A design method and system for the pylon cross-section of a long-span cable-stayed bridge based on big data
By using big data-driven methods, a mapping model between the pylon cross-sectional parameters and the main span length was constructed and combined with finite element analysis, which solved the problem of experience dependence in the design of cable-stayed bridge pylons and achieved efficient and accurate design results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN ROAD & BRIDGE CONSTRUCTION GROUP CO LTD
- Filing Date
- 2025-11-03
- Publication Date
- 2026-05-26
Smart Images
Figure CN121435342B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of general-purpose computers, and specifically relates to a design method and system for the cross-section of a long-span cable-stayed bridge tower based on big data. Background Technology
[0002] With the continuous advancement of transportation infrastructure construction, cable-stayed bridges are widely used in long-span bridge projects due to their beautiful structural form, strong span capacity, and reasonable stress distribution. Among them, the pylon, as one of the main load-bearing components of a cable-stayed bridge, directly affects the stress performance, material utilization efficiency, and construction feasibility of the entire bridge, making it a key aspect of bridge structural design.
[0003] In existing technologies, the design of cable-stayed bridge pylon sections typically relies on the experience of structural engineers and code guidelines, supplemented by static calculations and finite element simulations. Common methods include determining section dimensions based on empirical formulas, followed by adjustments through static verification and local structural optimization. In engineering practice, designers often perform parameter calculations iteratively, comprehensively considering material properties, cost, and construction factors while meeting code requirements. Some research has also attempted to introduce parametric modeling, optimization algorithms, and data fitting techniques to assist the design process and improve design efficiency and rationality.
[0004] However, the aforementioned traditional design methods have the following shortcomings. On the one hand, over-reliance on engineering experience leads to a lack of systematicity and repeatability in the design process, making it difficult to adapt to design requirements under complex conditions such as ultra-large spans and variable loads. On the other hand, design efficiency is limited by the complexity of trial calculations and iterations, making it difficult to achieve efficient response and rapid decision-making. Furthermore, the lack of systematic summarization of data patterns from a large number of existing bridges means that historical engineering achievements have not been fully transformed into design knowledge for future projects. Summary of the Invention
[0005] To address the problems in existing technologies, this invention provides a design method for the pylon cross-section of long-span cable-stayed bridges based on big data, comprising the following steps:
[0006] Obtain the design parameters of the target bridge, and determine the structural type of the target bridge based on the design parameters;
[0007] Multiple cable-stayed bridge instances with the same structural type as the target bridge are selected from the completed bridge design data. The cable-stayed bridge instances are divided into multiple structural levels according to the main span length. The structural levels include micro-small, small, medium, large and super-large. Each structural level contains at least three completed cable-stayed bridges.
[0008] Based on the main span length and corresponding tower section parameters of the cable-stayed bridge example, a mapping model of the tower section parameters changing with the main span length is constructed. The mapping model establishes a quantitative relationship between the tower section parameters and the structural scale using a power function, and establishes a coupling relationship between different tower section parameters to form a parameter evolution model.
[0009] When the main span length of the target bridge is close to the boundary of the adjacent structural level, a level transition factor is introduced to construct a smooth transition mechanism between adjacent structural levels and avoid parameter abrupt changes.
[0010] Based on the main span length of the target bridge, the initial cable tower section parameters are calculated using the parameter evolution model.
[0011] Construct a finite element model of the tower of the target bridge, input the design conditions of the target bridge, perform multi-condition structural response analysis on the initial tower section parameters, and determine whether the initial tower section parameters meet the preset design constraints.
[0012] If the initial tower section parameters do not meet the design constraints, historical design error samples at the structural level are extracted, a residual correction model is constructed, the corrected tower section parameters are calculated based on the correction model, and parameter trimming is performed in combination with material properties, construction technology and structural constraints to obtain the final recommended design parameters of the tower section that meet the design constraints.
[0013] Furthermore, a clustering analysis algorithm is used to adaptively group all the main span lengths of the bridges, automatically determining the span range boundaries for each level.
[0014] Furthermore, for cases where there are insufficient samples at the structural level, simulation design results or expert-modeled bridge types are introduced as supplementary samples.
[0015] Furthermore, in scenarios where sample density is insufficient or structural level division is uncertain, prediction models at three levels—the previous level, the current level, and the next level—are introduced simultaneously, and a comprehensive prediction model is constructed using weighted averaging or weighted regression.
[0016] Furthermore, the residual correction model is implemented using a gradient boosting tree-based regression framework.
[0017] In another aspect, the present invention provides a design system for the pylon cross-section of a long-span cable-stayed bridge based on big data, comprising the following modules:
[0018] The design parameter extraction module is used to obtain the design parameters of the target bridge and determine the structural type of the target bridge based on the design parameters;
[0019] The structural level filtering module is used to filter out multiple cable-stayed bridge instances with the same structural type as the target bridge from the completed bridge design data, and divide the cable-stayed bridge instances into multiple structural levels according to the main span length. The structural levels include micro-small, small, medium, large and super-large, and each structural level contains at least three completed cable-stayed bridges.
[0020] The parameter modeling module is used to construct a mapping model of the cable-stayed bridge example's main span length and corresponding tower section parameters, and to establish a quantitative relationship between the tower section parameters and the structural scale using a power function, and to establish a coupling relationship between different tower section parameters to form a parameter evolution model.
[0021] The transition control module is used to introduce a level transition factor when the main span length of the target bridge is close to the boundary of the adjacent structural level, to build a smooth transition mechanism between adjacent structural levels and avoid parameter abrupt changes.
[0022] The initial parameter calculation module is used to calculate the initial cable tower section parameters based on the main span length of the target bridge using the parameter evolution model.
[0023] The structural response analysis module is used to construct the finite element model of the tower of the target bridge, input the design conditions of the target bridge, perform multi-condition structural response analysis on the initial tower section parameters, and determine whether the initial tower section parameters meet the preset design constraints.
[0024] The parameter correction and trimming module is used to extract historical design error samples at the structural level when the initial cable tower section parameters do not meet the design constraints, construct a residual correction model, calculate the corrected cable tower section parameters based on the correction model, and trim the parameters in combination with material properties, construction technology and structural constraints to obtain the final recommended design parameters of the cable tower section that meet the design constraints.
[0025] Furthermore, a clustering analysis algorithm is used to adaptively group all the main span lengths of the bridges, automatically determining the span range boundaries for each level.
[0026] Furthermore, for cases where there are insufficient samples at the structural level, simulation design results or expert-modeled bridge types are introduced as supplementary samples.
[0027] Furthermore, in scenarios where sample density is insufficient or structural level division is uncertain, prediction models at three levels—the previous level, the current level, and the next level—are introduced simultaneously, and a comprehensive prediction model is constructed using weighted averaging or weighted regression.
[0028] Furthermore, the residual correction model is implemented using a regression framework based on gradient boosting trees.
[0029] This invention introduces a big data-based structural parameter evolution modeling method, which realizes the transformation of the design of the pylon section of long-span cable-stayed bridges from experience-driven to data-driven. It can fully explore and utilize the structural scale patterns and section parameter evolution trends in existing bridge engineering data, effectively improve the scientificity and rationality of parameter prediction, and significantly reduce the reliance on the subjective experience of designers.
[0030] This invention establishes a complete set of parameter reasoning mechanisms oriented towards engineering practice, including structural level classification, parameter mapping modeling, transition control, and residual correction. It enables multi-condition adaptation and rapid correction while meeting design specification constraints, improving design efficiency and the engineering adaptability of the results. Simultaneously, the introduction of a finite element model for multi-condition response analysis and iterative verification of design parameters facilitates pre-assessment and optimized control of structural performance in the early design phase. Attached Figure Description
[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments 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.
[0032] Figure 1 This is a flowchart of the method of the present invention;
[0033] Figure 2 This is a schematic diagram illustrating the evolution modeling principle of cable tower cross-section parameters in this invention;
[0034] Figure 3 This is a flowchart of the residual correction model construction and correction process in this invention. Detailed Implementation
[0035] The invention will now be described in preferred form with reference to the accompanying drawings and specific embodiments.
[0036] like Figure 1As shown, in one embodiment, this invention discloses a design method for the pylon section of a long-span cable-stayed bridge based on big data. The method aims to collect and analyze a large amount of data from existing cable-stayed bridge projects in a structured manner, uncovering the evolutionary relationship between the main span length of the bridge and the pylon section parameters. Combined with the design parameters of the target bridge, a scale mapping model and a parameter coupling model for section prediction are established. Through finite element analysis and residual correction mechanisms, the method achieves rapid determination of the pylon section parameters and verification of performance under multiple working conditions, thereby improving the scientific rigor, engineering adaptability, and intelligence level of the pylon section design. This method is applicable to the preliminary design and scheme comparison stages of pylon structures in new long-span cable-stayed bridge projects.
[0037] The method specifically includes the following steps:
[0038] Obtain the design parameters of the target bridge, and determine the structural type of the target bridge based on the design parameters;
[0039] Multiple cable-stayed bridge instances with the same structural type as the target bridge are selected from the completed bridge design data. The cable-stayed bridge instances are divided into multiple structural levels according to the main span length. The structural levels include micro-small, small, medium, large and super-large. Each structural level contains at least three completed cable-stayed bridges.
[0040] Based on the main span length and corresponding tower section parameters of the cable-stayed bridge example, a mapping model of the tower section parameters changing with the main span length is constructed. The mapping model establishes a quantitative relationship between the tower section parameters and the structural scale using a power function, and establishes a coupling relationship between different tower section parameters to form a parameter evolution model.
[0041] When the main span length of the target bridge is close to the boundary of the adjacent structural level, a level transition factor is introduced to construct a smooth transition mechanism between adjacent structural levels and avoid parameter abrupt changes.
[0042] Based on the main span length of the target bridge, the initial cable tower section parameters are calculated using the parameter evolution model.
[0043] Construct a finite element model of the tower of the target bridge, input the design conditions of the target bridge, perform multi-condition structural response analysis on the initial tower section parameters, and determine whether the initial tower section parameters meet the preset design constraints.
[0044] If the initial tower section parameters do not meet the design constraints, historical design error samples at the structural level are extracted, a residual correction model is constructed, the corrected tower section parameters are calculated based on the correction model, and parameter trimming is performed in combination with material properties, construction technology and structural constraints to obtain the final recommended design parameters of the tower section that meet the design constraints.
[0045] Through the above steps, existing cable-stayed bridge engineering case data are fully utilized. Samples are extracted according to structural type and bridge scale, and a power function mapping relationship between tower section parameters and bridge main span length is constructed. Parameter coupling modeling and level smooth transition mechanisms are introduced to effectively avoid the problems of parameter mutation and subjective experience dependence in traditional design. By combining residual correction models with construction and material constraints for adaptive adjustment, the accuracy of section parameter prediction and engineering adaptability are further improved. This enables rapid acquisition of tower section parameters, multi-condition performance verification, and intelligent optimization, significantly improving the efficiency, reliability, and intelligence level of long-span cable-stayed bridge structural design.
[0046] The following section will explain and illustrate the steps in the method in detail.
[0047] Obtain the design parameters of the target bridge, and determine the structural type of the target bridge based on the design parameters.
[0048] Obtaining the design parameters of the target bridge and determining its structural type accordingly is a crucial prerequisite for rapidly estimating cross-sectional parameters and establishing a preliminary design scheme. The fundamental principle is that the type of bridge structural system determines the force path of the towers within the overall structure, the cable arrangement, and structural constraints, directly influencing the selection range and evolution trend of tower cross-sectional dimensions. Therefore, before conducting data-driven design simulations, it is necessary to clearly define the structural category of the target bridge to ensure the comparability of subsequent data samples and avoid mapping distortion or estimation errors caused by inconsistencies in structural systems.
[0049] The design parameters of the target bridge refer to the set of core parameters directly related to the bridge's structural configuration and stress characteristics, including but not limited to the main span length, tower height, connection method between the towers and the bridge deck, cable-stayed bridge arrangement, and design load level. The structural type refers to the standard category of cable-stayed bridges in bridge engineering, classified according to their structural system. Common types include single-tower single-cable-stayed structures, double-tower double-cable-stayed structures, and multi-tower continuous structures. Determining the structural type depends not only on the geometric layout but also on the connection method between the towers and the main girder, the cable-stayed bridge anchorage method, and the bridge deck stiffness distribution characteristics.
[0050] In the specific implementation process, design parameters are first extracted from the initial design documents, project summary, or bridge layout drawings of the target project. Standard data extraction methods are then used for information structuring to obtain the basic design information set of the target bridge. During this process, key node parameters can be read through a bridge modeling platform or structural information database system and standardized for type matching with existing bridge samples in the database. The structural type is determined based on a multi-factor feature matching algorithm. Under the premise of ensuring basic consistency of design parameters, priority is given to identifying the spatial distribution of pylon type and cable arrangement, and this is further supplemented by the longitudinal stiffness distribution of the bridge for auxiliary determination.
[0051] Preferably, in the process of determining the structural type, a graph structure comparison method can be introduced in conjunction with the structural similarity recognition model. The structural topology of the target bridge is represented as a graph structure in the form of nodes and edges. A graph neural network is used to perform structural similarity analysis between the target bridge and the sample bridge, thereby improving the accuracy and robustness of structural classification in bridge types with complex node relationships.
[0052] The implementation of this step can significantly improve the relevance and effectiveness of bridge sample calls, ensure that the subsequent mapping relationship between cross-sectional parameters and structural scale has physical meaning and engineering applicability, avoid calculation deviations caused by differences in structural types, and enhance the data consistency and convergence speed of the design process.
[0053] For example, suppose a target bridge has a main span of 960 meters, a tower height of 210 meters, and adopts a double-tower, double-cable-stayed structure. The cables are arranged in a symmetrical fan shape, and the bridge deck is a monolithic steel box girder structure. The towers and the bridge deck are connected by bending hinges. Based on the analysis of design parameters and structural feature comparison, the bridge is determined to be a large-scale double-tower, double-cable-stayed symmetrical structure.
[0054] Multiple cable-stayed bridge examples with the same structural type as the target bridge are selected from the completed bridge design data. The cable-stayed bridge examples are divided into multiple structural levels according to the main span length. The structural levels include micro-small, small, medium, large and super-large. Each structural level contains at least three completed cable-stayed bridges.
[0055] Cable-stayed bridges of different spans exhibit measurable evolutionary trends in terms of stress characteristics, tower dimensions, and cable force distribution. Bridges of the same structural type share a high degree of consistency in structural layout and construction details, which facilitates the establishment of continuous mapping relationships of structural parameters among samples. Cross-scale hierarchical management ensures that sample data covers the entire scale range, enhancing the representativeness of data fitting and the accuracy of piecewise modeling.
[0056] The structural level refers to a classification method that discretizes the structural scale according to the main span length of the bridge. Different structural levels represent systematic changes in the bridge's dimensions, internal force levels, and structural complexity. The main span length is an important parameter for measuring the scale of a bridge's span and has clear physical meaning and engineering distinguishability as a classification criterion. The cable-stayed bridge examples refer to actual engineering bridge cases with completed structural designs and clearly recorded parameters. These are usually stored in a structural design database as a basic sample source for the big data-driven design process.
[0057] In the specific implementation process, firstly, based on the determined structural type of the target bridge, design data of all cable-stayed bridges with the same structural type are extracted from a structural engineering database. This database can be a bridge dataset built by the organization itself or a shared bridge structure sample library within the industry. For each extracted bridge instance, its main span length information is read and the structure is divided into structural levels according to a set span range. Preferably, the structural levels can be set into five categories: micro-small, small, medium, large, and super-large. The span range of each structural level can be set based on industry standards or big data clustering results. To ensure the effectiveness of the sample statistical analysis, at least three completed cable-stayed bridge instances are retained in each structural level.
[0058] To improve the accuracy of structural level classification, in one preferred implementation, a clustering analysis algorithm can be used to adaptively group all bridge main span lengths. The clustering algorithm can be either K-means or Gaussian mixture model, automatically determining the span range boundaries for each level, thus naturally matching the structural level classification with the sample distribution. In another preferred implementation, to address the issue of insufficient samples at the structural level, simulation design results or expert-modeled bridge types can be introduced as supplementary samples to improve the completeness of sample coverage at extreme scales.
[0059] The implementation of this step ensures that the parameter mapping model established subsequently has coverage, continuity and multi-scale adaptability, avoids local fitting errors caused by single-scale samples, and also provides a data foundation for achieving smooth interpolation and extrapolation of tower section design parameters between different bridge scales.
[0060] Continuing with the aforementioned target bridge, its structural type is a double-tower, double-cable-stayed symmetrical structure with a main span length of 960 meters. The system extracts all bridge samples with the same structural type from the bridge database, classifies their main span lengths by structural level, and obtains a span range of 800 to 1200 meters for large structures. Within this level, three cable-stayed bridge examples, including the Chongqing Egongyan Bridge, the Wuhan Yingwuzhou Yangtze River Bridge, and the Sutong Yangtze River Bridge, are selected as representative samples for the fitted parameter mapping model.
[0061] Based on the main span length and corresponding tower section parameters of the cable-stayed bridge example, a mapping model is constructed to show the variation of tower section parameters with the main span length. The mapping model establishes a quantitative relationship between tower section parameters and structural scale using a power function, and establishes a coupling relationship between different tower section parameters to form a parameter evolution model.
[0062] The pylon cross-sections of bridges with different main span lengths exhibit a certain continuous variation in terms of stress requirements, structural dimensions, and safety margins. This variation can be captured and expressed through model fitting. Establishing a quantitative relationship between the main span length and pylon cross-section parameters using power functions helps to accurately describe the nonlinear trend of parameter growth in actual engineering and supports scale extrapolation. Furthermore, by coupling modeling the cross-section parameters, the inherent correlation between parameters can be effectively controlled, avoiding physical inconsistencies caused by independent fitting of each parameter.
[0063] The cable tower section parameters refer to several key dimensional parameters describing the cable tower's geometry and structural boundaries, including section height, section width, tower foot stiffening thickness, cable anchorage zone dimensions, and transverse rib spacing. The mapping model refers to a mathematical model establishing the functional relationship between the main span length and the cable tower section parameters, used to automatically output various section parameters when the main span dimensions are input. The parameter coupling relationship refers to the functional or constraint relationship between different section parameters arising from structural layout, stress coordination, or structural requirements, used to ensure the physical rationality of parameter combinations.
[0064] In the specific implementation process, firstly, bridge instances with the same structural type and structural level as the target bridge are extracted from the database of completed cable-stayed bridge designs, and then sorted in ascending order according to the main span length of each instance. For each bridge instance, its main span length and corresponding tower section parameters are extracted, and a dataset containing mapping pairs of main span length and multiple section parameters is constructed. Each record in the basic dataset includes a set of structural input variables and a set of corresponding tower section output parameters.
[0065] The cable tower cross-sectional parameters include at least the cross-sectional height, cross-sectional width, cross-sectional thickness, local stiffening dimensions of the tower feet, and transverse and longitudinal structural dimensions of the anchorage zone. To improve the stability of parameter modeling and the engineering usability of the data, preferably, during the dataset construction phase, all cross-sectional parameters are standardized in units and scales, and outliers or sample points that do not conform to structural logic are removed to ensure that the data used for fitting modeling has sufficient representativeness and physical consistency.
[0066] After constructing the dataset, a univariate regression fitting was performed on each cable tower section parameter using a power function to establish a single-parameter mapping model. The power function has the following basic form:
[0067] p = a × L^b
[0068] Where p represents a certain pylon section parameter, L represents the main span length of the corresponding bridge, and a and b are constant coefficients obtained from the fitting, representing the scale scaling factor and scale growth exponent, respectively. The fitting process uses the least squares method for parameter estimation, that is, minimizing the sum of squared differences between the predicted values and the actual sample values, thereby obtaining the optimal fitting curve. The power function form has good cross-scale expressive ability and can reflect the nonlinear growth or decay trend of parameters under the condition of increasing span, making it suitable for cable-stayed bridges of different structural levels.
[0069] To further improve the overall consistency of the model and the physical rationality of the derived parameters, after completing the independent fitting of each section parameter, the inherent coupling trends between the section parameters are identified based on fitting residual analysis, correlation testing, and engineering law mining. For example, when a stable linear proportional relationship is found between the tower height and the tower top width, or when the fitting residuals between the transverse dimensions of the anchorage zone and the maximum cable spacing are highly correlated, an engineering coupling relationship can be determined.
[0070] Based on this, coupling equations between parameters are established using multivariate regression modeling or explicit constructor expression. Optional implementation schemes include: selecting significant variable pairs based on correlation coefficient thresholds to establish a multiple regression model, or determining the dominant dimension of change through principal component analysis, thereby constraining the modeling of subordinate parameters.
[0071] Preferably, in one parametric modeling implementation, instead of independently predicting all parameters, a hierarchical modeling approach is adopted, where the primary parameter dominates and the secondary parameters depend on it. That is, firstly, the section parameter with the strongest control over the overall structure or the most sensitive to stress response is selected as the primary parameter, such as the tower height, and its prediction is directly performed using the power function relationship between it and the main span length. Then, based on established coupling relationships, such as the ratio of the tower top width to the tower height, the remaining secondary parameters are derived from the primary parameter values. This modeling approach effectively prevents structural inconsistencies or instability caused by free parameter combinations, reduces model training complexity, and enhances the explicit expression of structural constraints.
[0072] In another preferred implementation, to further improve the model's prediction accuracy near the target bridge scale, a local weighting mechanism can be introduced into the fitting process, assigning higher fitting weights to samples with main span lengths close to the target bridge. Furthermore, a regularization term can be introduced into the fitting model to set a constraint range on the exponential parameter b of the power function, avoiding unacceptable extreme values in engineering.
[0073] The parameter evolution model constructed in the above manner can quantitatively express the law of change of tower cross-section parameters with structural scale, and maintain the coordination between parameters and structural adaptability, providing a clear input basis for subsequent structural performance verification and design optimization.
[0074] For example, in the design of the target bridge with a main span of 960 meters, the three selected large-scale structural bridge examples correspond to main spans of 880 meters, 980 meters, and 1120 meters, respectively. After extracting the tower height and tower top cross-sectional width, and establishing a power function fitting model, the fitting formula between the tower height and the main span length is obtained as h = 0.22 × L^0.85, and the coupling relationship between the tower top width and the tower height is w = 0.6 × h. Substituting the target bridge's main span length L = 960 meters, the calculated tower height h is approximately 122 meters, and the tower top width w is approximately 73 meters.
[0075] When the main span length of the target bridge is close to the boundary of the adjacent structural level, a level transition factor is introduced to construct a smooth transition mechanism between adjacent structural levels and avoid parameter abrupt changes.
[0076] Different structural levels typically correspond to different fitting sample intervals and model coefficients. If the transition logic between levels is not considered, even a small change in the main span length at the level's critical position can cause a sharp jump in the predicted parameters, thus affecting the design smoothness and structural stability. Therefore, when the target main span approaches the boundary of the structural level division, introducing a transition factor to construct a smooth transition mechanism between structural levels is a necessary technical means to ensure model stability and parameter continuity.
[0077] The level transition factor refers to a mathematical factor used to establish a weighted transition relationship between two adjacent structural levels. Essentially, it is a continuous function value ranging from zero to one, used to control the fusion ratio of prediction results from the upper and lower level models within the boundary interval. This factor is calculated based on the relative position of the target main span length with respect to the structural level boundary position, enabling a weighted combination of the output results from the two fitted models.
[0078] In the specific implementation process, the boundary main span lengths of the structural levels are first set to L1 and L2, where L1 is the maximum main span length of the next level and L2 is the minimum main span length of the previous level, forming a transition interval. When the main span length L0 of the target bridge falls into this transition interval, its normalized position α within the interval is calculated, defined as α=(L0−L1) / (L2−L1), where L0 represents the main span length of the target bridge and α represents the weight coefficient of the transition factor, with a value ranging from zero to one. Subsequently, the corresponding cross-sectional parameter values p1 and p2 are calculated using the fitted parameter prediction models at the two structural levels, respectively. Finally, the final predicted value is obtained through the linear weighted fusion formula p=(1−α)×p1+α×p2, where p represents the fused tower cross-sectional parameters, p1 represents the parameters predicted by the next level model, and p2 represents the parameters predicted by the previous level model.
[0079] This linear fusion mechanism can effectively avoid abrupt changes in cross-sectional parameters at structural level boundaries and ensure that design parameters change continuously with the length of the main span, thereby meeting the basic requirements for parameter continuity and response consistency in structural design.
[0080] In a preferred implementation, to improve the stability of the transition factor, the definition of α can be extended to a smoothing function, such as a tangent function or a sigmoid function, so that the change is more gradual in the middle of the boundary interval and converges more rapidly at both ends of the interval, thereby further improving the smoothness and naturalness of the parameter evolution curve.
[0081] In another preferred implementation, it can be extended to a multi-level fusion mechanism. In scenarios where the sample density is insufficient or the structural level division is uncertain, prediction models of the previous level, the current level and the next level are introduced at the same time. A comprehensive prediction model is constructed by weighted averaging or weighted regression to improve the robustness and fault tolerance of the prediction.
[0082] This smooth transition mechanism enables a natural transition of cross-sectional parameters at the structural scale boundary, avoiding abrupt modeling changes and design instability caused by structural level variations. It improves the continuity, stability, and engineering practicality of the overall parameter evolution model, and is particularly suitable for application scenarios where the main span length of the target bridge is close to the critical range.
[0083] For example, in the design of the target bridge with a main span of 960 meters, assuming the structural level boundary is divided into an upper limit of 900 meters for the medium level and a lower limit of 950 meters for the large level, the main span of the target bridge falls within this boundary range. Defining L1=900, L2=950, and L0=960, then α=(960−950) / (950−900)=1.2. Since it exceeds the upper limit, it automatically enters the fully applicable range of the large level without any transition. However, if the target main span is 940 meters, then α=(940−900) / (950−900)=0.8, and the corresponding parameter prediction result is p=0.2×p1+0.8×p2, where p1 is the predicted value for the medium level and p2 is the predicted value for the large level. This fusion result allows for a natural transition, ensuring smooth modeling and accurate prediction.
[0084] The initial cable tower section parameters are calculated using the parameter evolution model based on the main span length of the target bridge.
[0085] The variation trend of the pylon section parameters at the structural scale can be obtained through modeling with structured sample data. The main span length of the target bridge is input into the model as a scale variable, and a set of reasonable structural parameters that are suitable for it can be derived. Compared with traditional empirical estimation or multiple rounds of trial calculations, this method is more systematic, accurate and repeatable, which helps to improve design efficiency and reduce initial modeling costs.
[0086] The parameter evolution model refers to the mathematical mapping model between structural dimensions and cross-sectional parameters established based on historical bridge data in the aforementioned steps. Preferably, it employs a power function or a combination thereof, which can reflect the nonlinear growth law of the tower's geometric parameters with the main span scale. The coefficients of this model have been determined through regression fitting, and can be combined with parameter coupling relationships to achieve coordinated output of master and slave parameters.
[0087] In the specific implementation process, the main span length L0 of the target bridge is first input to determine its structural level. If this value is within the transition range between two structural levels, the fusion weight α in the upper and lower level models is calculated according to the aforementioned smooth transition mechanism. Subsequently, L0 is substituted into the power function expressions of each term in the parameter evolution model to calculate the initial estimated value of each dominant tower section parameter. The formula is expressed as p = a × L0^b, where p is a certain section parameter of the target tower, a is the fitted proportionality coefficient, b is the power exponent, and L0 is the target main span length.
[0088] For dependent parameters with coupling relationships, the obtained dominant parameter values are substituted into the corresponding functions based on the established parameter dependency expressions to derive all dependent section parameters. For example, if the width w of the tower top section is a function of the total tower height h, w = k × h, then the value of w can be directly derived after calculating the value of h.
[0089] In a preferred implementation, for cases where the main span length falls within the boundaries of the existing sample interval or exceeds the upper limit of the sample, a model extrapolation mechanism can be enabled, while simultaneously introducing a prediction boundary control module. This module sets reasonable upper and lower limits for each parameter value during the model prediction process to ensure that the predicted values do not violate structural limits and construction process constraints.
[0090] In another preferred implementation, to improve the model's accuracy in local regions, a weighted average or local regression approach can be used to combine the predicted values of multiple neighboring main span samples. For example, based on three sets of sample bridges with similar main span lengths, the corresponding parameters are predicted separately, and then weighted and summed to form an initial estimate, thereby improving model stability.
[0091] Using the above method, the initial cable tower section parameters corresponding to the target main span length can be calculated quickly without manual intervention, providing accurate input data for subsequent finite element analysis, structural optimization and detailed construction design.
[0092] For example, in the design of the target bridge with a main span of 960 meters, based on the established power function model h=0.22×L^0.85, and inputting L0=960, the total tower height h≈122 meters is calculated; based on the coupling relationship w=0.6×h, the tower top width w≈73.2 meters is calculated; the anchorage zone section width is set to 1.1 times w, resulting in m≈80.5 meters. This set of parameters is used as initial values for structural modeling, ensuring that the initial design scheme has structural rationality and computational convergence.
[0093] Construct a finite element model of the tower of the target bridge, input the design conditions of the target bridge, perform multi-condition structural response analysis on the initial tower section parameters, and determine whether the initial tower section parameters meet the preset design constraints.
[0094] To verify the structural rationality and mechanical adaptability of the calculated cross-sectional parameters under actual engineering conditions, a multi-condition structural response analysis is required. This analysis necessitates constructing a three-dimensional structural model of the target bridge using the finite element method and applying loads consistent with those used in the design phase. By evaluating whether the stress, deformation, stability, and local structural response of the pylon under various combinations of loads under the initial cross-sectional parameters meet the preset design constraints, it can be determined whether the initial scheme can be used for subsequent detailed design or requires further modification.
[0095] The finite element model refers to a structural calculation model composed of multiple elements, which is created by modeling the cable tower structure of the target bridge individually or as an integral part of the bridge deck system using the finite element method. This model should include the main tower structure, simplified equivalent models of the stay cables, support boundary constraints, and construction node details. The design load cases refer to combinations of various loads determined according to design specifications and usage scenarios, including at least dead load, self-weight, live load, wind load, temperature gradient load, and seismic load, and may include special load cases such as construction stages and cable tension adjustments as needed.
[0096] In the specific implementation process, the first step is to establish a finite element model of the target bridge's cable towers in the modeling software. When modeling, appropriate element types must be selected; beam elements or shell elements can be used for the tower columns. The structural treatment at the nodes should conform to the actual structural logic, and the constraints should consider the equivalent effects of foundation stiffness or the elastic modulus of the foundation. For the simulation of the stay cables, equivalent rod elements can be used with prestress applied, or a stiffness simulation method can be employed.
[0097] After the model is built, the initial tower section parameters are input to form a complete initial structural model. Then, various load combinations are applied according to specifications, such as dead load plus live load, dead load plus wind load, maximum temperature difference load, and seismic load. Under each load combination, the stress response, axial force distribution, displacement, horizontal displacement at the top of the tower, shear stress at the tower foot nodes, and local strain energy density are extracted and compared with preset design constraint values. These design constraints include, but are not limited to, the normal section stress not exceeding the design strength, the maximum horizontal displacement less than the specified limit, the overall stability coefficient greater than the stability limit, no local instability in the tower foot area, and no structural damage to the anchorage nodes.
[0098] In a preferred implementation, a constraint evaluation system based on an index matrix can be established, a standard threshold is set for each structural response, and an evaluation function J=max(σi / σ_allow,δi / δ_allow,μi / μ_allow) is introduced, where σi represents the maximum stress value under the i-th working condition, σ_allow represents the allowable stress limit, δi represents the maximum displacement, and μi represents the instability factor. When J is less than or equal to 1, the scheme is considered to meet the structural safety requirements.
[0099] In another preferred implementation, an automated modeling and calculation platform is employed to automate the input process of initial cross-sectional parameters and load case combinations. Multiple rounds of load case simulation and batch result extraction are achieved through scripts or parametric modeling languages such as Python+OpenSees, APDL, or MidasGSE. If the calculation results indicate that any structural index fails to meet the design constraints, subsequent parameter correction procedures are triggered.
[0100] This step uses engineering finite element methods to quantitatively verify the performance of the initial section parameters, ensuring that the data-driven results are not only empirically reasonable, but also meet the structural safety and usability requirements under engineering load conditions, providing a scientific basis for subsequent design finalization.
[0101] For example, in the design of the aforementioned target bridge with a main span of 960 meters, based on the preliminary predicted combination of parameters—tower height of 122 meters, tower top width of 73.2 meters, and anchorage zone width of 80.5 meters—an integrated finite element model of the tower and superstructure was established. Under conditions of dead load plus live load on the main lane, dead load plus vertical wind, summer-winter temperature difference, and seismic acceleration spectrum, the calculated maximum principal stress was 38.5 MPa, the maximum horizontal displacement was 138 mm, and the overall stability coefficient was 2.15. All indicators were within the allowable range specified in the code; therefore, the initial section parameters were considered to meet the design requirements. If the stress exceeds the specified limit under any condition, a parameter correction stage is initiated for reassessment.
[0102] If the initial tower section parameters do not meet the design constraints, historical design error samples at the structural level are extracted, a residual correction model is constructed, the corrected tower section parameters are calculated based on the correction model, and parameter trimming is performed in combination with material properties, construction technology and structural constraints to obtain the final recommended design parameters of the tower section that meet the design constraints.
[0103] When the initial cable tower section parameters calculated based on the parametric evolution model fail to meet the design constraints in the finite element structural response analysis, the initial parameter scheme needs to be modified in a targeted manner to ensure that the final design results meet the requirements of structural safety, applicability, and engineering feasibility. The principle of this modification process is to analyze the systematic deviations between the model's predicted values and the actual parameters used in historical design data under the same structural type and level, extract the residual patterns, construct a residual correction model, and combine it with construction technology, material limitations, and structural constraints to tailor the parameters, thereby achieving adaptive correction of the design parameters and avoiding blind iteration or non-target-oriented numerical calculations.
[0104] The residual correction model refers to an empirical fitting model used to describe the difference between the predicted parameter values and the final values used in actual engineering projects. This model identifies the causes of systematic deviations and performs compensation calculations by comparing the predicted values with historically validated design values, thereby improving the convergence and feasibility of the parameter prediction results. Parameter pruning refers to a secondary screening, constraint adjustment, or boundary truncation of the corrected parameter results to ensure that the output parameters are feasible within the limits of material strength, structural node forms, and construction methods.
[0105] In the specific implementation process, firstly, multiple bridge examples that have completed design at the same structural level are statistically analyzed, and the difference Δp between their model prediction values and the final actual cross-sectional parameters is extracted. This difference is then correlated with variables such as the bridge's main span length, design condition type, and feedback results from the construction phase. A residual correction model is constructed using least squares fitting, locally weighted regression, or machine learning regression methods. Preferably, multivariate regression or ensemble learning models are used to improve the accuracy of residual estimation. This correction model can be formalized as Δp=f(L,p0,σ_max,δ_max), where Δp is the prediction error value, L is the main span length, p0 is the initial prediction value, σ_max is the stress exceedance, and δ_max is the displacement exceedance.
[0106] The residual model is applied to the correction of the target bridge design parameters. Specifically, the residual value Δp for each initial predicted parameter is calculated and corrected to p1 = p0 + Δp, where p1 is the corrected section parameter value. Subsequently, the corrected parameter set is input into the parameter trimming module. This module sets the parameter value range, upper and lower limits of dimensions, and structural ratio constraints. Preferably, parameter trimming rules can be established by combining design specifications regarding tower column details, geometric limits of anchorage nodes, and material machinability dimensions to ensure that the corrected parameters can be practically adopted.
[0107] In a preferred implementation, the residual correction model is implemented using a gradient boosting tree-based regression framework, which can capture the nonlinear relationship between multiple source variables and has strong generalization ability, making it suitable for parameter error estimation under complex working conditions.
[0108] In another preferred implementation, to improve the engineering interpretability of parameter correction, a sensitivity analysis method can be introduced to calculate the influence weight of each section parameter on the structural response result, and prioritize the adjustment of parameters with a large influence, avoiding ineffective correction of minor parameters with low sensitivity.
[0109] Through the above correction mechanism, parameter sets that do not meet design constraints can be optimized into parameter solutions that are structurally feasible and construction adaptable, effectively improving the adaptability of the design process and the feasibility of engineering implementation.
[0110] For example, continuing with the aforementioned target bridge design with a main span of 960 meters, under the initial parameter input conditions, structural response analysis showed that the horizontal displacement at the top of the tower exceeded the limit by 18%, and the stress exceeded the limit by 9%. Querying the residual data between historical predictions and actual parameters for bridges of the same structural level revealed a general underestimation trend of 3% to 6% in tower height. Calculations using the residual correction model indicated that the current tower height needed to be increased by 6.5 meters, and the tower top width needed to be adjusted to 1.08 times the original width. The trimming module determined that the tower top section width was still within the constructible range. The corrected parameters were re-input into the finite element analysis model, and all indicators met the specifications. This parameter set was ultimately determined as the recommended tower section design scheme.
[0111] In another embodiment, the present invention also provides a design system for the pylon cross-section of a long-span cable-stayed bridge based on big data, comprising:
[0112] The design parameter extraction module is used to obtain the design parameters of the target bridge and determine the structural type of the target bridge based on the design parameters;
[0113] The structural level filtering module is used to filter out multiple cable-stayed bridge instances with the same structural type as the target bridge from the completed bridge design data, and divide the cable-stayed bridge instances into multiple structural levels according to the main span length. The structural levels include micro-small, small, medium, large and super-large, and each structural level contains at least three completed cable-stayed bridges.
[0114] The parameter modeling module is used to construct a mapping model of the cable-stayed bridge example's main span length and corresponding tower section parameters, and to establish a quantitative relationship between the tower section parameters and the structural scale using a power function, and to establish a coupling relationship between different tower section parameters to form a parameter evolution model.
[0115] The transition control module is used to introduce a level transition factor when the main span length of the target bridge is close to the boundary of the adjacent structural level, to build a smooth transition mechanism between adjacent structural levels and avoid parameter abrupt changes.
[0116] The initial parameter calculation module is used to calculate the initial cable tower section parameters based on the main span length of the target bridge using the parameter evolution model.
[0117] The structural response analysis module is used to construct the finite element model of the tower of the target bridge, input the design conditions of the target bridge, perform multi-condition structural response analysis on the initial tower section parameters, and determine whether the initial tower section parameters meet the preset design constraints.
[0118] The parameter correction and trimming module is used to extract historical design error samples at the structural level when the initial cable tower section parameters do not meet the design constraints, construct a residual correction model, calculate the corrected cable tower section parameters based on the correction model, and trim the parameters in combination with material properties, construction technology and structural constraints to obtain the final recommended design parameters of the cable tower section that meet the design constraints.
[0119] It should be noted that the explanation of the aforementioned design method embodiment for the cross-section of the pylon of a long-span cable-stayed bridge based on big data also applies to the apparatus of the embodiments of this application, and will not be repeated here.
[0120] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0121] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0122] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0123] The above description is merely a specific embodiment of this application. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the protection scope of this application. For some module structures not specifically defined in this invention, the content described in the prior art shall prevail. The prior art mentioned in the foregoing background and specific embodiments sections can be considered as part of this invention and used to understand the meaning of some technical features or parameters.
Claims
1. A design method for the section of a cable tower of a long-span cable-stayed bridge based on big data, characterized in that, The method includes the following steps: Obtain the design parameters of the target bridge, and determine the structural type of the target bridge based on the design parameters; Multiple cable-stayed bridge instances with the same structural type as the target bridge are selected from the completed bridge design data. The cable-stayed bridge instances are divided into multiple structural levels according to the main span length. The structural levels include micro-small, small, medium, large and super-large. Each structural level contains at least three completed cable-stayed bridges. Based on the main span length and corresponding tower section parameters of the cable-stayed bridge example, a mapping model of the tower section parameters changing with the main span length is constructed. The mapping model establishes a quantitative relationship between the tower section parameters and the structural scale using a power function, and establishes a coupling relationship between different tower section parameters to form a parameter evolution model. When the main span length of the target bridge is close to the boundary of the adjacent structural level, a level transition factor is introduced to construct a smooth transition mechanism between adjacent structural levels and avoid parameter abrupt changes. Based on the main span length of the target bridge, the initial cable tower section parameters are calculated using the parameter evolution model. Construct a finite element model of the tower of the target bridge, input the design conditions of the target bridge, perform multi-condition structural response analysis on the initial tower section parameters, and determine whether the initial tower section parameters meet the preset design constraints. If the initial tower section parameters do not meet the design constraints, historical design error samples at the structural level are extracted, a residual correction model is constructed, the corrected tower section parameters are calculated based on the correction model, and parameter trimming is performed in combination with material properties, construction technology and structural constraints to obtain the final recommended design parameters of the tower section that meet the design constraints.
2. The method for designing the section of the cable tower of the long-span cable-stayed bridge based on big data according to claim 1, characterized in that, Cluster analysis algorithms are used to adaptively group all bridge main span lengths and automatically determine the span range boundaries for each level. 3.The design method of the large-span cable-stayed bridge tower section based on big data according to claim 1, characterized in that, For cases where there are insufficient samples at the structural level, simulation design results or expert-modeled bridge types are introduced as supplementary samples. 4.The design method of a large-span cable-stayed bridge tower section based on big data according to claim 1, characterized in that, In scenarios where sample density is insufficient or structural level division is uncertain, prediction models at three levels—the previous level, the current level, and the next level—are introduced simultaneously, and a comprehensive prediction model is constructed using weighted averaging or weighted regression. 5.The design method of the large-span cable-stayed bridge tower section based on big data according to claim 1, characterized in that, The residual correction model is implemented using a regression framework based on gradient boosting trees.
6. A design system for the section of a cable tower of a long-span cable-stayed bridge based on big data, characterized in that, The system includes the following modules: The design parameter extraction module is used to obtain the design parameters of the target bridge and determine the structural type of the target bridge based on the design parameters; The structural level filtering module is used to filter out multiple cable-stayed bridge instances with the same structural type as the target bridge from the completed bridge design data, and divide the cable-stayed bridge instances into multiple structural levels according to the main span length. The structural levels include micro-small, small, medium, large and super-large, and each structural level contains at least three completed cable-stayed bridges. The parameter modeling module is used to construct a mapping model of the cable-stayed bridge example's main span length and corresponding tower section parameters, and to establish a quantitative relationship between the tower section parameters and the structural scale using a power function, and to establish a coupling relationship between different tower section parameters to form a parameter evolution model. The transition control module is used to introduce a level transition factor when the main span length of the target bridge is close to the boundary of the adjacent structural level, to build a smooth transition mechanism between adjacent structural levels and avoid parameter abrupt changes. The initial parameter calculation module is used to calculate the initial cable tower section parameters based on the main span length of the target bridge using the parameter evolution model. The structural response analysis module is used to construct the finite element model of the tower of the target bridge, input the design conditions of the target bridge, perform multi-condition structural response analysis on the initial tower section parameters, and determine whether the initial tower section parameters meet the preset design constraints. The parameter correction and trimming module is used to extract historical design error samples at the structural level when the initial cable tower section parameters do not meet the design constraints, construct a residual correction model, calculate the corrected cable tower section parameters based on the correction model, and trim the parameters in combination with material properties, construction technology and structural constraints to obtain the final recommended design parameters of the cable tower section that meet the design constraints.
7. The big data based design system of cable tower section of long-span cable-stayed bridge of claim 6, wherein, Cluster analysis algorithms are used to adaptively group all bridge main span lengths and automatically determine the span range boundaries for each level.
8. The big data based design system of cable tower section of long-span cable-stayed bridge of claim 6, wherein, For cases where there are insufficient samples at the structural level, simulation design results or expert-modeled bridge types are introduced as supplementary samples.
9. The design system for the pylon section of a long-span cable-stayed bridge based on big data as described in claim 6, characterized in that, In scenarios where sample density is insufficient or structural level division is uncertain, prediction models at three levels—the previous level, the current level, and the next level—are introduced simultaneously, and a comprehensive prediction model is constructed using weighted averaging or weighted regression.
10. The design system for the pylon section of a long-span cable-stayed bridge based on big data according to claim 6, characterized in that, The residual correction model is implemented using a regression framework based on gradient boosting trees.