posterior distribution without explicit computation .

[0230] The parameterized finite element model is a high-dimensional nonlinear model. Variational inference methods are used to find a tractable variational distribution belonging to a specific distribution family Q. To approximate the true posterior distribution. The metric for measuring the degree of approximation is... and KL divergence between Minimizing the KL divergence is equivalent to maximizing the lower bound of the evidence. : ,in, To improve the expected log-likelihood, we need to improve the variational distribution. Place the probability mass in a position that enables the observed data The region of parameters with high likelihood values; for With prior distribution The KL divergence between them is used as a regularization term to prevent variational distributions. Excessive deviation from prior knowledge .

[0231] We choose the Gaussian distribution under the mean-field assumption as the family of variational distributions Q, that is, we assume... The components are independent multidimensional normal distributions: ,in, The mean is variance is The one-dimensional normal distribution, and These are the variational parameters to be optimized. This serves as the parameter index. At this point, the optimization problem is transformed into finding the optimal variational parameters. and To maximize the lower bound of evidence .

[0232] Solving for the maximum lower bound of evidence Problems like this are typically solved using gradient-based optimization algorithms (such as stochastic gradient descent and Adam); the key step is computing the objective function. For variational parameters and gradient and The adjoint method (inverse mode automatic differentiation) is used to efficiently compute these gradients. Its core steps are as follows: Given a sample of parameters... Run a complete parametric finite element model simulation to calculate all and log-likelihood The scalar objective function is obtained in one step by solving an adjoint problem (an additional solution to the linear system). For all parameters gradient ;Utilize probability distribution transformation techniques (such as reparameterization techniques) to transform the gradient in the parameter space. Convert to variational parameters The gradient.

[0233] When a certain amount of calibration data has been accumulated Afterwards (e.g., upon completion of construction on a bridge span), initiate an offline batch calibration. Obtain the calibration dataset from the completed processes. Let the mean of the variational distribution be... Equal to the mean and variance of the prior distribution of the parameters. Equals the prior variance. From the current variational distribution. Extraction Parameter samples ( Typically, a value of 50 to 200 is used for each sample. Calculate the gradient using the adjoint method Estimate gradients using reparameterization techniques. and Update variational parameters using an optimizer (such as Adam): ,in The learning rate (usually set to) arrive (between); repeat the steps until convergence, the convergence criterion can be set as The change is less than the threshold (usually set as) arrive The optimization is complete when the maximum number of iterations (e.g., 1000) is reached.

[0234] After optimization, the optimal variational distribution is obtained. Variational distribution mean As a parameter Calibration point estimates Update the calibrable parameter values ​​in the digital structure module to This yields a new version of the parametric finite element model.

[0235] To verify the calibration effect, the root mean square error (RMSE) of the model before and after calibration is calculated on an independent validation dataset (e.g., reserved data from a construction process that was not included in the calibration); the degree of error reduction is defined. To quantify the calibration effect: ,in and are the root mean square errors of the model on the validation dataset before and after calibration, respectively; if If the expected target is achieved (e.g., a reduction of more than 10%), the calibration is considered successful.

[0236] To accommodate the continuity of construction and ensure that the parametric finite element model can promptly reflect the gradual changes in the physical moving formwork structure's state (such as loosening of connections and material creep), an online sequential update strategy is adopted. When a new construction procedure is obtained... Monitoring data Instead of performing time-consuming batch calibration, it is based on the current posterior distribution. (from variational distribution) (Approximate) and new data are then rapidly updated recursively. The update formula is based on the idea of ​​exponentially weighted moving average and is approximated as follows: ,in, Indicates the posterior distribution after online update; The region ∈(0,1] is called the forgetting factor, and new data... The weight in the update is The weight of historical information is .

[0237] The specific method for online sequential updates is to maintain the variational distribution. It is a Gaussian distribution with independent parameters, i.e. ; New data Treat it as a micro dataset, based on the current parameter point estimation (Right now The gradient of new data is calculated using the adjoint method. Update the mean of the variational parameters. ,in It is an approximation of the negative of the Hessian matrix of the current posterior logarithmic density (i.e., the Fisher information matrix), and in practice it is often simplified to... (Diagonal matrix); Variance Can maintain with Same, or according to (1− The proportions were slightly reduced to reflect the increased information; As a new parameter point estimate Update to the digital structure module. The purpose of this online update process is to use monitoring data from the new construction process to quickly and incrementally adjust the model parameters, thereby enabling the parametric finite element model to continuously track the changes in the physical moving formwork structure's state during actual construction.

[0238] Forgetting factor Usually taken Smaller (e.g., 0.1) makes updates smoother, preventing model parameters from fluctuating drastically due to noise from a single observation; larger values... (e.g., 0.3) makes the model more sensitive to new data and can track sudden changes in structural state more quickly; the specific value can be adjusted according to the characteristics of the construction stage: take a smaller value in the stable pouring stage, and take a larger value in the stage of drastic changes such as physical movement of formwork through holes.

[0239] Please see Figure 2 As shown, this invention is a mobile formwork construction monitoring system based on finite element simulation, comprising the following modules:

[0240] Parametric Finite Element Model Establishment Module: Establishes a parametric finite element model corresponding to the physical moving mold frame. The parametric finite element model includes a digital structure module with calibrable parameters.

[0241] Simulation analysis and deployment planning module: Based on the digital structure module, the target construction load condition is simulated and analyzed. According to the mechanical cloud map generated by the simulation analysis, the corresponding theoretical high-risk area in the physical moving formwork is identified, and the optimized deployment scheme of the physical sensors is planned accordingly.

[0242] Pre-simulation and prediction module: Before the start of each preset construction procedure, the procedure parameters are input into the parameterized finite element model for pre-simulation, and the mechanical response prediction dataset of the physical moving formwork under the preset construction procedure is output.

[0243] Data acquisition and comparison module: When executing the preset construction procedure, it acquires the measured response values ​​of the physical sensors installed according to the optimized layout scheme, and compares the measured response values ​​with the corresponding mechanical response prediction dataset in real time to generate a response deviation value time series;

[0244] Safety assessment and early warning module: Based on the time series of the measured response value and response deviation value, the module processes the sequence to extract time-domain statistical features, and then performs a dynamic safety assessment of the physical moving frame, and outputs an early warning signal based on the assessment results;

[0245] Model reverse calibration module: Uses the response deviation value to reverse calibrate the calibrable parameters in the digital structure module to update the parameterized finite element model.

[0246] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for monitoring the construction of a moving formwork based on finite element simulation, characterized in that, Includes the following steps: S1: Establish a parametric finite element model corresponding to the physical moving frame, wherein the parametric finite element model includes digital structural modules with calibrable parameters; S2: Based on the digital structure module, the target construction load condition is simulated and analyzed. Based on the mechanical cloud map generated by the simulation analysis, the corresponding theoretical high-risk area in the physical moving formwork is identified, and the optimized layout scheme of the physical sensors is planned accordingly. S3: Before each preset construction procedure begins, the procedure parameters are input into the parameterized finite element model for pre-simulation, and the mechanical response prediction dataset of the physical moving formwork under the preset construction procedure is output. S4: When performing the preset construction procedure, obtain the measured response values ​​of the physical sensors installed according to the optimized layout scheme, and compare the measured response values ​​with the corresponding mechanical response prediction dataset in real time to generate a response deviation value time series. S5: Based on the measured response value and response deviation value time series, the series is processed to extract time domain statistical features, and then a dynamic safety assessment is performed on the physical moving frame, and an early warning signal is output according to the assessment results; S6: Use the response deviation value to perform reverse calibration on the calibrable parameters in the digital structure module to update the parameterized finite element model; including the following steps: Obtain the calibrable parameter vector in the digital structure module As the object to be calibrated, each parameter is obtained simultaneously. Prior probability distribution ,in, M is the total number of calibrable parameters; Collect monitoring data corresponding to multiple completed pre-set construction procedures from historical construction cycles to form a calibration dataset. Where c is the index of the completed process, , This represents the total number of processes that have been completed. For the first The measured response value vector of each process, For the corresponding mechanical response prediction dataset, The response deviation value vector, where N is the total number of planned measurement points. For the first The planned measurement point is at the first Average deviation index of each process window , ; Within the Bayesian inference framework, the calibration dataset is utilized. Calculate the parameter vector posterior probability distribution Its calculation is performed by maximizing the lower bound of evidence. The problem is solved, and the lower bound of the evidence is determined. The expression is: in, This is a variational distribution used to approximate the posterior distribution; Let be the likelihood function, representing the likelihood given parameters. The observed response deviation vector The probability of; For parameters Prior probability distribution ; For variational distribution The mathematical expectation; KL is the Kullback-Leibler divergence; The adjoint method is used to calculate the gradients required in the variational inference process. To address the high computational cost of the parameterized finite element model; to solve the high computational cost problem; and to find the maximizing lower bound of evidence using a gradient optimization algorithm. The problem is to obtain the optimal variational distribution. and average them As parameter calibration value Use the calibration value Update the corresponding parameters in the digital structure module to form a new version of the parameterized finite element model, and evaluate the prediction error of the calibrated parameterized finite element model on the validation dataset, and calculate the degree of reduction of its root mean square error relative to before calibration. An online sequential update strategy is adopted; when new monitoring data is obtained... At that time, based on the current posterior distribution Likelihood with new data The posterior distribution of the parameters is recursively updated; the update rule is based on the idea of ​​exponentially weighted moving average, and the updated posterior distribution is... Approximately expressed as: ,in, It is a forgetting factor.

2. The method for monitoring the construction of a moving formwork based on finite element simulation according to claim 1, characterized in that, S1 includes the following steps: Construct a three-dimensional geometric model corresponding to the physical moving formwork. The three-dimensional geometric model includes the geometric entities of the main beam system, guide beam system, support leg system, and key connection nodes. The three-dimensional geometric model is meshed using finite element methods to generate a digital structural module composed of shell elements, beam elements, and rod elements. The physical properties of interest in the impact mechanical response in the digital structure module are defined as a calibrable parameter set. , where each parameter Having a prior probability distribution ; The caliable parameter set It includes at least one of the following parameter categories: material parameters, boundary condition parameters, and connection stiffness parameters; wherein, the material parameters include at least the equivalent elastic modulus, Poisson's ratio, and yield strength; the boundary condition parameters include at least the contact support stiffness between the outrigger and the pier, the shear stiffness of the pin connection, and the initial clearance between the connectors; and the connection stiffness parameters include at least the shear stiffness of the bolt connection and the coefficient of friction of the contact surface. The digital structure module and its calibrable parameter set will be used. The structure formed by the collective elements is defined as the parametric finite element model. Based on the initial values ​​of the calibrable parameter set, an initial verification simulation is performed on the digital structure module. The simulation results are compared with the design pre-camber or initial measured data. If the deviation is less than a preset threshold, the parameterized finite element model is determined to be reliable; otherwise, the initial parameter values ​​are adjusted and the verification simulation is performed again.

3. The method for monitoring the construction of a moving formwork based on finite element simulation according to claim 1, characterized in that, S2 includes the following steps: Define at least one target construction load condition, and apply the corresponding load and boundary conditions to the digital structure module. Perform static finite element simulation calculations to obtain the stress and displacement response data of the digital structure module under the target construction load condition. Based on the stress and displacement response data, stress field and displacement field distribution data of the digital structure module are generated, and further processed to generate Mises stress distribution cloud map and displacement distribution cloud map of the digital structure module under the target construction load condition. Based on preset identification criteria, the system automatically analyzes the Mises stress distribution cloud map and displacement distribution cloud map to identify the corresponding theoretical high-risk areas in the physical moving formwork. The preset identification criteria include high-stress area identification criteria and large deformation area identification criteria; The high-stress zone identification criterion refers to identifying all Mises stress values ​​in the stress distribution contour map. satisfy The element that meets this condition is defined as the continuous structural part corresponding to the element on the physical moving formwork as a high-stress zone; stress threshold Based on the allowable stress of the material of the physical moving mold frame Compared with the preset importance coefficient To be determined jointly; The criterion for identifying large deformation zones refers to identifying the absolute values ​​of all displacements in the displacement distribution contour map. satisfy For nodes that meet this condition, the corresponding continuous structural parts on the physical moving formwork are defined as large deformation zones; displacement threshold. Based on the span of the physical moving frame Compared with the preset ratio coefficient To be determined jointly; Based on theoretical high-risk areas, an optimized deployment scheme for physical sensors is generated; wherein, the theoretical high-risk areas are a set of high-stress areas and large-deformation areas identified by preset identification criteria; the generation of the optimized deployment scheme is based on preset optimization criteria for coverage, sensitivity and redundancy.

4. The method for monitoring the construction of a moving formwork based on finite element simulation according to claim 3, characterized in that, S3 includes the following steps: The continuous construction process is broken down into multiple sequentially executed pre-set construction procedures. Each pre-set construction procedure has a unique procedure code, a clear start and end state, load conditions, and structural configuration. Before each preset construction procedure begins, the procedure parameters corresponding to that procedure are extracted. The procedure parameters include at least the load increment, the boundary condition change matrix, and the structural configuration update parameters. The process parameters are input into the parameterized finite element model, and the mechanical state of the model at the end of the previous preset construction process is used as the initial condition to perform the pre-static finite element simulation calculation of the current process. After the preliminary static simulation calculations are completed and the stress and displacement field distribution data are generated, all the data for installing the physical sensors, determined according to the optimized layout scheme, are acquired. The spatial coordinates of each planned measuring point are used; based on these spatial coordinates, the mechanical response at each planned measuring point is extracted from the stress and displacement field distribution data using interpolation methods; thus forming a mechanical response prediction dataset for this process that corresponds one-to-one with all planned measuring points. ,in, To plan the total number of measurement points, For the first Predicted values ​​for each planned measurement point ; This is the index for the preset process, and its value range is... ,in This represents the total number of pre-set procedures within the construction period. For each predicted value in the mechanical response prediction dataset Assess its uncertainty and provide a confidence interval, which is expressed as follows: ,in, The standard deviation of the uncertainty of the predicted value is... The coefficient is determined based on the target confidence level.

5. The method for monitoring the construction of a moving formwork based on finite element simulation according to claim 1, characterized in that, S4 includes the following steps: At a frequency no lower than the preset sampling frequency The system synchronously acquires the output signals of all physical sensors installed according to the optimized deployment scheme using a data acquisition device. This synchronous acquisition is triggered by a unified clock on the data acquisition device. The synchronous acquisition process generates a series of discrete, equally timed sampling moments, denoted as... ,in , The start time of data collection The sampling interval is k, where k = 0, 1, 2, ... represents the sampling point number. The raw physical sensor data collected is preprocessed, including at least outlier removal, digital filtering and noise reduction, engineering unit conversion, and temperature compensation, at each sampling time. Forming the measured response value vector ; The measured response value vector The mechanical response prediction dataset after spatiotemporal alignment Real-time comparison is performed to obtain the residual between the measured values ​​and the predicted values ​​of each planned measurement point; Based on the residuals between the measured and predicted values ​​of each planning measurement point, the response deviation value of each planning measurement point is obtained by calculating the standardized confidence bias index, where the i-th Each planned measurement point is located at Response deviation value at time The calculation formula is: in, and The first Each planned measurement point is at time [time] Measured and predicted values ​​after spatiotemporal alignment; For the first Measurement noise variance of physical sensors and their signal acquisition links installed at each planned measurement point; For parametric finite element model for the first Each planned measurement point is at time [time] Predicted value The variance of the forecast uncertainty; Based on all planned measurement points at a series of discrete sampling times Calculated Generate a discrete time series for each measurement point, denoted as the response deviation value time series. .

6. The method for monitoring the construction of a moving formwork based on finite element simulation according to claim 5, characterized in that, S5 includes the following steps: Define a sliding time window Through response deviation value time series In the sliding time window The calculations are performed internally to extract the time-domain statistical characteristics of each planned measurement point, including the window average deviation index. Deviation trend coefficient Among them, the window average deviation index , For measuring points At a historical moment The standardized confidence bias index, where k is the sampling point number; by analyzing the time series of response bias values ​​within the time window. Linear regression was performed on the data points to obtain the deviation trend coefficient. ; Based on the time-domain statistical characteristics of each planned measurement point and the measured response value, a comprehensive index for evaluating the overall safety status of the physical moving formwork is calculated, which includes the global comprehensive deviation index, the measured response safety index, the deviation significance index, and the trend hazard index. Global comprehensive deviation index The calculation formula is: in, For the first Structural importance weights for each measurement point; For the norm order, This represents the total number of planned measurement points; Actual Response Safety Index ,in For the first Safety threshold for each measuring point; Deviation significance index ,in This is the preset critical deviation threshold; Trend Risk Index ,in, This is a set of key structural measurement points; Based on the measured response safety index, deviation significance index, and trend hazard index, a comprehensive risk index is calculated. ; Comprehensive risk index It performs real-time matching with preset multi-level risk threshold ranges, triggers corresponding level of early warning signals based on the matching results, outputs early warning signals, and generates a structured early warning report that includes risk level, key anomaly measurement point location, potential cause analysis, and matching response plan.

7. A mobile formwork construction monitoring system based on finite element simulation, characterized in that, Includes the following modules: Parametric Finite Element Model Establishment Module: Establishes a parametric finite element model corresponding to the physical moving mold frame. The parametric finite element model includes a digital structure module with calibrable parameters. Simulation analysis and deployment planning module: Based on the digital structure module, the target construction load condition is simulated and analyzed. According to the mechanical cloud map generated by the simulation analysis, the corresponding theoretical high-risk area in the physical moving formwork is identified, and the optimized deployment scheme of the physical sensors is planned accordingly. Pre-simulation and prediction module: Before the start of each preset construction procedure, the procedure parameters are input into the parameterized finite element model for pre-simulation, and the mechanical response prediction dataset of the physical moving formwork under the preset construction procedure is output. Data acquisition and comparison module: When executing the preset construction procedure, it acquires the measured response values ​​of the physical sensors installed according to the optimized layout scheme, and compares the measured response values ​​with the corresponding mechanical response prediction dataset in real time to generate a response deviation value time series; Safety assessment and early warning module: Based on the time series of the measured response value and response deviation value, the module processes the sequence to extract time-domain statistical features, and then performs a dynamic safety assessment of the physical moving frame, and outputs an early warning signal based on the assessment results; Model reverse calibration module: This module uses the response deviation value to reverse-calibrate the calibrable parameters in the digital structure module to update the parameterized finite element model; it includes the following steps: Obtain the calibrable parameter vector in the digital structure module As the object to be calibrated, each parameter is obtained simultaneously. Prior probability distribution ,in, M is the total number of calibrable parameters; Collect monitoring data corresponding to multiple completed pre-set construction procedures from historical construction cycles to form a calibration dataset. Where c is the index of the completed process, , This represents the total number of processes that have been completed. For the first The measured response value vector of each process, For the corresponding mechanical response prediction dataset, The response deviation value vector, where N is the total number of planned measurement points. For the first The planned measurement point is at the first Average deviation index of each process window , Within the Bayesian inference framework, the calibration dataset is utilized. Calculate the parameter vector posterior probability distribution Its calculation is performed by maximizing the lower bound of evidence. The problem is solved, and the lower bound of the evidence is determined. The expression is: in, This is a variational distribution used to approximate the posterior distribution; Let be the likelihood function, representing the likelihood given parameters. The observed response deviation vector The probability of; For parameters Prior probability distribution ; For variational distribution The mathematical expectation; KL is the Kullback-Leibler divergence; The adjoint method is used to calculate the gradients required in the variational inference process. To address the high computational cost of the parameterized finite element model; to solve the high computational cost problem; and to find the maximizing lower bound of evidence using a gradient optimization algorithm. The problem is to obtain the optimal variational distribution. and average them As parameter calibration value Use the calibration value Update the corresponding parameters in the digital structure module to form a new version of the parameterized finite element model, and evaluate the prediction error of the calibrated parameterized finite element model on the validation dataset, and calculate the degree of reduction of its root mean square error relative to before calibration. An online sequential update strategy is adopted; when new monitoring data is obtained... At that time, based on the current posterior distribution Likelihood with new data The posterior distribution of the parameters is recursively updated; the update rule is based on the idea of ​​exponentially weighted moving average, and the updated posterior distribution is... Approximately expressed as: ,in, It is a forgetting factor.

Images