Correction method of bridge finite element model

Abstract

The invention provides a correction method of a bridge finite element model, and relates to the field of bridge engineering. According to the method, bridge design drawings, construction records and actual detection data are collected, an initial finite element model is established, key parameters are initialized, a trained neural network model is utilized to evaluate initial grid quality, a self-adaptive grid division technology and engineering experience are combined to optimize grids, parameter sensitivity analysis is carried out through a machine learning algorithm, and the quality of the grid is evaluated. Key correction objects are screened, an optimal parameter combination is automatically searched by using a deep learning model or an optimization algorithm, and finally, the corrected model is comprehensively verified by adopting an independent verification data set. The method improves the model correction efficiency and precision, enhances the adaptability to a complex bridge, and provides a reliable basis for bridge safety assessment and maintenance.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge engineering, and in particular to a method for correcting a finite element model of a bridge. Background Art

[0002] In bridge engineering, finite element analysis (FEA) is a key method for assessing bridge safety, reliability, and durability. However, traditional FEA model modification methods have numerous shortcomings. For example, when it comes to meshing, engineers rely on manual, empirical methods, making it difficult to balance mesh density with computational efficiency. This often results in meshes that are either too sparse or too dense in critical areas, compromising the accuracy of the results.

[0003] When modifying parameters, traditional methods rely on limited test data and require extensive trial calculations to determine parameters. This process is cumbersome and inefficient, making it difficult to fully consider the impact of each parameter on the model. Model validation primarily relies on single-condition comparisons, lacking adaptability to the multiple conditions of complex bridge structures, making it difficult to ensure the accuracy and reliability of the model under various actual conditions. Furthermore, for complex bridge structures, traditional methods struggle to handle the coupling of multiple physical fields, such as the interaction between the structure and temperature and flow fields. This leads to the neglect of these important factors during model modification, affecting overall performance evaluation. Summary of the Invention

[0004] (1) Technical problems solved In view of the deficiencies of the prior art, the present invention provides a method for correcting a bridge finite element model, which solves the problems raised by the above-mentioned background technology.

[0005] (2) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for correcting a bridge finite element model, comprising the following steps: S1. Collect the design drawings, construction records, and actual inspection data of the bridge. Based on the design drawings, establish an initial finite element model of the bridge in the finite element analysis software and initialize the key parameters in the model. S2. Use the trained neural network model to evaluate the quality of the initial mesh, quickly identifying areas of poor mesh quality due to complex geometry and unreasonable meshing. Based on the AI ​​evaluation results, adaptive meshing technology is used for optimization. Combined with the actual stress characteristics of the bridge and previous engineering experience, the mesh of some known key stress-bearing areas is manually strengthened to increase the mesh density. S3. Perform finite element analysis on the meshed model to obtain stress, strain, and other results. These results are used as training data to further train and optimize the neural network model, enabling it to more accurately evaluate mesh quality and guide meshing. After completing local mesh optimization, rerun the finite element analysis to compare the stress, strain, and displacement results of key locations before and after the correction. S4. Use machine learning algorithms to conduct sensitivity analysis on the parameters in the model to identify the parameters that have the greatest impact on the model's output. By changing the parameter values, observe the changes in the model response, select sensitive parameters as key correction targets, and compare the results with the finite element model's calculation results using actual bridge experimental data. S5. Use deep learning models or optimization algorithms to automatically search for the optimal parameter combination, with the objective function of minimizing the error between the model calculation results and the experimental data. For example, during the parameter correction process, the least squares method can be combined with a neural network model to quickly find the parameter value that minimizes the error. During the parameter correction process, the consistency goals of these model calculation results with the experimental data and the actual project goals are transformed into a multi-objective optimization problem. By weighing the relationship between different goals, a set of parameter values ​​that meet the project requirements and are coordinated with each other is found to ensure that the corrected model can accurately reflect the actual situation and meet the design and use standards of the bridge. The multi-objective optimization algorithm is combined with AI technology to solve the multi-objective optimization problem. Through the intelligent search and optimization capabilities of the AI ​​algorithm, the Pareto optimal solution set is quickly found, providing engineers with multiple feasible parameter combination schemes to assist in decision-making; S6. Use an independent validation data set (different from the experimental data used for parameter modification) to fully validate the modified finite element model. Based on the validation results, adjust the local meshing, optimize parameter values, or modify certain assumptions and boundary conditions of the model until the model passes all validation tests and can be reliably used for subsequent work such as bridge performance evaluation, defect diagnosis, and reinforcement design.

[0006] Preferably, in step S1, the actual detection data includes the geometric dimensions, material properties, and load conditions of the bridge, providing accurate basic data for model establishment. The operating methods in the finite element software include dividing the grid, defining material properties, applying boundary conditions and loads, and ensuring that the model is consistent with the design drawings. The initialization setting parameters include the elastic modulus, Poisson's ratio, density of the material, the boundary conditions, load size and distribution of the structure, and these parameters will serve as the basis for subsequent corrections.

[0007] Preferably, in step S2, the adaptive meshing technology is to set appropriate error tolerance and mesh refinement standards, so that the software automatically refines the mesh in key locations where stress concentration areas and structural mutations require higher precision based on the geometric and physical properties of the model, while maintaining a coarser mesh in non-critical areas.

[0008] Preferably, in step S2, the known key stress-bearing areas such as the connection between the piers and abutments and the mid-span area of ​​the beam body are used to ensure that the calculation results of these areas have higher accuracy.

[0009] Preferably, in step S3, if there is still a large deviation between the corrected result and the actual situation or expected result, or the calculation accuracy of some areas still does not meet the requirements, the grid division strategy is further adjusted using the optimized neural network model, and the above steps are repeated until a satisfactory correction effect is obtained.

[0010] Preferably, in step S4, the influencing parameters of the output results of the model include deformation and stress distribution of the structure, and the experimental data include displacement, strain, and vibration frequency obtained by static load tests, dynamic load tests, non-destructive testing, etc. For parameters with large differences, parameter identification and optimization algorithms are used for adjustment.

[0011] Preferably, in step S5, a proxy model is established for the complex finite element model, and the proxy model is used to quickly evaluate the model performance under different parameter combinations, preliminarily determine the optimal parameter range, and then apply these parameters to the original finite element model for accurate calculation and verification, which can reduce the resource consumption of directly performing a large amount of calculations on the complex finite element model.

[0012] Preferably, the verification content includes the static response, dynamic response and long-term performance of the model. The static response includes displacement and stress under different load conditions, the dynamic response includes natural frequency, mode shape, etc., and the long-term performance includes damage accumulation under fatigue load, to ensure that the model can accurately simulate the actual behavior of the bridge under various working conditions.

[0013] (3) Beneficial effects The present invention provides a method for correcting a bridge finite element model, which has the following beneficial effects: This correction method combines AI with finite element model correction, uses neural networks to evaluate and optimize mesh quality, and uses machine learning algorithms and optimization algorithms to automatically search for the optimal parameter combination, reducing manual trial and error, improving correction efficiency, and making the model more accurately reflect actual bridge performance.

[0014] This method can adapt to bridge structures of different complexities and types. By continuously training and optimizing the neural network model, it can automatically process and analyze complex bridge data, improve the level of intelligence, and provide support for the digital and intelligent development of bridge engineering.

[0015] When modifying parameters, the consistency between the model calculation results and the experimental data as well as the various actual engineering objectives are comprehensively considered. A multi-objective optimization algorithm is used to find parameter values ​​that meet the engineering requirements and are coordinated with each other to ensure the feasibility and reliability of the modified model in practical applications. DETAILED DESCRIPTION

[0016] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] Example 1: An embodiment of the present invention provides a method for correcting a bridge finite element model, comprising the following steps: S1. Collect the bridge's design drawings, construction records, and actual inspection data. Based on the design drawings, create an initial finite element model of the bridge in the finite element analysis software and initialize the key parameters in the model. This step is the foundation of the entire correction method, ensuring that the model is consistent with the design drawings and providing accurate data support for subsequent correction work. S2. The quality of the initial mesh is evaluated through the trained neural network model, and areas with poor mesh quality due to complex geometric shapes and unreasonable mesh division are quickly identified. The use of AI technology can more efficiently and accurately find mesh areas that need to be optimized. Compared with traditional mesh division methods, the manual trial and error process is reduced and work efficiency is improved. According to the AI ​​evaluation results, adaptive mesh division technology is used for optimization. Combined with the actual stress characteristics of the bridge and previous engineering experience, the mesh of some known key stress areas is manually strengthened to increase the mesh density. This method of combining AI with engineering experience not only takes advantage of the advantages of intelligent technology, but also takes into account the specific conditions of the actual project, making the mesh division more reasonable and accurate.

[0018] S3. Perform finite element analysis on the meshed model to obtain stress, strain and other results. Use these results as training data to further train and optimize the neural network model. Through continuous training and optimization, the neural network model can better evaluate the mesh quality, guide subsequent meshing work, form a closed-loop optimization process, and gradually improve the accuracy of the model. After completing the local mesh optimization, re-run the finite element analysis and compare the stress, strain and displacement results of key positions before and after correction. If the corrected results still have a large deviation from the actual situation or expected results, or the calculation accuracy of certain areas still does not meet the requirements, use the optimized neural network model to further adjust the meshing strategy and repeat the above steps until a satisfactory correction effect is obtained.

[0019] S4. Use machine learning algorithms to perform sensitivity analysis on model parameters, identify parameters that have a significant impact on the model's output, and then observe changes in the model's response by changing the parameter values. This method can more scientifically determine which parameters require correction, avoiding blindly adjusting all parameters and improving the relevance and efficiency of corrections. S5. Compare the experimental data of the actual bridge with the calculation results of the finite element model. Use a deep learning model or optimization algorithm, with minimizing the error between the model calculation results and the experimental data as the objective function, and automatically search for the optimal parameter combination. During the parameter correction process, transform the consistency goals of these model calculation results with the experimental data and the actual engineering goals into a multi-objective optimization problem. By weighing the relationship between different goals, find a set of parameter values ​​that meet the engineering requirements and are coordinated with each other. Use a multi-objective optimization algorithm combined with AI technology to solve the multi-objective optimization problem. This multi-objective optimization method comprehensively considers the model accuracy and various actual engineering needs, making the corrected parameters more reasonable and practical.

[0020] S6. Fully validate the revised finite element model using an independent validation dataset. This validation includes the model's static response, dynamic response, and long-term performance, ensuring that the model accurately simulates the actual bridge behavior under various operating conditions. Based on the validation results, if any deficiencies are identified, further updates and improvements will be made to the model until it passes all validation tests and can be reliably used for subsequent work such as bridge performance assessment, defect diagnosis, and reinforcement design.

[0021] Example 2: The difference between this embodiment and the first embodiment is that: In the process of model building, MIDAS finite element analysis software was used; For the initial meshing, hexahedral meshing is adopted, which can better adapt to the geometric shape of the regular beam structure and improve the mesh quality.

[0022] In mesh quality assessment, the neural network model structure used is different. Using a convolutional neural network to process mesh data can better capture the spatial characteristics and topological relationships of the mesh, improving the accuracy of the assessment. In parameter sensitivity analysis, a neural network algorithm is used instead of a random forest algorithm. By building a neural network model to learn the mapping relationship between parameters and model outputs, sensitive parameters are determined, further enriching the parameter analysis method. In the process of parameter optimization and adjustment, the particle swarm optimization algorithm is used instead of the genetic algorithm. By simulating the flight and optimization process of particles in the solution space, the optimal parameter combination is quickly searched, thereby improving the optimization efficiency. During the model validation phase, long-term monitoring data was used as part of the validation dataset. This included actual response data of the bridge under different seasons and traffic flows, enabling a more comprehensive assessment of the performance and reliability of the model during long-term operation.

[0023] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for correcting a bridge finite element model, characterized in that: The following steps are involved: S1. Collect the design drawings, construction records, and actual inspection data of the bridge. Based on the design drawings, establish an initial finite element model of the bridge in the finite element analysis software and initialize the key parameters in the model. S2. Use the trained neural network model to evaluate the quality of the initial mesh, quickly identifying areas of poor mesh quality due to complex geometry and unreasonable meshing. Based on the AI ​​evaluation results, adaptive meshing technology is used for optimization. Combined with the actual stress characteristics of the bridge and previous engineering experience, the mesh of some known key stress-bearing areas is manually strengthened to increase the mesh density. S3. Perform finite element analysis on the meshed model to obtain stress, strain and other results. These results are used as training data to further train and optimize the neural network model. After completing local mesh optimization, rerun the finite element analysis to compare the stress, strain and displacement results of key positions before and after the correction. S4. Use machine learning algorithms to conduct sensitivity analysis on the parameters in the model to identify the parameters that have the greatest impact on the model's output. By changing the parameter values, observe the changes in the model response, select sensitive parameters as key correction targets, and compare the results with the finite element model's calculation results using actual bridge experimental data. S5. Use deep learning models or optimization algorithms to automatically search for the optimal parameter combination, with the objective function of minimizing the error between the model calculation results and the experimental data. During the parameter correction process, transform the consistency between these model calculation results and the experimental data and the actual engineering goals into a multi-objective optimization problem. By weighing the relationship between different goals, find a set of parameter values ​​that meet the engineering requirements and are coordinated with each other. Use the multi-objective optimization algorithm combined with AI technology to solve the multi-objective optimization problem; S6. Use an independent validation data set to fully validate the modified finite element model. Based on the validation results, adjust the local meshing, optimize the parameter values, or modify some assumptions and boundary conditions of the model until the model passes all validation tests.

2. The method for correcting a bridge finite element model according to claim 1, characterized in that: In step S1, the actual detection data includes the geometric dimensions, material properties, and load conditions of the bridge. The operation method in the finite element software includes meshing, defining material properties, applying boundary conditions and loads. The initialization setting parameters include the elastic modulus, Poisson's ratio, and density of the material, the boundary conditions of the structure, and the load size and distribution.

3. The method for correcting a bridge finite element model according to claim 1, wherein: In step S2, the adaptive meshing technology sets appropriate error tolerance and mesh refinement standards, refines the mesh in stress concentration areas and key locations where structural mutations require higher precision, and maintains a coarser mesh in non-critical areas.

4. The method for correcting a bridge finite element model according to claim 1, wherein: In step S2, the known key stress-bearing areas include the connection between the piers and abutments, and the mid-span area of ​​the beam.

5. The method for correcting a bridge finite element model according to claim 1, wherein: In step S3, if there is still a large deviation between the corrected result and the actual situation or expected result, or the calculation accuracy of some areas still does not meet the requirements, the optimized neural network model is used to further adjust the grid division strategy, and the above steps are repeated until a satisfactory correction effect is obtained.

6. The method for correcting a bridge finite element model according to claim 1, wherein: In step S4, the influencing parameters of the output results of the model include deformation and stress distribution of the structure, and the experimental data include displacement, strain, and vibration frequency obtained from static load tests, dynamic load tests, non-destructive testing, etc. For parameters with large differences, parameter identification and optimization algorithms are used to adjust them.

7. The method for correcting a bridge finite element model according to claim 1, characterized in that: In step S5, a proxy model is established for the complex finite element model. The proxy model is used to quickly evaluate the model performance under different parameter combinations, preliminarily determine the optimal parameter range, and then these parameters are applied to the original finite element model for accurate calculation and verification.

8. The method for correcting a bridge finite element model according to claim 1, wherein: The verification content includes the static response, dynamic response and long-term performance of the model.