Optimization Method and System for Prestressing Tensioning Parameters of Precast Segments in Steel-Concrete Composite Beam Bridges

Abstract

This invention relates to the field of parameter optimization technology, and discloses a method and system for optimizing prestressed tensioning parameters of precast segments in steel-concrete composite beam bridges. The method includes: acquiring a design BIM model and classifying precast segments; constructing an initial digital twin model and acquiring monitoring data; processing the monitoring data to obtain corrected data; comparing the corrected data with the model simulation results, calibrating the model to generate an accurate digital twin model; evaluating multi-dimensional robustness indicators to construct feature vectors; and combining the feature vectors with the accurate model to generate an adaptive tensioning strategy sequence through a graph neural network. This invention achieves intelligent optimization of tensioning parameters, significantly improving the construction accuracy and structural safety of steel-concrete composite beam bridges.

Description

Technical Field

[0001] This invention relates to the field of parameter optimization technology, and more specifically, to a method and system for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges. Background Technology

[0002] In the prestressing tensioning construction of precast segments of steel-concrete composite beam bridges, the optimization of prestressing tensioning parameters is crucial for ensuring the bridge's construction accuracy and structural safety. Chinese patent application CN114201796A discloses a method for accurately calculating prestress loss considering the influence of tensioning sequence. This method calculates the actual tension stress value after prestressing tendon loss by constructing a tension stress matrix and a prestress loss calculation matrix. Chinese patent application CN102941623A provides a method for controlling the tensioning parameters of a box girder prestressing tensioning system. Through equipment such as controllers, pressure sensors, oil pump trucks, hydraulic jacks, and displacement sensors, real-time monitoring and adjustment of the tensioning process are achieved.

[0003] However, existing technologies have shortcomings in constructing temporal dependencies and spatial coupling relationship diagrams between tension control units. Traditional methods struggle to identify implicit stress influences across segments through intelligent models, such as the indirect effect of tensioning of type A segments on the preload of type C nodes. This limits tensioning strategy generation to single segments or fixed processes, failing to adapt to complex stress scenarios involving dynamic coupling between segments. When type A segments are tensioned, stress transfer can cause changes in the preload of type C nodes. Without an effective relationship diagram, these changes cannot be accurately identified and quantified, creating blind spots in tensioning strategy optimization. Furthermore, existing technologies lack sufficient collaboration between decision models and digital twin models, failing to fully utilize the simulation capabilities of precise digital twin models and the autonomous learning capabilities of graph neural networks-reinforcement learning. This makes it difficult to dynamically adjust tensioning parameters based on real-time robustness states, resulting in deviations between strategy execution and the actual stress state of the structure, affecting construction accuracy and structural safety. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of existing technologies, this invention provides a method and system for optimizing prestressed tensioning parameters of precast segments in steel-concrete composite beam bridges. It constructs a complete process integrating segment classification, data correction, model calibration, and strategy generation. By building a temporal dependency and spatial coupling correlation graph between tension control units, and combining a precise digital twin model with graph neural network-reinforcement learning technology, dynamic optimization of tensioning parameters is achieved. This invention can accurately identify the implicit stress influence across segments, significantly improve the accuracy and efficiency of prestressed tensioning, effectively ensure the construction accuracy and structural safety of steel-concrete composite beam bridges, and achieve intelligent and precise operation throughout the entire process from data acquisition to strategy execution.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Methods for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges include:

[0007] Obtain the design BIM model, classify the prefabricated segments based on the design BIM model, construct the initial digital twin model of the prefabricated segments based on the classification results, set up monitoring points, and obtain the original monitoring data stream;

[0008] The original monitoring data stream is filtered, verified, and corrected to obtain the corrected original monitoring data stream;

[0009] The corrected original monitoring data stream is compared with the simulation results of the initial digital twin model of the prefabricated segment to determine whether a calibration signal is triggered. If a calibration signal is triggered, dynamic calibration of the initial digital twin model is performed to generate an accurate digital twin model of the prefabricated segment.

[0010] Based on the corrected original monitoring data stream, a multi-dimensional robustness index assessment is conducted, and a robustness index feature vector is constructed.

[0011] Based on the robustness index feature vector, a tension control unit is constructed and an association graph of the control unit is established. Combined with the accurate digital twin model of the prefabricated segment, a graph neural network tension decision model is constructed and trained to generate an adaptive tension strategy sequence.

[0012] Furthermore, the classification of prefabricated segments based on the design BIM model includes:

[0013] Structural parameters of the steel-concrete composite beam bridge were extracted from the design BIM model. Based on the structural parameters, the precast segments were divided into three types: Class A segments (negative bending moment zone segments), Class B segments (mid-span segments), and Class C segments (connection node segments).

[0014] Furthermore, the initial digital twin model for constructing the prefabricated segment includes:

[0015] Based on the segment type classification results, detailed construction parameters of each segment type are extracted to generate segment classification identification data; based on the segment classification identification data, an initial digital twin model of the prefabricated segment is constructed on the basis of the design BIM model.

[0016] Furthermore, the method for obtaining the corrected original monitoring data stream includes:

[0017] The raw monitoring data stream is initially screened to remove abnormal data, resulting in a filtered raw monitoring data stream.

[0018] Adjacent comparisons and secondary verifications were performed on the filtered raw monitoring data streams to obtain the reliability strain data and reliability displacement data of Class A segments, the reliability strain data and reliability displacement data of Class B segments, and the reliability temperature data and reliability preload data of Class C segments.

[0019] Based on the reliability temperature data of Class C segments, the reliability strain and displacement data of Class A segments, the reliability strain and displacement data of Class B segments, and the reliability preload data of Class C segments are corrected for temperature effects to obtain the corrected original monitoring data stream. The corrected original monitoring data stream includes the corrected strain and displacement data of Class A segments, the corrected strain and displacement data of Class B segments, and the corrected preload data of Class C segments.

[0020] Furthermore, the method for determining whether a calibration signal has been triggered includes:

[0021] The corrected original monitoring data stream is compared with the simulation results of the initial digital twin model of the prefabricated segment to determine whether the Class A segment strain calibration signal, Class A segment displacement calibration signal, Class B segment strain calibration signal, Class B segment displacement calibration signal, and Class C segment connection calibration signal are triggered. If any of the Class A segment strain calibration signal, Class A segment displacement calibration signal, Class B segment strain calibration signal, Class B segment displacement calibration signal, and Class C segment connection calibration signal are triggered, then the calibration signal is determined to be triggered.

[0022] Furthermore, the method for constructing the robustness indicator feature vector includes:

[0023] Based on the corrected strain and displacement data of Class A segments in the corrected original monitoring data stream, a crack risk assessment and deformation compatibility assessment are conducted in the negative bending moment zone, and a comprehensive risk status record table for Class A segments is generated.

[0024] Based on the corrected strain and displacement data of Class B segments in the corrected original monitoring data stream, stress level and linear state assessments are performed in the mid-span zone, and a comprehensive state record table for Class B segments is generated.

[0025] Based on the corrected preload data and reliability temperature data of Class C segments in the corrected original monitoring data stream, connection reliability assessment and temperature stability assessment are performed, and a comprehensive status record table of Class C segments is generated.

[0026] Based on the Class A segment comprehensive risk status record table, the Class B segment comprehensive status record table, and the Class C segment comprehensive status record table, a robustness indicator feature vector is constructed.

[0027] Furthermore, the method for constructing the tension control unit includes:

[0028] Prestressed steel strand parameters are extracted from the design BIM model; based on the segment classification and identification data, a unique segment number is assigned to each Class A, Class B, and Class C segment;

[0029] Based on the prestressed steel strand parameters, each prestressed steel strand is spatially matched with the segment number sequence to determine the segment type and segment number to which each prestressed steel strand belongs;

[0030] Based on the preset tensioning levels, the prestressed steel strands in each segment are grouped according to the tensioning levels to form a tensioning control unit consisting of segment numbers and tensioning levels.

[0031] Furthermore, the method for establishing the control unit association diagram includes:

[0032] Analyze the temporal dependencies of different tensioning level control units within the same segment to form an irreversible sequence;

[0033] The spatial coupling relationship between adjacent segments was analyzed, and the structural correlation between different segments was identified based on the classification results of three types of segments.

[0034] Using the tension control unit as a node, the temporal dependency relationship forms a directed edge, and the spatial coupling relationship forms an undirected edge, thus constructing the control unit association graph.

[0035] Furthermore, the method for generating the adaptive tensioning strategy sequence includes:

[0036] The robustness index feature vector in the current state is extended by using the state of the tensioned control unit and the initial state of the untensioned control unit to obtain the extended robustness index feature vector; the extended robustness index feature vector is input into the trained graph neural network tensioning decision model to output the tensioning decision, which is the next control unit to be tensioned and its tension force value;

[0037] The structural response after the execution of the tensioning decision is predicted using a precise digital twin model. The results are then verified to see if the tensioning constraints set during the training of the graph neural network tensioning decision model are met. If they are met, the tensioning decision is retained; otherwise, the tensioning decision is regenerated. The above process is iteratively executed until a complete tensioning strategy sequence covering all tensioning control units is generated, which serves as the adaptive tensioning strategy sequence.

[0038] A prestressing tensioning parameter optimization system for precast segments of steel-concrete composite beam bridges, used to implement the aforementioned method for optimizing prestressing tensioning parameters of precast segments of steel-concrete composite beam bridges, the system comprising:

[0039] Segment classification module: used to obtain the design BIM model and classify the prefabricated segments based on the design BIM model;

[0040] Data acquisition module: Based on the classification results, construct the initial digital twin model of the prefabricated segment and set up monitoring points to acquire the raw monitoring data stream;

[0041] The data processing module is used to filter, verify, and correct the raw monitoring data stream to obtain the corrected raw monitoring data stream.

[0042] Model calibration module: Compares the corrected original monitoring data stream with the simulation results of the initial digital twin model of the prefabricated segment to determine whether a calibration signal is triggered. If a calibration signal is triggered, dynamic calibration of the initial digital twin model is performed to generate an accurate digital twin model of the prefabricated segment.

[0043] Robustness Index Assessment Module: Based on the corrected original monitoring data stream, multi-dimensional robustness index assessment is performed, and a robustness index feature vector is constructed;

[0044] Strategy generation module: Based on the robustness index feature vector, a tension control unit is constructed and a control unit association graph is established. Combined with the accurate digital twin model of the prefabricated segment, a graph neural network tension decision model is constructed and trained to generate an adaptive tension strategy sequence.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] This invention achieves precise optimization of prestressing tension parameters for precast segments of steel-concrete composite beam bridges by acquiring the design BIM model, classifying precast segments, and constructing initial digital twin models. It combines a complete process of monitoring data processing, model calibration, robustness assessment, and adaptive tensioning strategy generation. Specifically, the screening, verification, and correction of the original monitoring data stream effectively eliminates abnormal interference, ensuring the data accurately reflects the structural stress state and providing a reliable foundation for subsequent analysis. By comparing and calibrating the corrected data with the simulation results of the initial model, the model parameters are dynamically optimized, enabling the generated precise digital twin model to accurately map the actual structural characteristics and improve simulation reliability. Multi-dimensional robustness index assessment and feature vector construction based on the corrected data quantitatively integrate the risk status and structural performance of different segments, providing clear risk guidance and quantitative basis for tensioning strategy formulation, avoiding the one-sidedness and subjectivity of traditional assessments. By constructing a relationship graph between the tension control unit and the control unit based on feature vectors, and combining a precise digital twin model with a graph neural network-reinforcement learning method to generate an adaptive tensioning strategy sequence, the spatial relationship and temporal dependency of each segment can be fully considered. This allows the tensioning operation to focus on high-risk areas while also taking into account the overall stress coordination of the structure. The strategy can be dynamically adjusted according to the real-time status, overcoming the limitations of traditional fixed tensioning processes that are difficult to adapt to dynamic changes in the structure. Ultimately, this significantly improves the accuracy and efficiency of prestressed tensioning, ensures the construction accuracy and structural safety of steel-concrete composite beam bridges, and achieves full-process intelligence and precision from data acquisition to strategy execution. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of the method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges according to the present invention.

[0049] Figure 2 A flowchart illustrating the principle of generating a precise digital twin model of prefabricated segments for this invention;

[0050] Figure 3 A flowchart illustrating the method for constructing a robustness indicator feature vector to evaluate multi-dimensional robustness indicators in this invention;

[0051] Figure 4 This is a schematic diagram of the control unit association diagram of the present invention;

[0052] Figure 5This is a functional module diagram of the prestressing tensioning parameter optimization system for precast segments of steel-concrete composite beam bridges in this invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Example 1

[0055] Please see Figure 1 As shown in the figure, this embodiment provides a method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges, including:

[0056] Step S10: Obtain the design BIM model, classify the prefabricated segments based on the design BIM model, construct the initial digital twin model of the prefabricated segments based on the classification results, set up monitoring points, and obtain the original monitoring data stream.

[0057] Further, step S10 includes:

[0058] Step S11: Obtain the design BIM model and extract the structural parameters of the steel-concrete composite beam bridge from the design BIM model;

[0059] Step S12: Based on the construction parameters, the precast segments are divided into three types of segments, including type A segments (negative bending moment zone segments), type B segments (mid-span segments), and type C segments (connection node segments).

[0060] Step S13: Based on the segment classification results of the three types of segments, extract the detailed construction parameters of each type of segment to generate segment classification identification data; the segment classification identification data includes the detailed construction parameters of type A segments, type B segments and type C segments.

[0061] Step S14: Based on the segment classification and identification data, construct the initial digital twin model of the prefabricated segment on the basis of the design BIM model;

[0062] Further, step S14 includes:

[0063] Step S141: Based on the detailed structural parameters of Class A segments in the segment classification identification data, establish a three-dimensional spatial model of shear stud arrangement and calculate the initial combined section stiffness parameter EIA of Class A segments.

[0064] Step S142: Based on the detailed structural parameters of Class B segments in the segment classification identification data, establish a collaborative stress relationship model between the main beam and the profiled steel plate bridge deck.

[0065] Step S143: Based on the detailed structural parameters of Class C segments in the segment classification identification data, establish a toothed connection stiffness calculation model and determine the toothed connection stiffness K. C .

[0066] Step S144: Combine the initial composite section stiffness parameter EIA of type A segment, the cooperative force relationship model of type B segment, and the toothed connection stiffness K of type C segment. C The components are integrated to form an initial digital twin model of the prefabricated segments.

[0067] Step S15: Based on the initial digital twin model of the prefabricated segment, monitoring points are set up for the key stress-bearing parts of different segment types to obtain the original monitoring data stream. The original monitoring data stream includes strain data εA(t) and displacement data δA(t) for type A segments, strain data εB(t) and displacement data δB(t) for type B segments, and temperature data TC(t) and preload data FC(t) for type C segments, where t represents time.

[0068] Specifically, a design BIM model containing the three-dimensional structural information of the entire steel-concrete composite beam bridge is obtained. This model includes data such as geometric dimensions, material properties, and component connection relationships. The structural parameters of the steel-concrete composite beam bridge are extracted through the secondary development interface of the BIM software. These parameters include the bridge span length, the three-dimensional coordinate information, cross-sectional dimensions, material mechanical properties, and connection structure types within the precast segments. Cross-sectional dimensions include, for example, beam height, web thickness, and flange width. Material mechanical properties include, for example, the elastic modulus of steel and the compressive strength grade of concrete. Connection structure types within the precast segments include, for example, whether they include toothed joints or bolt assemblies. Based on these structural parameters, the precast segments are divided into three categories: A, B, and C, representing segments in the negative bending moment zone, mid-span segments, and connection node segments, respectively. The criterion for class A segments is that the distance from the center point of the segment to the pier support is less than a preset negative bending moment influence distance threshold L. neg L neg The influence range of Class A segments is determined by multiplying the span length by KA. KA is a coefficient used to define the influence range of Class A segments. Specifically, it is the ratio of the distance from which the negative bending moment significantly affects the segment in the negative bending moment zone to the bridge span length. This is determined based on the distribution characteristics of negative bending moment in simply supported beams in structural mechanics. Negative bending moment has a significant impact near the piers, and its influence range is typically 0.15 to 0.25 times the span. These segments require close monitoring of the tensile stress in the bridge deck. Class B segments are determined by the distance from the segment center point to the mid-span position being less than the preset mid-span influence distance threshold L. mid L midThe product of the span length and KB is used to define the influence range of Class B segments. Specifically, KB is the ratio of the distance at which the vertical deflection of the mid-span segment is significantly affected to the bridge span length. KB is determined based on the deflection variation pattern in the mid-span region. The vertical deflection variation is most significant within a range of 0.2 to 0.3 times the span near the mid-span, requiring close monitoring of vertical deflection. Class C segments are determined based on the presence of toothed connection structures and bolted assembly structures within the segment, requiring monitoring of the connection status. Based on the above classification results, detailed structural parameters for each type of segment are extracted to generate segment classification identification data. For Class A segments, the ratio of the moment of inertia of the steel main beam section about the neutral axis to the distance of the farthest fiber in the negative bending moment zone (i.e., the section modulus parameter WA of the steel main beam section in the negative bending moment zone) and the ratio of the total cross-sectional area of ​​the longitudinal reinforcement of the bridge deck to the effective cross-sectional area of ​​the bridge deck (i.e., the reinforcement ratio parameter ρA of the bridge deck) are extracted as detailed structural parameters for Class A segments. For Class B segments, the product of the elastic modulus of the minor main beam material and the moment of inertia of the section (i.e., the bending stiffness parameter EIB of the minor main beam) and the cross-sectional area moment of the profiled steel sheet and concrete combination (i.e., the composite section parameter SB of the profiled steel sheet bridge deck) are extracted as detailed structural parameters for Class B segments. For Class C segments, the geometric dimension parameters of the tooth groove, including the tooth groove depth, width, and slope (GC), and the design value parameter of the bolt preload determined according to the bolt diameter, material strength grade, and connection stress requirements are extracted as detailed structural parameters for Class C segments.

[0069] Based on the segment classification and identification data, an initial digital twin model is constructed on the basis of the design BIM model. Specifically, for Class A segments, according to their detailed structural parameters, shear stud 3D models are arranged in 3D modeling software according to the design spacing and quantity. By establishing the contact relationship between the steel main beam and the concrete bridge deck, the initial combined section stiffness parameter EIA is calculated using the superposition method. That is, the section stiffness of the steel main beam and the section stiffness of the concrete bridge deck are calculated separately, and the connection stiffness reduction coefficient determined by the ratio of the shear bearing capacity obtained from shear stud tests to the design value is considered. The two are then superimposed to obtain the EIA. For Class B segments, according to their detailed structural parameters, by defining the friction coefficient and bond strength parameters between the main beam and the profiled steel bridge deck determined by material tests, a cooperative stress relationship model is established to reflect the common deformation characteristics. For Class C segments, according to their detailed structural parameters, based on the toothed contact area and bolt preload, a spring-damping model is used to simulate the toothed connection stiffness. The calculation formula is: ,in, This refers to the tooth groove contact area. The elastic modulus of the toothed concrete. The design value for bolt preload and the stiffness of the toothed connection. Contact area with tooth groove Elastic modulus of toothed concrete and bolt preload design value They are positively correlated. The coefficient of friction of the tooth groove contact surface is determined experimentally based on the material properties and surface treatment of the tooth groove contact surface. This refers to the tooth groove contact depth. For the number of bolts, The stiffness coefficient of a single bolt is determined through bolt mechanical property tests. This formula comprehensively considers the contact stiffness of the toothed concrete and the stiffness of the bolt connection, making... Follow and The size increases with the size of the segment. The model parameters of the above-mentioned segments A, B, and C are integrated to form an initial digital twin model containing the prefabricated segments of the entire bridge.

[0070] Based on the initial digital twin model of the precast segments, monitoring points are set up for key stress-bearing parts of different segment types. For Class A segments: strain monitoring points are set up at equal intervals on the top surface of the bridge deck to obtain strain data εA(t) of Class A segments, and vertical displacement monitoring points are set up at both ends of the segment to obtain displacement data δA(t) of Class A segments. For Class B segments: strain monitoring points are set up at equal intervals on the lower flange of the minor main beam to obtain strain data εB(t) of Class B segments, and vertical displacement monitoring points are set up at the bottom of the mid-span to obtain displacement data δB(t) of Class B segments. For Class C segments: temperature sensors are set up on both sides of the tooth groove to obtain temperature data TC(t) of Class C segments, and preload monitoring plates are installed on the bolt heads to obtain preload data FC(t) of Class C segments. All monitoring data are uploaded to the digital twin system to form the original monitoring data stream.

[0071] Compared to the traditional method of roughly classifying segments based solely on location, the classification based on specific structural parameters makes the definition of segments A, B, and C more consistent with the structural stress characteristics. For example, segment A is classified based on a specific distance threshold from the pier, ensuring accurate identification of negative bending moment zones and making monitoring point deployment more targeted. The extraction of detailed structural parameters from the segment classification and identification data provides clear direction for the subsequent construction of the digital twin model. The WA and ρA of segment A are directly related to its crack resistance monitoring requirements, the EIB and SB of segment B directly affect the accuracy of vertical deflection calculation, and the tooth groove geometry parameter GC and bolt preload design value parameter of segment C ensure the effectiveness of connection status monitoring. This correspondence ensures a high degree of matching between the initial digital twin model parameters and the monitoring targets, improving the simulation accuracy of the model. In the initial construction of the digital twin model, differentiated modeling methods were adopted for different segment types. The combined section stiffness calculation of Class A segments took into account the actual effect of shear studs. The collaborative force model of Class B segments reflected the overall deformation characteristics of the structure. The connection stiffness model of Class C segments took into account the combined effect of toothed grooves and bolts. The integration of the three enabled the model to reflect both the independent characteristics of each segment and the overall force relationship of the entire bridge, making up for the shortcomings of traditional models in characterizing the characteristics of different parts. The monitoring points are deployed based on the key stress-bearing parts of each segment type. For Class A segments, the bridge deck strain and end displacement are monitored; for Class B segments, the main beam strain and mid-span deflection are monitored; and for Class C segments, the temperature and preload are monitored. This forms a targeted monitoring network. Compared to the method of uniformly deploying monitoring points throughout the entire bridge, the sensors of this invention are deployed directly to the risk points of each segment. This targeted approach allows the collected data to directly reflect the stress state of the key structural parts, avoiding interference from non-critical data in the analysis process. At the same time, resource allocation is more focused on areas that require key monitoring, which reduces the overall cost of the monitoring system and the data transmission load, while ensuring that key parameters that play a decisive role in structural safety can be accurately captured. The accuracy of segment classification provides a clear direction for the extraction of detailed structural parameters, which in turn lays the foundation for the accurate construction of the digital twin model. The accuracy of the model guides the optimized layout of monitoring points, and the targeted nature of the monitoring data can in turn feed back into the subsequent calibration of the model. This closed-loop relationship makes the whole process, from data acquisition to model construction to monitoring implementation, an organic whole. It not only realizes the effective monitoring of key parts of precast segments of steel-concrete composite beam bridges, but also provides a precise data foundation for the subsequent optimization of prestressing tensioning parameters. This is because the monitoring data of different segments can directly reflect their sensitivity to tensioning, providing a basis for the formulation of differentiated tensioning strategies.

[0072] Step S20: Filter, verify and correct the original monitoring data stream to obtain the corrected original monitoring data stream;

[0073] Further, step S20 includes:

[0074] Step S21: Perform preliminary screening on the raw monitoring data stream, remove abnormal data, and obtain the screened raw monitoring data stream; the screened raw monitoring data stream includes strain data and displacement data after screening of Class A segments, strain data and displacement data after screening of Class B segments, and temperature data and preload data after screening of Class C segments.

[0075] Step S22: Perform adjacent comparison and secondary verification on the screened raw monitoring data stream to obtain the reliability strain data and reliability displacement data of Class A segments, the reliability strain data and reliability displacement data of Class B segments, and the reliability temperature data and reliability preload data of Class C segments.

[0076] Further, step S22 includes:

[0077] Step S221: Perform strain verification and displacement verification on the strain data and displaceable data of the selected Class A segments to obtain the reliability strain data ε of the Class A segments. reli A(t) and reliability displacement data δε reli A(t);

[0078] Step S222: Perform strain verification and displacement verification on the strain data and displaceable data of the selected Class B segments to obtain the reliability strain data ε of the Class B segments. reli B(t) and reliability displacement data δε reli B(t);

[0079] Step S223: Perform temperature verification and preload verification on the temperature data and preload data after screening the C-class segments to obtain the reliability temperature data T for the C-class segments. reli C(t) and reliability preload data F reli C(t);

[0080] Step S23: Based on the reliability temperature data of Class C segments, perform temperature effect correction on the reliability strain and displacement data of Class A segments, the reliability strain and displacement data of Class B segments, and the reliability preload data of Class C segments to obtain the corrected original monitoring data stream; the corrected original monitoring data stream includes the corrected strain and displacement data of Class A segments, the corrected strain and displacement data of Class B segments, and the corrected preload data of Class C segments.

[0081] Specifically, the raw monitoring data stream is first preliminarily screened to remove abnormal data. Abnormal data refers to monitoring values ​​that significantly deviate from the normal stress range or physical laws of the structure, such as sudden values ​​caused by loose sensor cables, high-frequency oscillation values ​​caused by electromagnetic interference, or saturation values ​​exceeding the sensor's range. A statistical method is used to identify abnormal data. By calculating the mean and standard deviation of data from the same monitoring point over multiple consecutive acquisition cycles, values ​​deviating from the mean by more than three times the standard deviation are considered abnormal. Step-like abrupt changes in the time series, such as a difference between two adjacent acquisition cycles greater than twice the historical maximum fluctuation value of the monitoring point, are also considered abnormal. After the preliminary screening, the raw monitoring data stream is obtained, including strain and displacement data from Class A segments, strain and displacement data from Class B segments, and temperature and preload data from Class C segments. Based on this, adjacent comparisons and secondary verifications are performed on the screened raw monitoring data streams. Strain and displacement verifications are performed on the strain and displacement data from Class A segments: the strain difference Δε on both sides of the same cross-section of the bridge deck top surface is calculated. sec A, strain difference Δε sec A is the difference between the strain sensor readings at the left and right edges of the bridge deck within the same cross-section on the top surface of the bridge deck; if Δε sec A is greater than the strain difference threshold ε thre A will then be used to calculate the strain difference Δε sec A set of strain data is marked as requiring verification; calculate the displacement difference Δδ between the two ends of the segment. ends A, Δδ ends A represents the difference between the displacement sensor readings at both ends of segment A; Δδ ends If A is greater than the displacement difference threshold δ thre If segment A is selected, it is marked as requiring verification. The strain and displacement data of segments not marked as requiring verification (after screening) are then marked as the reliability strain data ε of segment A. reli A(t) and reliability displacement data δε reli A(t); strain difference threshold ε thre A calculates the maximum asymmetric strain of this type of segment under the design load using finite element simulation, and uses 90% of this strain as the strain difference threshold. Displacement difference threshold δ thre A is determined based on the segment length and the maximum allowable torsion angle. The maximum allowable torsion angle is obtained from the material crack resistance test and is converted into the displacement difference threshold by multiplying the segment length and the maximum torsion angle.

[0082] Strain and displacement verifications were performed on the strain and displacement data of the B-class segments after screening: the strain gradient ε between adjacent measuring points was calculated. gradB, the strain gradient is the ratio of the strain difference between two adjacent points to the distance between the two points. If the strain gradient abruptly exceeds the first gradient threshold, it is marked as needing verification. The first gradient threshold is determined by analyzing the strain distribution law of this type of segment under normal stress conditions, that is, by statistically analyzing multiple sets of strain data under normal working conditions and taking the 95th percentile of the strain gradient as the threshold. Calculate the ratio of the mid-span displacement to the theoretical value of the mid-span displacement simulated by the initial digital twin model. If the ratio is greater than the set ratio range [R], it is considered a failure. min ,R max If ], then it is marked as requiring verification, where R min R is the lower limit of the ratio range. max The ratio range is the upper limit; the theoretical value of mid-span displacement is calculated by inputting the load parameters and material parameters of the segments into the initial digital twin model, R. min and R max The simulation error range is determined based on the initial digital twin model, for example, [0.85, 1.15]. The strain and displacement data, after being filtered from the unmarked Class B segments requiring verification, are marked as the reliability strain data ε for Class B segments. reli B(t) and reliability displacement data δε reli B(t).

[0083] Temperature and preload data after screening for Class C segments were verified separately: the temperature difference ΔT was calculated as the difference between the temperature sensor readings on both sides of the tooth groove. slot C, if the temperature difference ΔT slot C duration T dura Greater than the temperature difference threshold T dura If thre is selected, an abnormal temperature gradient is marked; the temperature difference threshold is determined based on the temperature stress limit of the toothed concrete material. This is achieved by calculating the tensile stress in the concrete caused by the temperature gradient, and using the temperature difference at which the tensile stress approaches the material's tensile strength as the temperature difference threshold; the difference in preload force ΔF between bolts in the same group is calculated. bolt group, if the preload difference ΔF bolt group is greater than the preload difference threshold F diff If the preload is abnormal, it is marked as such. Bolts in the same group refer to symmetrically distributed bolts with the same tooth structure. The difference in bolt preload is the maximum difference in the monitored preload values ​​of bolts in the same group. The preload difference threshold is determined based on the requirement for uniform stress distribution in the bolt connection; for example, 10% of the bolt preload design value is taken as the allowable difference range. Temperature and preload data from C-class segments that are not marked as having temperature gradient or preload anomalies are then labeled as the reliability temperature data (T) for C-class segments. reli C(t) and reliability preload data F reli C(t).

[0084] Step S23 corrects other data for temperature effects based on the reliability temperature data of Class C segments. The principle of temperature correction is that the strain caused by temperature changes in materials is positively correlated with the amount of temperature change and the material's linear expansion coefficient. For the strain data of Class A and Class B segments, the correction formula is: monitored strain value - (temperature change × material linear expansion coefficient), where the temperature change is the difference between the current monitored temperature and the initial temperature, the initial temperature is the temperature at which the segment was installed, and the material linear expansion coefficient is determined according to the material properties of steel or concrete. The monitored strain value includes the monitored strain values ​​of Class A segments and Class B segments, obtained from the reliability strain data of Class A segments and Class B segments. For displacement data, the displacement caused by temperature is calculated by multiplying the temperature change, segment length, and linear expansion coefficient. Subtracting the temperature-induced displacement from the monitored displacement value yields the corrected displacement data. The monitored displacement values ​​of Class A segments and Class B segments are obtained from the reliability displacement data of Class A segments and Class B segments. For the preload data of segment C, temperature changes cause the bolts to expand and contract. The correction method is to calculate the temperature correction amount of the preload based on the linear expansion coefficient of the bolt material and the amount of temperature change, and then add it to the monitored preload value to obtain the corrected preload data. Through the above process, the corrected original monitoring data stream is obtained, which includes the corrected strain and displacement data of segments A and B, and the corrected preload data of segment C.

[0085] The initial screening in step S20 identifies and removes abnormal data through statistical characteristics, reducing the impact of sensor failures or environmental interference on subsequent analysis and increasing the proportion of effective signals in the data. Secondary verification employs differentiated judgment criteria for the stress characteristics of different segments. For segment A, the strain-displacement difference verification utilizes structural symmetry to ensure data reliability; for segment B, the strain gradient-to-theoretical ratio verification combines local change rate and overall simulation deviation; and for segment C, the temperature difference-preload difference verification focuses on the environment and stress uniformity of the connection points. This categorized verification method makes data reliability judgment more closely aligned with the physical nature of each segment, reducing the data misjudgment rate compared to a unified verification standard. Temperature influence correction uses temperature data from segment C to calibrate the strain and displacement of segments A and B, eliminating non-stress deformation interference caused by temperature changes. This makes the corrected strain data closer to the actual stress state of the structure, and the displacement data more accurately reflects the deformation caused by the load. The initial screening reduced the amount of invalid data processed for secondary verification, which in turn provided reliable basic data for temperature correction. Temperature correction further eliminated systematic errors caused by environmental factors. The data processing chain formed by these three processes ensured that the final corrected data stream simultaneously possessed the characteristics of low noise, high reliability, and high fidelity. The temperature data of the C-segment was not only used for its own verification but also served as the correction benchmark for the A and B-segment data, achieving cross-segment data correlation optimization. Through temperature correction, monitoring data under different ambient temperatures became comparable, providing a consistent benchmark for long-term performance evaluation. Step S20 helps improve the accuracy of subsequent digital twin model calibration and the reliability of data-based risk assessment and tensioning strategy formulation.

[0086] Step S30: Compare the corrected original monitoring data stream with the simulation results of the initial digital twin model of the prefabricated segment to determine whether a calibration signal is triggered. If a calibration signal is triggered, perform dynamic calibration of the initial digital twin model to generate an accurate digital twin model of the prefabricated segment.

[0087] Please see Figure 2 As shown, step S30 further includes:

[0088] Step S31: Compare the corrected original monitoring data stream with the simulation results of the initial digital twin model of the prefabricated segment to determine whether the Class A segment strain calibration signal, Class A segment displacement calibration signal, Class B segment strain calibration signal, Class B segment displacement calibration signal and Class C segment connection calibration signal are triggered.

[0089] Step S32: If any of the following signals is triggered: Type A segment strain calibration signal, Type A segment displacement calibration signal, Type B segment strain calibration signal, Type B segment displacement calibration signal, and Type C segment connection calibration signal, then the trigger calibration signal is determined, and the initial digital twin model dynamic calibration is performed to obtain the accurate digital twin model of the prefabricated segment; otherwise, the initial digital twin model remains unchanged.

[0090] Specifically, the process of comparing the corrected original monitoring data stream with the simulation results of the initial digital twin model of the precast segment to determine the deviation and generate an accurate digital twin model includes: First, comparing the corrected original monitoring data stream with the simulation results of the initial digital twin model. The simulation results of the initial digital twin model refer to the theoretical values ​​calculated based on the initial digital twin model constructed in step S14, with input design loads, boundary conditions, and material property parameters, including strain simulation data and displacement simulation data of Class A segments, strain simulation data and displacement simulation data of Class B segments, and preload simulation data of Class C segments.

[0091] During the comparison process, the deviation values ​​for various segments are calculated: For segment A, the strain deviation of segment A is obtained by subtracting the strain simulation data of segment A from the corrected strain data; the displacement deviation of segment A is obtained by subtracting the displacement simulation data of segment A from the corrected displacement data of segment A. The strain deviation reflects the degree of deviation between the actual force and the theoretical calculation, while the displacement deviation reflects the difference between the actual deformation and the theoretical deformation. When the absolute value of the strain deviation of segment A exceeds the first strain deviation threshold, the strain calibration signal for segment A is triggered; when the absolute value of the displacement deviation exceeds the first displacement deviation threshold, the displacement calibration signal for segment A is triggered. The determination of the first strain deviation threshold is based on the allowable stress error range of the material. The ultimate tensile strain of the segmental bridge deck material is obtained through material mechanics tests, and 5% of it is taken as the threshold. For example, the ultimate tensile strain of a C50 concrete bridge deck is 200με, then the first strain deviation threshold is set to 10με. The first displacement deviation threshold is determined according to the maximum allowable deformation value in the design, which is the product of the segment length and the allowable deflection limit. The allowable deflection limit is calculated from the span ratio specified in the specification. The specified span ratio can be, for example, 1:500.

[0092] For Class B segments, the strain deviation is obtained by subtracting the strain simulation data from the corrected strain data of the Class B segment. If the absolute value of the strain deviation of the Class B segment is greater than the preset second strain deviation threshold, a strain calibration signal for the Class B segment is triggered. Similarly, the displacement deviation is obtained by subtracting the displacement simulation data from the corrected displacement data of the Class B segment. If the absolute value of the displacement deviation is greater than the preset second displacement deviation threshold, a displacement calibration signal for the Class B segment is triggered. The second strain deviation threshold is determined based on the fatigue strain limit of the main beam material. The allowable strain fluctuation range of this type of segment main beam within its design life is obtained through fatigue testing, and 10% of this range is taken as the threshold. The second displacement deviation threshold is 8% of the mid-span design deflection, which is determined by structural calculations based on the span length and load level. For Class C segments, the preload deviation is obtained by subtracting the preload simulation data from the corrected preload data. If the absolute value of the preload deviation exceeds a preset preload deviation threshold, a Class C segment connection calibration signal is triggered. The preload deviation threshold is determined based on the safety factor of the bolt connection. The safety factor is determined by the ratio of the tensile strength of the bolt material to the design preload. The allowable preload deviation range is derived from the safety factor; for example, if the design preload is 100kN, the preload deviation threshold is set to 10kN.

[0093] When any signal is triggered, a calibration signal is determined, and initial digital twin model dynamic calibration is performed. If a Class A segment strain calibration signal is triggered, the initial combined section stiffness parameter EIA is adjusted. The adjustment method is based on the linear relationship between the strain deviation of the Class A segment and EIA, where strain is inversely proportional to stiffness. The required EIA correction is calculated, and iterative calculations are used to ensure that the deviation between the simulated strain corresponding to the corrected EIA and the reliability strain data is within a threshold. If a Class A segment displacement calibration signal is triggered, the number or diameter of shear studs in the three-dimensional spatial model of shear stud arrangement is adjusted to reduce displacement deviation by increasing connection stiffness. If a Class B segment strain calibration signal is triggered, the bending stiffness parameter EIB of the main beam is corrected. Based on the inverse relationship between the strain deviation of the Class B segment and EIB, an increase in EIB results in a decrease in strain. EIB is gradually adjusted until the simulated strain meets the requirements. If a Class B segment displacement calibration signal is triggered, the profiled steel plate bridge deck combined section parameter SB is optimized by increasing the moment of inertia of the combined section and adjusting the thickness of the profiled steel plate to reduce mid-span deflection. For Class C segments, if a Class C segment connection calibration signal is triggered, adjust the tooth connection stiffness K. C Based on the preload deviation of segment C and K C The positive correlation between K and C Increasing the preload transmission efficiency improves K by modifying the bolt preload design parameters or the tooth groove geometry parameter GC. CThe corresponding deviations between the simulated preload and the reliability preload data are within the threshold. After multiple rounds of iterative calibration, until all deviations are within the corresponding thresholds, an accurate digital twin model is obtained; when no calibration signal is triggered, the initial digital twin model remains unchanged.

[0094] Step S40: Based on the corrected original monitoring data stream, conduct a multi-dimensional robustness index assessment and construct a robustness index feature vector.

[0095] Please see Figure 3 As shown, step S40 further includes:

[0096] Step S41: Based on the corrected strain and displacement data of Class A segments in the corrected original monitoring data stream, conduct a crack risk assessment and deformation compatibility assessment in the negative bending moment zone, and generate a comprehensive risk status record table for Class A segments.

[0097] Specifically, crack risk assessment needs to be based on the strain data of Class A segments, establishing a crack risk determination system that includes low-risk and high-risk strain thresholds. The low-risk strain threshold is 50% of the ultimate tensile strain of the bridge deck concrete material, obtained through axial tensile tests of concrete, by multiplying the average ultimate tensile strain of concrete specimens of the same strength grade by 50%. The high-risk strain threshold is 80% of the ultimate tensile strain, similarly calculated from test data. When the strain value corresponding to the strain data of Class A segments is less than the low-risk strain threshold, the crack risk is classified as low-risk; when the strain value corresponding to the strain data of Class A segments is greater than or equal to the low-risk strain threshold and less than the high-risk strain threshold, the crack risk is classified as low-risk; when the strain value corresponding to the strain data of Class A segments is greater than or equal to the high-risk strain threshold, it is classified as high-risk. Deformation compatibility assessment is based on the corrected displacement data of Class A segments. The relative rotation angle between the two ends of the segment is calculated. The relative rotation angle is the ratio of the displacement difference between the two ends of the segment to the segment length, where the displacement difference is the difference between the values ​​monitored by displacement sensors at both ends, and the segment length is extracted from the precise digital twin model. The rotation angle threshold is determined according to the maximum allowable torsional angle for this type of segment in the structural design code. This value is obtained by calculating the maximum allowable torsional angle of the segment under the design load using finite element simulation of the precise digital twin model. When the relative rotation angle is greater than the rotation angle threshold, it is marked as a deformation incompatibility state. The crack risk level and deformation compatibility state are integrated to generate a comprehensive risk status record table for Class A segments that includes both crack risk level and deformation compatibility state.

[0098] Step S42: Based on the corrected strain and displacement data of Class B segments in the corrected original monitoring data stream, perform stress level assessment and linear state assessment in the mid-span area, and generate a comprehensive state record table for Class B segments.

[0099] Stress level assessment requires calculating the ratio of the actual stress value to the design allowable stress. The design allowable stress is determined based on the yield strength and safety factor of the main beam material. The safety factor is selected according to bridge design specifications. For example, for Q355 steel, the yield strength is 355 MPa, the safety factor is 1.2, and the design allowable stress is 296 MPa. When the ratio of the actual stress value corresponding to the strain data after correction for Class B segments to the design allowable stress is greater than the warning ratio, it is marked as a stress warning state. The warning ratio is determined based on material fatigue performance tests and can be set to 90%. When it is greater than 90%, the material fatigue life decreases significantly. Alignment status assessment requires calculating alignment deviation, which is the difference between the displacement data after correction for Class B segments and the design deflection value. The design deflection value is extracted from the simulation results of the accurate digital twin model. The alignment deviation threshold is determined by comparing the correlation data between alignment deviations of multiple sets of actual bridges and structural safety, and is usually set to 10% of the design deflection value. When the absolute value of the alignment deviation is greater than the alignment deviation threshold, it is marked as an alignment adjustment state. The stress level assessment results and the alignment status assessment results are integrated to form a comprehensive status record table for Class B segments that includes both stress level status and alignment status.

[0100] Step S43: Based on the corrected preload data and reliability temperature data of Class C segments in the corrected original monitoring data stream, perform connection reliability assessment and temperature stability assessment, and generate a comprehensive status record table for Class C segments.

[0101] Connection reliability assessment requires calculating the preload ratio, which is the ratio of the preload value in the corrected preload data to the bolt preload design value, extracted from a precise digital twin model. The reliability threshold is determined based on bolt connection tests and is generally set to 0.9. When the preload ratio is less than the reliability threshold, the connection is marked as unreliable. Temperature stability assessment requires analyzing the spatiotemporal distribution characteristics of the temperature gradient. The temperature gradient is the ratio of the temperature difference between the two sides of the tooth groove to the tooth groove contact depth. The second gradient threshold is determined through concrete temperature stress calculation; the gradient value when the tensile stress in the concrete caused by the temperature gradient approaches its tensile strength is the second gradient threshold. The duration is determined by statistically analyzing the shortest time for cracks to appear in the structure after the temperature gradient exceeds the second threshold. For example, in a certain project, statistics showed that microcracks would appear after a duration greater than 2 hours; therefore, the preset duration is 2 hours. When the temperature gradient exceeds the second threshold and the duration exceeds the preset duration, it is marked as a temperature unstable state. The connection reliability assessment results and temperature stability assessment results are integrated to form a Class C segment comprehensive status record table containing both connection reliability and temperature stability states.

[0102] Step S44: Based on the Class A segment comprehensive risk status record table, the Class B segment comprehensive status record table, and the Class C segment comprehensive status record table, construct a robustness indicator feature vector.

[0103] Specifically, the crack risk level and deformation compatibility status in the Class A segment comprehensive risk status record table are vectorized and encoded. For crack risk level, low risk is coded as 00, medium risk as 01, and high risk as 10. For deformation compatibility status, compatibility is coded as 0 and incompatibility as 1. These two are combined to form the Class A segment feature vector φA; for example, low risk and compatibility are coded as 000. The stress level and alignment status in the Class B segment comprehensive status record table are vectorized and encoded. Normal stress level is coded as 0 and warning as 1. Normal alignment status is coded as 0 and adjustment required as 1. These are combined to form the Class B segment feature vector φB; for example, stress warning and alignment requiring adjustment are coded as 11. The connection reliability status and temperature stability status in the Class C segment comprehensive status record table are vectorized and encoded. Reliable connection is coded as 0 and unreliable as 1. Stable temperature is coded as 0 and unstable as 1. These are combined to form the Class C segment feature vector φC; for example, reliable connection and unstable temperature are coded as 01. The robustness index feature vector Φ = [φA, φB, φC] is constructed by concatenating the feature vectors of segment A, segment B, and segment C.

[0104] In step S40, the evaluation indicators for each segment precisely correspond to the segment characteristics. For segment A, the crack risk assessment is directly related to the crack resistance requirements of the bridge deck in the negative bending moment zone, while the deformation compatibility assessment reflects the overall stress balance of the segment. The combination of these two assessments provides a comprehensive view of the risk status of segment A. For segment B, the stress level assessment focuses on the load-bearing capacity of the mid-span main girder, while the alignment status assessment focuses on the overall structural alignment accuracy. The two assessments work together to achieve dual control over the stress and deformation in the mid-span region. For segment C, the connection reliability assessment targets the bolt preload, a core connection parameter, while the temperature stability assessment considers the impact of environmental factors on connection performance. The combination of these two assessments ensures a comprehensive judgment of the node connection status. Step S40 works in conjunction with the corrected original monitoring data stream from step S20. The corrected data eliminates temperature interference and outlier effects, making the evaluation results closer to the actual structural state. Combined with the precise digital twin model from step S30, the design parameters provided by the precise digital twin model provide a benchmark for the evaluation, improving its accuracy. The construction of robustness index feature vectors transforms various state information into quantifiable vector forms, providing structured input data for the subsequent tensioning strategy generation in step S50. This allows the adaptive tensioning strategy generation process to directly relate to the actual risk state of the structure. Without this step, the subsequent tensioning strategy would lack a clear risk orientation, potentially leading to a disconnect between the strategy and the actual structural needs. For example, targeted tensioning measures might not be taken for high-risk crack areas, or adjustments might not be made to segments with inconsistent deformation. In summary, step S40, through multi-dimensional evaluation and feature quantification, achieves a precise mapping from monitoring data to structural state. Its synergy with the previous data processing and model building steps creates a closed loop in the entire prestressing optimization process. This not only improves the reliability of risk assessment but also lays a data foundation for subsequent strategy generation. Furthermore, the feature vector encoding method allows different types of state information to be processed uniformly, enabling comprehensive analysis across segments.

[0105] Step S50: Based on the robustness index feature vector, construct the tension control unit and establish the control unit association graph. Combined with the accurate digital twin model of the prefabricated segment, construct and train the graph neural network tension decision model to generate an adaptive tension strategy sequence.

[0106] Further, step S50 includes:

[0107] Step S51: Extract prestressed steel strand parameters from the design BIM model and construct a tension control unit by combining segment classification and identification data;

[0108] Specifically, prestressed steel strand parameters are extracted from the design BIM model, including the spatial coordinates of the prestressed steel strands, strand number, strand specifications, and design tension value. The spatial coordinates of the prestressed steel strands are obtained through the three-dimensional coordinate system of the BIM model, accurate to the millimeter level. The strand numbers are assigned according to the numbering rules of the design drawings, such as W for web strands and T for top strands. The strand specifications include diameter and material strength grade, such as φ15.2mm steel strand corresponding to a tensile strength of 1860MPa. The design tension value is determined based on the stress calculation of the steel strands and is specified by the design documents, for example, the design tension value of a certain web strand is 1500kN. Based on segment classification and identification data, a unique segment number is assigned to each Class A, Class B, and Class C segment. The numbering rule is the first letter of the segment type plus a serial number, such as A01, B03, and C02, forming a segment number sequence. According to the prestressed steel strand parameters, each prestressed steel strand is spatially matched with the segment number sequence. The collision detection function of the design BIM model is used to determine the segment range where the steel strand is located, and to determine the segment type and segment number of each steel strand. Based on the preset tensioning levels, such as initial tensioning, intermediate tensioning, and final tensioning, the steel strands in each segment are grouped according to the tensioning level. Initial tensioning is for steel strands with temporary fixing requirements, intermediate tensioning is used to adjust the structural alignment, and final tensioning achieves the design stress state. Each group forms a tension control unit composed of segment number and tensioning level, such as A01-initial and B03-final.

[0109] Step S52: Based on the tension control unit and the robustness index feature vector Φ, establish the control unit association diagram;

[0110] Specifically, the temporal dependencies of different tensioning levels within the same segment are analyzed. Initial tensioning must be completed before intermediate tensioning, and intermediate tensioning must be completed before final tensioning, forming an irreversible sequence. The spatial coupling relationships between adjacent segments are also analyzed; the tensioning of steel strands in adjacent segments will affect each other. For example, tensioning segment A01 will cause additional stress in segment A02. Based on the classification of the three segment types, the structural correlations between different segments are identified. Segments of the same type have strong correlations due to similar stress characteristics; for example, A01 and A02 are both segments in the negative bending moment region, and their tensioning significantly affects each other. Segments of different types, such as A01 and B03, have weak correlations due to different stress mechanisms. A control unit association graph is constructed with tension control units as nodes. Temporal dependencies form directed edges, with arrows pointing to subsequent tensioning levels, such as A01-initial → A01-intermediate. Spatial coupling relationships form undirected edges, connecting tension control units of adjacent segments, such as A01-final - B01-final. Edge weights are determined by the spatial distance between segments and structural correlation; the closer the distance and the stronger the correlation, the greater the weight. For example, A01 and A02 are 5 meters apart and of the same type, so the weight is set to 0.8; A01 and B03 are 20 meters apart and of different types, so the weight is set to 0.3. The robustness index feature vector Φ is used as the global state feature of the control unit association graph, assigning a current risk state attribute to each node, such as the A01-final node being associated with a high-risk crack attribute.

[0111] For example, such as Figure 4 As shown, the circles represent the tension control units, including A01-initial, A01-intermediate, A01-final, A02-initial, B01-initial, and A02-final.

[0112] Table 1 shows the... Figure 4 The directed edges in the diagram are explained in detail.

[0113] Table 1 Figure 4 Detailed description of directed edges in the control unit association graph

[0114]

[0115] Table 2 pairs Figure 4 The undirected edges in the diagram are described in detail.

[0116] The tension control unit forms a control unit association graph using directed edges representing temporal dependencies and undirected edges representing edge weights.

[0117] Table 2 Figure 4 Detailed description of undirected edges in the control unit association graph

[0118]

[0119] Step S53: Based on the control unit association graph and robustness indicator feature vector, and combined with the precise digital twin model, construct and train a graph neural network tensor decision model;

[0120] The graph neural network tensioning decision model uses the control unit association graph as the input structure and robustness index feature vectors as node features. The tensioning decision objective function is defined, including minimizing crack risk, minimizing alignment deviation, and maximizing connection reliability. In the crack risk minimization objective function, segments with higher crack risk levels have greater weights. Alignment deviation minimization uses the sum of squared differences between the actual and designed alignment as the indicator. Connection reliability maximization uses the closeness of the preload ratio to the reliability threshold as the indicator. Tensioning constraints are set: the single tensioning force does not exceed the designed tensioning force value to avoid over-tensioning and breakage of the steel strand; the displacement difference between adjacent segments does not exceed the ratio range [R]. min ,R max To prevent excessive displacement differences from causing connection damage, a precise digital twin model is used as the simulation environment. Inputting the current tension state parameters outputs the structural response, such as strain, displacement, and preload changes. The parameters of the graph neural network tension decision model are trained using a reinforcement learning algorithm: after initializing the model parameters, the model outputs a tension decision in each training round. The digital twin model returns the structural response and objective function value under that decision. If the objective function value increases, the parameters are adjusted to enhance the probability of that decision; conversely, they are decreased. Iterative training continues until the objective function converges; for example, the objective function value fluctuation is less than 5% after 100 consecutive training rounds.

[0121] Step S54: Obtain the robustness index feature vector under the current state, and generate an adaptive tensioning strategy sequence using the trained graph neural network tensioning decision model.

[0122] The robustness indicator feature vector under the current state is obtained. This vector is then expanded using the state of the tensioned control unit and the initial state of the untensioned control unit to obtain the expanded robustness indicator feature vector. For example, the state of the tensioned control unit is A01 - Initially tensioned, corresponding to a tension force of 1000kN; the initial state of the untensioned control unit is A01 - Ultimately not tensioned, with an initial tension force of 0kN. The robustness indicator feature vector essentially provides a quantitative description of the structure's current risk state, such as the crack risk of Class A segments, the linear deviation of Class B segments, and the connection reliability of Class C segments, but it does not include dynamic information about the tensioning process. The generation of the tensioning strategy depends on both the structural risk status and the tensioning implementation progress. The completion status of tensioned units affects the stress environment of untensioned units. Therefore, two types of information need to be added to the robustness index feature vector: first, the status of tensioned control units, including their segment number, tensioning level, actual tensioning force value, and corresponding structural response changes, such as the reduction in crack risk after tensioning; second, the initial status of untensioned control units, including their preset tensioning level, design tensioning force value, and the risk status of the currently associated segments, whether the corresponding segment is in a high-risk state. The robustness index feature vector under the current state incorporates the status information of tensioned control units and the initial state information of untensioned control units to form a complete feature vector under the current state, i.e., the expanded robustness index feature vector. The extended robustness index feature vector still uses φA, φB, and φC as the core framework, with added information, namely the state of the tensioned control unit and the initial state of the untensioned control unit, embedded as additional dimensions. For example, the identifier of the corresponding tensioned unit segment is added to φA to ensure consistency with the input structure when training the graph neural network.

[0123] The expanded robustness index feature vector is input into the trained graph neural network (GNN) tensioning decision model. The GNN extracts node association features through the GNN layers and outputs the tensioning decision, i.e., the next control unit to be tensioned and its tension force value. A precise digital twin model is used to predict the structural response after the tensioning decision is executed, verifying whether the tensioning constraints set during the training of the GNN tensioning decision model are met. If met, the tensioning decision is retained; otherwise, a new tensioning decision is generated. This process is iteratively executed until a complete tensioning strategy sequence covering all tensioning control units is generated, serving as the adaptive tensioning strategy sequence.

[0124] Step S55: Perform closed-loop execution based on the adaptive tensioning strategy sequence, and make dynamic corrections based on real-time feedback data.

[0125] Tensioning operations are executed sequentially according to an adaptive tensioning strategy sequence. After tensioning each tensioning control unit is completed, real-time strain, displacement, and preload data are acquired through a monitoring system to update the robustness indicator feature vector. For example, after the final tensioning of a segment (Type A), its crack risk level changes from high to medium risk. Based on the updated robustness indicator feature vector, it is determined whether the subsequent tensioning strategy needs adjustment: if the deviation between the real-time risk status of a certain segment and the prediction result of the graph neural network tensioning decision model exceeds a preset threshold (e.g., the model predicts low risk while actual monitoring shows medium risk), the updated robustness indicator feature vector is re-inputted into the trained graph neural network tensioning decision model to generate a new subsequent tensioning strategy. After completing the tensioning operations of all tensioning control units, the tensioning effect is evaluated based on the full-process monitoring data, including the reduction in crack risk level and the degree of control over linear deviation. The parameters of the graph neural network tensioning decision model are optimized based on the evaluation results, such as adjusting the weights of indicators like crack risk, linear deviation, and connection reliability in the objective function.

[0126] In step S50, the combination of prestressed steel strand parameters and segment classification gives the tension control unit a clear structural focus, avoiding local stress imbalances caused by traditional tensioning in a uniform sequence. The control unit's correlation graph simultaneously incorporates temporal dependencies and spatial coupling relationships, enabling decisions to consider both construction sequence constraints and the overall structural stress correlation. Compared to tensioning strategies that only consider a single segment, this reduces the generation of secondary stresses across segments. The synergy between graph neural networks-reinforcement learning methods and precise digital twin models achieves a dynamic mapping from structural state to tensioning decisions: the graph neural network's ability to handle complex relationships allows the model to identify implicit effects across segments, such as the indirect effect of tensioning a segment of type A on the preload of a node of type C. Reinforcement learning continuously optimizes the strategy through simulation feedback from the digital twin model, overcoming the shortcomings of traditional empirical tensioning in dealing with dynamic structural changes. The closed-loop execution mechanism of the adaptive tensioning strategy sequence allows the strategy to adjust according to real-time conditions, avoiding strategy failures caused by construction errors or material property fluctuations. For example, when the actual stiffness of a segment is less than the model prediction, the strategy will automatically reduce the tension force to control strain within a safe range. The robustness index feature vectors in steps S50 and S40 work together to provide precise risk guidance for tensioning strategy generation, prioritizing tensioning operations in high-risk areas such as Class A segments with high crack risk. This, combined with the precise digital twin model in step S30, ensures the effectiveness of reinforcement learning training by providing accurate simulation results, making the generated tensioning strategy feasible in practice. Without this step, prestressing tensioning would rely on a fixed process, unable to adjust according to the real-time state of the structure, potentially leading to uncontrolled high-risk areas or over-tensioning causing new damage. This invention demonstrates significant effectiveness in the collaborative optimization of different segment types; for example, adjusting the tensioning sequence of Class B segments can indirectly improve the deformation coordination of Class A segments. This cross-segment correlation control mechanism is difficult to achieve in traditional tensioning strategies. Through mutual feedback and dynamic adaptation of the stress states of multiple segment types, it significantly improves the overall robustness of the structure.

[0127] Example 2

[0128] This embodiment, based on Embodiment 1, provides a prestressing tensioning parameter optimization system for precast segments of steel-concrete composite beam bridges, such as... Figure 5 Shown, including:

[0129] Segment classification module: used to obtain the design BIM model and classify the prefabricated segments based on the design BIM model;

[0130] Data acquisition module: Based on the classification results, construct the initial digital twin model of the prefabricated segment and set up monitoring points to acquire the raw monitoring data stream;

[0131] The data processing module is used to filter, verify, and correct the raw monitoring data stream to obtain the corrected raw monitoring data stream.

[0132] Model calibration module: Compares the corrected original monitoring data stream with the simulation results of the initial digital twin model of the prefabricated segment to determine whether a calibration signal is triggered. If a calibration signal is triggered, dynamic calibration of the initial digital twin model is performed to generate an accurate digital twin model of the prefabricated segment.

[0133] Robustness Index Assessment Module: Based on the corrected original monitoring data stream, multi-dimensional robustness index assessment is performed, and a robustness index feature vector is constructed;

[0134] Strategy generation module: Based on the robustness index feature vector, a tension control unit is constructed and a control unit association graph is established. Combined with the accurate digital twin model of the prefabricated segment, a graph neural network tension decision model is constructed and trained to generate an adaptive tension strategy sequence.

[0135] In the robustness indicator assessment module, the specific process of constructing the robustness indicator feature vector includes:

[0136] Step S41: Based on the corrected strain and displacement data of Class A segments in the corrected original monitoring data stream, conduct a crack risk assessment and deformation compatibility assessment in the negative bending moment zone, and generate a comprehensive risk status record table for Class A segments.

[0137] Step S42: Based on the corrected strain and displacement data of Class B segments in the corrected original monitoring data stream, perform stress level assessment and linear state assessment in the mid-span area, and generate a comprehensive state record table for Class B segments.

[0138] Step S43: Based on the corrected preload data and reliability temperature data of Class C segments in the corrected original monitoring data stream, perform connection reliability assessment and temperature stability assessment, and generate a comprehensive status record table for Class C segments.

[0139] Step S44: Based on the Class A segment comprehensive risk status record table, the Class B segment comprehensive status record table, and the Class C segment comprehensive status record table, construct a robustness indicator feature vector.

[0140] In the strategy generation module, the specific process of constructing a tension control unit, establishing a control unit association graph, and combining it with a precise digital twin model of the prefabricated segment to construct and train a graph neural network tension decision model to generate an adaptive tension strategy sequence includes:

[0141] Step S51: Extract prestressed steel strand parameters from the design BIM model and construct a tension control unit by combining segment classification and identification data;

[0142] Step S52: Based on the tension control unit and the robustness index feature vector Φ, establish the control unit association diagram;

[0143] Step S53: Based on the control unit association graph and robustness indicator feature vector, and combined with the precise digital twin model, construct and train a graph neural network tensor decision model;

[0144] Step S54: Obtain the robustness index feature vector under the current state, and generate an adaptive tensioning strategy sequence using the trained graph neural network tensioning decision model.

[0145] Step S55: Perform closed-loop execution based on the adaptive tensioning strategy sequence, and make dynamic corrections based on real-time feedback data.

[0146] The methods and systems of this application may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the method is for illustrative purposes only, and the steps of the method of this application are not limited to the order specifically described above, unless otherwise specifically stated.

[0147] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0148] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges, characterized in that, The method includes: Obtain the design BIM model, classify the prefabricated segments based on the design BIM model, construct the initial digital twin model of the prefabricated segments based on the classification results, set up monitoring points, and obtain the original monitoring data stream; The classification of prefabricated segments based on the design BIM model includes: extracting the structural parameters of the steel-concrete composite beam bridge from the design BIM model; and classifying the prefabricated segments into three types according to the structural parameters, including type A segments (negative bending moment zone segments), type B segments (mid-span segments), and type C segments (connection node segments). The original monitoring data stream is filtered, verified, and corrected to obtain the corrected original monitoring data stream; the corrected original monitoring data stream includes the corrected stress and displacement data of Class A segments, the corrected stress and displacement data of Class B segments, and the corrected preload data of Class C segments. The corrected original monitoring data stream is compared with the simulation results of the initial digital twin model of the prefabricated segment to determine whether a calibration signal is triggered. If a calibration signal is triggered, dynamic calibration of the initial digital twin model is performed to generate an accurate digital twin model of the prefabricated segment. Based on the corrected original monitoring data stream, a multi-dimensional robustness index assessment is conducted, and a robustness index feature vector is constructed. The method for constructing the robustness indicator feature vector includes: based on the corrected stress and displacement data of Class A segments in the corrected original monitoring data stream, performing a crack risk assessment and deformation compatibility assessment in the negative bending moment zone, and generating a comprehensive risk status record table for Class A segments; based on the corrected stress and displacement data of Class B segments in the corrected original monitoring data stream, performing a stress level assessment and alignment status assessment in the mid-span zone, and generating a comprehensive status record table for Class B segments; based on the corrected preload data and reliability temperature data of Class C segments in the corrected original monitoring data stream, performing a connection reliability assessment and temperature stability assessment, and generating a comprehensive status record table for Class C segments; and constructing a robustness indicator feature vector based on the comprehensive risk status record table for Class A segments, the comprehensive status record table for Class B segments, and the comprehensive status record table for Class C segments; the reliability temperature data of the Class C segments is obtained through preliminary screening, adjacent comparison, and secondary verification of the original monitoring data stream. Based on the robustness index feature vector, a tension control unit is constructed and an association graph of the control unit is established. Combined with the accurate digital twin model of the prefabricated segment, a graph neural network tension decision model is constructed and trained to generate an adaptive tension strategy sequence.

2. The method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges according to claim 1, characterized in that, The initial digital twin model for constructing the prefabricated segment includes: Based on the segment type classification results, detailed construction parameters of each segment type are extracted to generate segment classification identification data; based on the segment classification identification data, an initial digital twin model of the prefabricated segment is constructed on the basis of the design BIM model.

3. The method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges according to claim 2, characterized in that, The method for obtaining the corrected original monitoring data stream includes: The raw monitoring data stream is initially screened to remove abnormal data, resulting in a filtered raw monitoring data stream. Adjacent comparisons and secondary verifications were performed on the filtered raw monitoring data streams to obtain the reliability strain data and reliability displacement data of Class A segments, the reliability strain data and reliability displacement data of Class B segments, and the reliability temperature data and reliability preload data of Class C segments. Based on the reliability temperature data of Class C segments, the reliability strain and displacement data of Class A segments, the reliability strain and displacement data of Class B segments, and the reliability preload data of Class C segments are corrected for temperature effects to obtain the corrected original monitoring data stream.

4. The method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges according to claim 3, characterized in that, The method for determining whether a calibration signal has been triggered includes: The corrected original monitoring data stream is compared with the simulation results of the initial digital twin model of the prefabricated segment to determine whether the Class A segment strain calibration signal, Class A segment displacement calibration signal, Class B segment strain calibration signal, Class B segment displacement calibration signal, and Class C segment connection calibration signal are triggered. If any of the Class A segment strain calibration signal, Class A segment displacement calibration signal, Class B segment strain calibration signal, Class B segment displacement calibration signal, and Class C segment connection calibration signal are triggered, then the calibration signal is determined to be triggered.

5. The method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges according to claim 4, characterized in that, The method for constructing the tension control unit includes: Prestressed steel strand parameters are extracted from the design BIM model; based on the segment classification and identification data, a unique segment number is assigned to each Class A, Class B, and Class C segment; Based on the prestressed steel strand parameters, each prestressed steel strand is spatially matched with the segment number sequence to determine the segment type and segment number to which each prestressed steel strand belongs; Based on the preset tensioning levels, the prestressed steel strands in each segment are grouped according to the tensioning levels to form a tensioning control unit consisting of segment numbers and tensioning levels.

6. The method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges according to claim 5, characterized in that, The method for establishing the control unit association diagram includes: Analyze the temporal dependencies of different tensioning level control units within the same segment to form an irreversible sequence; The spatial coupling relationship between adjacent segments was analyzed, and the structural correlation between different segments was identified based on the classification results of three types of segments. Using the tension control unit as a node, the temporal dependency relationship forms a directed edge, and the spatial coupling relationship forms an undirected edge, thus constructing the control unit association graph.

7. The method for optimizing the prestressing tensioning parameters of precast segments in steel-concrete composite beam bridges according to claim 6, characterized in that, The method for generating the adaptive tensioning strategy sequence includes: The robustness index feature vector in the current state is extended by using the state of the tensioned control unit and the initial state of the untensioned control unit to obtain the extended robustness index feature vector; the extended robustness index feature vector is input into the trained graph neural network tensioning decision model to output the tensioning decision, which is the next control unit to be tensioned and its tension force value; The structural response after the execution of the tensioning decision is predicted using a precise digital twin model. The results are then verified to see if the tensioning constraints set during the training of the graph neural network tensioning decision model are met. If they are met, the tensioning decision is retained; otherwise, the tensioning decision is regenerated. The above process is iteratively executed until a complete tensioning strategy sequence covering all tensioning control units is generated, which serves as the adaptive tensioning strategy sequence.

8. A prestressing tensioning parameter optimization system for precast segments of steel-concrete composite beam bridges, used to implement the prestressing tensioning parameter optimization method for precast segments of steel-concrete composite beam bridges as described in any one of claims 1-7, characterized in that, The system includes: Segment classification module: used to obtain the design BIM model and classify the prefabricated segments based on the design BIM model; Data acquisition module: Based on the classification results, construct the initial digital twin model of the prefabricated segment and set up monitoring points to acquire the raw monitoring data stream; The data processing module is used to filter, verify, and correct the raw monitoring data stream to obtain the corrected raw monitoring data stream. Model calibration module: Compares the corrected original monitoring data stream with the simulation results of the initial digital twin model of the prefabricated segment to determine whether a calibration signal is triggered. If a calibration signal is triggered, dynamic calibration of the initial digital twin model is performed to generate an accurate digital twin model of the prefabricated segment. Robustness Index Assessment Module: Based on the corrected original monitoring data stream, multi-dimensional robustness index assessment is performed, and a robustness index feature vector is constructed; Strategy generation module: Based on the robustness index feature vector, a tension control unit is constructed and a control unit association graph is established. Combined with the accurate digital twin model of the prefabricated segment, a graph neural network tension decision model is constructed and trained to generate an adaptive tension strategy sequence.

Images