Optimization method and system for reinforced concrete support design

Abstract

The invention provides a reinforced concrete support design optimization method and system, and the method comprises the steps: carrying out the multi-dimensional analysis of a load expected to be borne by a building, so as to determine a key stress point in a building structure; on the basis of the key stress points, behavior characteristics of the reinforced concrete material under different building structure layers are simulated, and a multi-layer mechanical model is generated; establishing an intelligent decision process by using a genetic algorithm, particle swarm optimization and a reinforcement learning framework on the basis of data obtained by a multi-level mechanical model and monitoring a building structure so as to generate a reinforced concrete support design scheme; and establishing a comprehensive evaluation system by using a Bayesian network and a grey system theory, quantifying the reinforced concrete support design scheme to obtain feedback information, optimizing the reinforced concrete support design scheme by using the feedback information, and generating an optimized reinforced concrete support design scheme. The scientificity and rationality of the design scheme of the reinforced concrete support are improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical fields of building structure engineering and intelligent optimization algorithms, and in particular, to an optimization method and system for the design of reinforced concrete supports. Background Art

[0002] In modern building structure engineering, as the height and complexity of buildings continue to increase, more stringent requirements are imposed on the design of support structures. Especially in projects such as high-rise buildings, bridges, and large public facilities, it is necessary to accurately analyze various loads that the building is expected to bear, including static loads, dynamic loads, and seismic loads, etc., to determine the key stress points in the structure. This is not only to ensure the safety and stability of the building, but also to optimize the use of materials, reduce construction costs, and improve economic efficiency. Accurately identifying the key stress points is crucial for designing a safe and economical reinforced concrete support structure. In addition, in the face of the increasingly complex engineering environment and technical requirements, traditional methods are gradually showing their limitations, and there is an urgent need for a new method that can comprehensively consider the influence of multi-dimensional loads and has high optimization capabilities.

[0003] Traditional reinforced concrete support design mainly relies on empirical formulas and finite element analysis methods. Designers make a preliminary design based on specifications and standards, combined with historical data and experimental results, and optimize the design scheme through repeated trial and error. In recent years, some advanced numerical simulation technologies and intelligent optimization algorithms have begun to be applied in this field. For example, hybrid modeling technologies that combine deep learning and physical rules are used to simulate the behavior characteristics of materials, and methods such as genetic algorithms and particle swarm optimization are used to find the optimal solution. These methods have improved the design accuracy and efficiency to a certain extent, but there is still room for improvement. Nevertheless, the existing technologies have greatly promoted the progress of building structure design and provided strong support for complex projects.

[0004] However, the existing solutions still face many challenges in practical applications. First, it is difficult for traditional methods to comprehensively consider the influence of multi-dimensional loads, resulting in inaccurate identification of key stress points, which in turn affects the reliability of the overall design. Second, the traditional optimization process based on trial and error is time-consuming and prone to falling into local optimal solutions, unable to guarantee the global optimum. More importantly, the existing solutions often lack an effective quantitative evaluation system and feedback mechanism, making it difficult to comprehensively evaluate and continuously optimize the design scheme, especially in the face of complex and changeable actual working conditions, the design flexibility is insufficient. These problems limit the ability of existing methods to provide the best solutions in high-performance and complex environments. Summary of the Invention

[0005] The embodiments of the present application provide an optimization method and system for the design of reinforced concrete supports, so as to solve the problems of low scientificity and low rationality of the existing reinforced concrete support design scheme.

[0006] In a first aspect, an embodiment of the present application provides an optimization method for the design of reinforced concrete supports, including: Conduct multi-dimensional analysis of the loads expected to be borne by the building to determine the key stress points in the building structure; Based on the key stress points, adopt a hybrid modeling technique integrating deep learning and physical rules to simulate the behavioral characteristics of reinforced concrete materials at different levels of the building structure, and generate a multi-level mechanical model; Based on the multi-level mechanical model and the data obtained by monitoring the building structure, establish an intelligent decision-making process using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to generate a design plan for reinforced concrete supports; Establish a comprehensive evaluation system using Bayesian networks and grey system theory, quantitatively analyze the design plan for reinforced concrete supports to obtain feedback information, and use the feedback information to optimize the design plan for reinforced concrete supports to generate an optimized design plan for reinforced concrete supports.

[0007] Optionally, the step of establishing an intelligent decision-making process using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks based on the multi-level mechanical model and the data obtained by monitoring the building structure to generate a design plan for reinforced concrete supports includes: Use the finite element analysis method to simulate the multi-level mechanical model to generate simulation data. Based on the simulation data, verify the behavioral patterns and potential risk points of the key stress points under different loading conditions, and use the random forest algorithm to classify and regression predict the simulation data to identify the key parameters and change trends affecting the performance of the building structure, and obtain the behavioral characteristics of the key stress points; According to the behavioral characteristics, the data obtained by monitoring the building structure, and the simulation data, apply the principal component analysis dimensionality reduction technique to extract key feature vectors, and use the time series modeling method to analyze the change trend of the building structure response under preset working conditions to identify the dynamic stress situation, and generate a structure response prediction model according to the dynamic stress situation and the change of the expected load of the building structure; Based on the structure response prediction model, select the initial support members according to the behavioral characteristics of the key stress points and the change of the expected load, encode the design parameters of the initial support members using the genetic algorithm to generate an initial population, and use the differential evolution algorithm to analyze the initial population to generate a set of optimized design parameters; Define a fitness function according to the changing trend and dynamic stress conditions of the structural response prediction model, and use the particle swarm optimization algorithm to iteratively search for the best combination of design parameters. Based on the best combination of design parameters and the set of optimized design parameters, analyze the set of optimal design solutions corresponding to the fitness function. Based on the set of optimal design solutions and the ant colony optimization algorithm, generate the best combination of optimized design parameters; According to the reinforcement learning framework, use the best combination as the action space of the agent in the building structure, and use the safety, stability, and economy of the building structure as the reward signal to adjust the policy parameters of the agent to generate a reinforced concrete support design solution.

[0008] Optionally, the step of defining a fitness function according to the changing trend and dynamic stress conditions of the structural response prediction model, and using the particle swarm optimization algorithm to iteratively search for the best combination of design parameters, based on the best combination of design parameters and the set of optimized design parameters, analyzing the set of optimal design solutions corresponding to the fitness function, and based on the set of optimal design solutions and the ant colony optimization algorithm, generating the best combination of optimized design parameters, includes: Define a fitness function according to the changing trend and dynamic stress conditions of the structural response prediction model, combined with different factors of the building structure, where the different factors of the building structure include safety, stability, and economy; Based on the fitness function, use the particle swarm optimization algorithm and the differential evolution algorithm to iteratively search for the best combination of design parameters, calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm, and apply an adaptive inertia weight to obtain a set of best design parameters. Screen out the design solutions with fitness function values higher than the preset threshold from the set of best design parameters to generate a set of optimal design solutions; Based on the set of optimal design solutions, combine with the ant colony optimization algorithm to generate the best combination of optimized design parameters.

[0009] Optionally, the step of using the particle swarm optimization algorithm combined with the differential evolution algorithm based on the fitness function to iteratively search for the best combination of design parameters, calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm, and apply an adaptive inertia weight to obtain a set of best design parameters, and screening out the design solutions with fitness function values higher than the preset threshold from the set of best design parameters to generate a set of optimal design solutions, includes: Calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm; Apply an adaptive inertia weight adjustment strategy to dynamically adjust the velocity update rule of the particles to generate an optimized particle swarm, and based on the optimized particle swarm, obtain a set of best design parameters; Select the design solutions with fitness function values higher than the preset threshold from the set of optimal design parameters, and generate a set of optimal design solutions based on the design solutions.

[0010] Optionally, based on the structural response prediction model, select the initial support members according to the behavioral characteristics of the key stress points and the expected load changes, encode the design parameters of the initial support members using a genetic algorithm to generate an initial population, and use a differential evolution algorithm to analyze the initial population to generate a set of optimized design parameters, including: According to the structural response prediction model, select the initial support members suitable for each of the key stress points according to the behavioral characteristic description of the key stress points and the expected load change situation. Encode the design parameters of the initial support members using a genetic algorithm to obtain a set of initial support members, construct an initial population based on the set of initial support members, and apply a differential evolution algorithm to perform the construction and mutation operations of the differential vectors for the individuals in the initial population to obtain an optimized population. Generate a set of optimized design parameters based on the optimized population.

[0011] Optionally, based on the key stress points, adopt a hybrid modeling technique integrating deep learning and physical rules to simulate the behavioral characteristics of reinforced concrete materials at different building structural levels to generate a multi-level mechanical model, including: Based on the key stress points, adopt a hybrid modeling technique integrating deep learning and physical rules to collect the key data of each key stress point to generate a basic data set, and use the basic data set to construct a hybrid modeling framework. The key data includes relevant experimental data, historical monitoring data, and theoretical calculation results. Based on the hybrid modeling framework, use a deep learning model to simulate the behavior of reinforced concrete materials at the detailed level, predict the performance changes of the reinforced concrete materials under different stress states, and obtain a detailed-level material behavior model. According to the detailed-level material behavior model, combine the structural characteristics of the reinforced concrete materials at the member level to simulate the behavior of the reinforced concrete members at the structural detail level, and generate a member-level behavior model by introducing physical rules to guide model training. Extend the behavior in the member-level behavior model to the overall structural level, simulate the response of the entire building structure under different loading conditions, and use physical rules to constrain the output of the deep learning model to generate an overall structural level response model. Establish a multi-level mechanical model based on the detailed-level material behavior model, the member-level behavior model, and the overall structural level response model.

[0012] Optionally, establishing a comprehensive evaluation system using Bayesian networks and grey system theory, quantitatively analyzing the reinforced concrete support design scheme to obtain feedback information, and using the feedback information to optimize the reinforced concrete support design scheme to generate an optimized reinforced concrete support design scheme, including: Defining evaluation indicators according to the key factors of the reinforced concrete support design scheme to construct a comprehensive evaluation index system, where the key factors include safety, stability, and economy; Using Bayesian networks to conduct risk assessment on the comprehensive evaluation index system, dynamically adjusting the probability relationship between different risk factors in the reinforced concrete support design scheme to obtain a risk assessment result; Applying grey system theory to analyze the uncertainty of the reinforced concrete support design scheme, and using the grey relational analysis method to generate an uncertainty analysis report; Collating the risk assessment result and the uncertainty analysis report to form a comprehensive evaluation system, and quantitatively scoring the reinforced concrete support design scheme according to the comprehensive evaluation system to obtain a quantitative scoring result; Analyzing the data obtained from monitoring the building structure, expert review opinions, and feedback information from relevant parties to obtain a summary of feedback information; Based on the summary of feedback information and the quantitative scoring result, using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to adjust the reinforced concrete support design scheme to generate an optimized reinforced concrete support design scheme.

[0013] In a second aspect, an embodiment of the present application provides an optimization system for reinforced concrete support design, including: An analysis module for performing multi-dimensional analysis on various loads that the building is expected to bear to determine the key stress points in the building structure; A simulation module for simulating the behavior characteristics of reinforced concrete materials at different building structure levels based on the key stress points, using a hybrid modeling technology that integrates deep learning and physical rules to generate a multi-level mechanical model; A generation module for generating a reinforced concrete support design scheme based on the multi-level mechanical model in combination with the data obtained from monitoring the building structure, and using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to establish an intelligent decision-making process; A quantification module for establishing a comprehensive evaluation system using Bayesian networks and grey system theory, quantitatively analyzing the reinforced concrete support design scheme to obtain feedback information, and using the feedback information to optimize the reinforced concrete support design scheme to generate an optimized reinforced concrete support design scheme.

[0014] In a third aspect, an embodiment of the present application provides a computing device, including a processor and a memory. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the optimization method for the design of reinforced concrete supports according to any one of the first aspects.

[0015] In a fourth aspect, an embodiment of the present application provides a computer storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the optimization method for the design of reinforced concrete supports according to any one of the first aspects is implemented.

[0016] In the embodiment of the present application, the loads expected to be borne by the building are analyzed multi-dimensionally to determine the key stress points in the building structure; based on the key stress points, a hybrid modeling technology integrating deep learning and physical rules is adopted to simulate the behavioral characteristics of reinforced concrete materials at different building structure levels, and a multi-level mechanical model is generated; based on the multi-level mechanical model and the data obtained by monitoring the building structure, an intelligent decision-making process is established using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to generate a design plan for reinforced concrete supports; a comprehensive evaluation system is established using Bayesian networks and grey system theory to quantitatively analyze the design plan for reinforced concrete supports to obtain feedback information, and the feedback information is used to optimize the design plan for reinforced concrete supports to generate an optimized design plan for reinforced concrete supports.

[0017] The technical solution of the present application has the following beneficial effects: Through multi-dimensional analysis of various loads that a building is expected to bear, this application can accurately identify the key stress points in the structure, providing a scientific basis for subsequent design optimization. Based on the key stress points, a hybrid modeling technique is used to simulate the behavior characteristics of reinforced concrete materials at different structural levels, generating a multi-level mechanical model to ensure that the design takes into account the behavior of materials from the micro to the macro levels, improving the accuracy and reliability of the design. An intelligent decision-making process is established using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks, which not only speeds up the optimization process but also improves the quality of the solutions. This method can quickly find the optimal solution among a large number of candidate solutions, significantly improving the design efficiency and accuracy. The Bayesian network and grey system theory are introduced to establish a comprehensive evaluation system to comprehensively quantify and evaluate the design scheme of the reinforced concrete support, and feedback information is obtained. These feedback information are used to further optimize the design scheme to ensure the feasibility and superior performance of the final scheme in practical applications. Through the close combination of the above steps, this method can generate and optimize the design scheme of the reinforced concrete support to ensure its economy while meeting the requirements of safety and stability. This not only improves the overall performance of the building but also reduces the long-term maintenance cost and enhances the reliability and durability of the structure. This method is applicable to the design requirements of various complex building structures, can flexibly handle different working conditions and environmental conditions, and ensures that the design scheme not only meets the specification requirements but also can effectively cope with the uncertain factors in actual engineering.

[0018] Furthermore, the embodiments of this application also use finite element analysis to simulate the multi-level mechanical model, generating simulation data to verify the behavior patterns and potential risk points of the key stress points under different loading conditions, and identifying the key parameters and change trends that affect the structural performance of the building through the random forest algorithm. Further, the principal component analysis dimensionality reduction technique is applied to extract the key feature vectors, and the time series modeling method is combined to analyze the change trends of the building structure response, generating a structural response prediction model. Based on this model, the initial support members are selected and the design parameters are encoded using genetic algorithms, and the differential evolution algorithm is used to generate an optimized set of design parameters. Subsequently, the fitness function is defined and the particle swarm optimization algorithm is used to iteratively search for the best combination of design parameters, and the ant colony optimization algorithm is combined to generate the best combination of the optimized design parameters. Finally, a reinforcement learning framework is introduced, taking the best combination as the action space of the intelligent agent, and using the safety, stability, and economy of the building as the reward signal to adjust the strategy parameters, and finally generating the design scheme of the reinforced concrete support.

[0019] Through the above method, the scientificity and rationality of the reinforced concrete support design scheme are improved. First, through finite element analysis and the random forest algorithm, the behavior patterns of key stress points and potential risk points are accurately identified, ensuring the reliability of the design. Second, the application of the principal component analysis dimensionality reduction technique and the time series modeling method enables the accurate capture of the changing trends of the building structure response, thereby improving the prediction accuracy of the dynamic stress conditions. In addition, the combined use of genetic algorithms, differential evolution algorithms, particle swarm optimization algorithms, and ant colony optimization algorithms not only speeds up the optimization process but also improves the quality of the scheme, ensuring the finding of the global optimal solution. Finally, by introducing a reinforcement learning framework, the best combination is used as the action space of the intelligent agent, and safety, stability, and economy are used as reward signals to adjust the strategy parameters, achieving adaptive optimization. This method not only improves the design efficiency and quality but also ensures the overall performance and long-term reliability of the building, reduces the maintenance cost, and enhances the durability and flexibility of the structure.

[0020] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or in the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0022] Figure 1 It is a flowchart of an optimization method for the design of a reinforced concrete support provided by an embodiment of the present application; Figure 2 It is a schematic structural diagram of an optimization system for the design of a reinforced concrete support provided by an embodiment of the present application; Figure 3 It is a schematic structural diagram of a computing device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application.

[0024] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0025] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0026] Figure 1 A flowchart of a method for optimizing reinforced concrete support design is provided for an embodiment of the present application. Figure 1 As shown, the method includes: Step 101: Analyze the load that the building is expected to bear in multiple dimensions to determine the key stress points in the building structure; In this step, multi-dimensional analysis refers to the comprehensive consideration of various load conditions that a building may face during the design phase, including static loads (such as deadweight), dynamic loads (such as wind and earthquake forces), and long-term forces (such as temperature changes). It does not only focus on a single type of load, but combines multiple factors under different working conditions to identify which parts are the most vulnerable or critical stress points in the structure through numerical simulation and theoretical calculation. These critical stress points refer to the locations that bear the maximum stress or deformation under specific conditions, and they are directly related to the safety and stability of the entire building.

[0027] In actual operation, first, we collect and integrate information from multiple sources such as historical data, geological survey reports, and meteorological data to build a comprehensive load database. Then, we use finite element analysis software to model the building structure, input various possible load combinations, and run the simulation program to generate detailed stress distribution diagrams. Finally, based on the simulation results, we identify key locations that are subject to greater pressure under extreme conditions, providing a basis for subsequent design optimization.

[0028] For example, in a high-rise residential project, engineers first collected meteorological data for the past fifty years in the area, including information such as wind speed and precipitation, and conducted a geological exploration to understand the foundation conditions. Then, they used finite element analysis software to create a three-dimensional model of the entire building structure and input all the above parameters for simulation calculations. The results showed that the frame columns on the bottom few floors of the building endured relatively large shear forces under strong winds, while the top floors mainly faced vertical pressures. Based on this, the designers decided to use stronger materials and technical measures at these critical positions to ensure structural safety.

[0029] Step 102: Based on the key stress points, adopt a hybrid modeling technique integrating deep learning and physical rules to simulate the behavioral characteristics of reinforced concrete materials at different building structural levels and generate a multi-level mechanical model; In this step, the hybrid modeling technique integrating deep learning and physical rules is a method that combines the powerful pattern recognition ability of machine learning and the accurate description ability of traditional engineering mechanics. Deep learning can process a large amount of complex data sets and automatically extract features from them; physical rules are based on known scientific principles and provide accurate predictions of system behavior. The combination of the two can simulate the behavioral characteristics of reinforced concrete materials at three different scales: micro, meso, and macro, from cement hydration reactions to the overall structural response, and generate a flexible and reliable multi-level mechanical model.

[0030] In actual operation, first, prepare experimental data and theoretical formulas as training samples to train a deep neural network to capture the implicit relationship between the internal structure and external performance of the material. At the same time, establish a physical model according to classical mechanics theory and define the behavioral laws at each level. Next, combine the two and introduce appropriate constraint conditions at each level to ensure that the model can capture complex behaviors and conform to actual physical laws. Finally, after repeated verification and adjustment, ensure that the generated multi-level mechanical model can accurately reproduce the true characteristics of reinforced concrete materials.

[0031] For example, continuing with the example of the aforementioned high-rise residential project, based on the key stress points determined in the first step, the designers further carried out a detailed study of the material properties. They collected samples of steel bars and concrete used at the construction site, conducted comprehensive mechanical tests, and obtained data at all levels from micro to macro. Then, the team developed a hybrid model containing a deep learning module and a physical rules module, where the former is responsible for learning the implicit relationship between the internal structure and external performance of the material, and the latter ensures that the prediction results conform to the basic physical laws. After multiple iterations and optimizations, the model successfully simulated the non-linear response of reinforced concrete under different loading conditions, providing a reliable basis for subsequent designs.

[0032] Step 103: Based on the multi-level mechanical model and the data obtained from monitoring the building structure, use genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to establish an intelligent decision-making process to generate a reinforced concrete support design plan; In this step, the intelligent decision-making process refers to constructing a system that can automatically explore the optimal solution space and make the best choice by integrating multiple intelligent optimization algorithms, such as genetic algorithms, particle swarm optimization, and reinforcement learning frameworks. Genetic algorithms are used for global search, particle swarm optimization is used for local refinement, and the reinforcement learning framework is responsible for adjusting the policy parameters according to real-time feedback. This combination can not only quickly find the best design plan that meets multiple objectives but also adapt to the changing actual working conditions to achieve adaptive optimization.

[0033] In actual operation, first, use the multi-level mechanical model generated in Step 102 as the basis, and combine with actual monitoring data (such as strain gauge readings, displacement sensor outputs, etc.) to define the initial design variables and constraint conditions. Then, use genetic algorithms to generate a group of random initial populations and gradually evolve better solutions through crossover and mutation operations. Subsequently, use the particle swarm optimization algorithm to further refine these solutions to find local optimal solutions. On this basis, introduce the reinforcement learning framework, regard the current design plan as the action space of the "agent", and use the safety, stability, and economy of the building as the reward signal to dynamically adjust the policy parameters until convergence to the global optimal solution. Finally, generate a set of comprehensively optimized reinforced concrete support design plans.

[0034] For example, continuing with the previous high-rise residential project, the designers initiated the intelligent decision-making process using the previously established multi-level mechanical model and the real-time monitoring data provided by the sensor network installed on-site. They first generated a group of preliminary designs through genetic algorithms, covering different types of support members and their layout methods. As the number of iterations increased, the particle swarm optimization algorithm gradually narrowed the search range and found several potential optimal solutions. At the same time, the reinforcement learning framework continuously adjusted the policy parameters according to the actual monitoring data to ensure that each solution could achieve a balance among safety, stability, and economy. Finally, the team selected a design plan that could meet the strength requirements and control costs and successfully applied it to the actual construction.

[0035] Step 104: Use Bayesian networks and grey system theory to establish a comprehensive evaluation system, quantitatively analyze the reinforced concrete support design plan to obtain feedback information, and use the feedback information to optimize the reinforced concrete support design plan to generate an optimized reinforced concrete support design plan.

[0036] In this step, the Bayesian network is a probabilistic graphical model used to represent causal relationships between variables and update the probability distributions of these relationships based on new observed data. The grey system theory is applicable to dealing with uncertainty and fuzzy information, helping to quantify factors that are difficult to measure directly. The combination of the two can effectively evaluate the technical feasibility, reliability, and economy of design solutions, provide objective and quantitative feedback information, and support further optimization.

[0037] In actual operation, first, a series of evaluation indicators are defined based on existing technologies and experiences, covering aspects such as safety, stability, and economy. Then, the Bayesian network is used to establish the logical relationships between the indicators, forming a dynamically updated probability model. For factors with uncertainty or those difficult to measure directly, the grey system theory is applied for quantification. Next, by comparing the performances of different design solutions in this system, their advantages and disadvantages are identified. Finally, based on the obtained feedback information, the design parameters are adjusted accordingly until the optimal state is reached. The entire process emphasizes continuous improvement and closed-loop management to ensure that the design solution is always in the best state.

[0038] For example, in a high-rise residential project, designers established a comprehensive evaluation system based on the Bayesian network and grey system theory using multiple candidate design solutions generated in Step 103. They set several evaluation indicators, including compressive strength, seismic performance, material cost, etc., and obtained relevant data through expert reviews and on-site tests. The Bayesian network helped the team clearly show the mutual influences between the indicators, while the grey system theory effectively solved the problem of incomplete data for some indicators. Through quantitative analysis, the team found that although one design solution had a higher initial cost, it showed better stability and durability in long-term operation. Therefore, they decided to adopt this solution and further optimize the details according to the feedback information, such as adjusting the size and layout of the support members, and finally achieved the established goals.

[0039] Through the close combination of the above four steps, this method not only improves the scientificity and rationality of the reinforced concrete support design solution but also significantly enhances the intelligent level of the design process. The multi-dimensional analysis ensures the accurate identification of key stress points, the hybrid modeling technology enables the comprehensive simulation of material behavior characteristics, the intelligent decision-making process speeds up the optimization speed and improves the solution quality, and the comprehensive evaluation system provides strong support for continuous improvement. Ultimately, this method not only guarantees the safety and stability of the building but also takes into account economic benefits, reduces maintenance costs, enhances the durability and flexibility of the structure, and provides a new solution for modern construction projects.

[0040] To address the limitations of traditional design methods when faced with complex structures and variable load conditions, in some embodiments, based on the multi-level mechanical model and the data obtained from monitoring the building structure in step 103, a genetic algorithm, particle swarm optimization, and reinforcement learning framework are used to establish an intelligent decision-making process to generate a reinforced concrete support design solution, including: Using the finite element analysis method, simulate the multi-level mechanical model to generate simulation data. Based on the simulation data, verify the behavior patterns and potential risk points of the key stress points under different loading conditions, and use the random forest algorithm to classify and perform regression prediction on the simulation data to identify the key parameters and change trends affecting the performance of the building structure, and obtain the behavior characteristics of the key stress points; According to the behavior characteristics, the data obtained from monitoring the building structure, and the simulation data, apply the principal component analysis dimensionality reduction technique to extract the key feature vectors, and use the time series modeling method to analyze the change trend of the building structure response under preset working conditions to identify the dynamic stress conditions, and according to the dynamic stress conditions and the expected load change of the building structure, generate a structure response prediction model; Based on the structure response prediction model, select the initial support members according to the behavior characteristics of the key stress points and the expected load change, and use the genetic algorithm to encode the design parameters of the initial support members to generate an initial population. Use the differential evolution algorithm to analyze the initial population to generate an optimized design parameter set; According to the change trend and dynamic stress conditions of the structure response prediction model, define the fitness function, and use the particle swarm optimization algorithm to iteratively search for the best design parameter combination. Based on the best design parameter combination and the optimized design parameter set, analyze the optimal design solution set corresponding to the fitness function. Based on the optimal design solution set and the ant colony optimization algorithm, generate the best combination of optimized design parameters; According to the reinforcement learning framework, use the best combination as the action space of the intelligent agent in the building structure, and use the safety, stability, and economy of the building structure as the reward signal to adjust the strategy parameters of the intelligent agent to generate a reinforced concrete support design solution.

[0041] In this embodiment, finite element analysis is a numerical simulation method that discretizes a continuum into a finite number of units or elements to facilitate the solution of complex physical problems by a computer. The simulation data includes the behavior patterns of key stress points and information on potential risk points under different loading conditions, which are used to verify the theoretical model and provide the behavior characteristics under actual working conditions. The random forest algorithm is an ensemble learning method that classifies and regresses by voting on a large number of decision trees, and can effectively process high-dimensional data sets to identify the key parameters and changing trends affecting the building structure performance. Principal component analysis is a statistical method used to reduce the data dimension while retaining as much of the original data information as possible, thereby extracting the key feature vectors. Time series modeling helps to understand the changing pattern of the structural response over time, which is crucial for capturing dynamic stress conditions. The fitness function defines the criteria for evaluating the quality of individuals and is used here to evaluate the quality of different support member design parameters. The ant colony optimization algorithm simulates the foraging behavior of ants and finds the optimal path or solution through a pheromone update mechanism.

[0042] In the embodiment of the present application, first, the multi-level mechanical model is simulated through finite element analysis to generate detailed simulation data, and then the random forest algorithm is used to classify and regress these data to determine which factors most affect the building structure performance. Next, the principal component analysis dimensionality reduction technique is applied to extract the most important feature vectors, and the time series modeling method is combined to predict the changing trend of the structural response under preset working conditions. According to these prediction results, the initial support members are selected and the design parameters are encoded through the genetic algorithm to form an initial population. Subsequently, the differential evolution algorithm is used to analyze the initial population to obtain a set of optimized design parameters. Then, the particle swarm optimization algorithm is used to iteratively search for the best combination of design parameters and select the corresponding set of optimal design schemes from the set of optimized design parameters. Finally, the ant colony optimization algorithm is introduced to further refine the best combination of design parameters, and the strategy parameters of the intelligent agent are adjusted through the reinforcement learning framework to ensure that the final scheme is optimal in terms of safety, stability, and economy.

[0043] The following is a specific embodiment: In a design project of a large commercial complex, engineers faced a highly complex building structure with multiple functional areas. First, they used finite element analysis software to construct a 3D model of the entire building and conducted detailed simulation calculations to obtain the performance of key stress points under various possible working conditions. Then, the team used the random forest algorithm to analyze the simulation data and found that the layout of support members at certain specific locations significantly affected the overall structural performance. To better understand and predict the long-term performance, they applied principal component analysis and time series modeling methods, successfully simplifying the data analysis process and accurately predicting the changing trends of the structural response in the next few years. Based on this, the engineers selected several candidate support members and found the best design solution that met both strength requirements and cost control through a series of intelligent optimization algorithms. Finally, by continuously adjusting the strategy parameters with the reinforcement learning framework, the safety, stability, and economic benefits of the design solution were ensured, laying a solid foundation for the success of the project.

[0044] To solve the problem that traditional optimization methods are difficult to balance multi-objective requirements under complex working conditions, in some embodiments, according to the changing trend and dynamic stress conditions of the structural response prediction model in step 103, a fitness function is defined, and the particle swarm optimization algorithm is used to iteratively search for the best combination of design parameters. Based on the best combination of design parameters and the set of optimized design parameters, the optimal set of design solutions of the fitness function is analyzed. Based on the optimal set of design solutions and the ant colony optimization algorithm, the best combination of optimized design parameters is generated, and it further includes: According to the change trend and dynamic stress condition of the structural response prediction model, and in combination with different factors of the building structure, a fitness function is defined, where the different factors of the building structure include safety, stability, and economy. Based on the fitness function, a particle swarm optimization algorithm and a differential evolution algorithm are used to iteratively search for the optimal combination of design parameters, calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm, and apply an adaptive inertia weight to obtain the optimal set of design parameters. Design schemes with fitness function values higher than a preset threshold are selected from the optimal set of design parameters to generate an optimal set of design schemes; based on the optimal set of design schemes, in combination with an ant colony optimization algorithm, the optimal combination of optimized design parameters is generated. Optionally, the step of using a particle swarm optimization algorithm in combination with a differential evolution algorithm based on the fitness function to iteratively search for the optimal combination of design parameters, calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm, and apply an adaptive inertia weight to obtain the optimal set of design parameters, and selecting design schemes with fitness function values higher than a preset threshold from the optimal set of design parameters to generate an optimal set of design schemes includes: calculating the fitness function value corresponding to each particle in the particle swarm optimization algorithm; applying an adaptive inertia weight adjustment strategy to dynamically adjust the velocity update rule of the particles to generate an optimized particle swarm, and based on the optimized particle swarm, obtaining the optimal set of design parameters; selecting design schemes with fitness function values higher than the preset threshold from the optimal set of design parameters, and based on the design schemes, generating the optimal set of design schemes.

[0045] In this embodiment, the fitness function is a quantitative index for evaluating the quality of an individual. In this embodiment, it comprehensively considers multiple factors such as the safety, stability, and economy of the building structure. Safety includes compressive strength, seismic performance, etc.; stability involves structural deformation control, stiffness maintenance, etc.; economy covers factors such as material cost and construction difficulty. These different factors are assigned different weights to form a comprehensive scoring system to guide the optimization process. By combining the change trend of the structural response prediction model and the dynamic stress condition, the fitness function can more accurately reflect the performance of the design scheme in actual applications. In addition, the particle swarm optimization algorithm simulates the foraging behavior of birds and finds the optimal solution through group cooperation, while the differential evolution algorithm introduces a mutation operation to increase the diversity of the population. The adaptive inertia weight adjustment strategy dynamically changes the velocity update rule of the particles to help the algorithm achieve a balance between exploration and exploitation.

[0046] In the embodiments of the present application, first, according to the change trend of the structural response prediction model and the dynamic stress conditions, combined with different factors of the building structure, a fitness function is defined. Then, the particle swarm optimization algorithm is combined with the differential evolution algorithm to iteratively search for the optimal combination of design parameters. Specifically, the fitness function value corresponding to each particle is calculated, and an adaptive inertia weight adjustment strategy is applied to dynamically adjust the velocity update rule of the particle to generate an optimized particle swarm. Based on the optimized particle swarm, the optimal set of design parameters is obtained. From this set, the design schemes with fitness function values higher than the preset threshold are selected to generate the optimal set of design schemes. Finally, based on the optimal set of design schemes, the ant colony optimization algorithm is combined to further optimize the optimal combination of design parameters. The ant colony optimization algorithm enhances the global search ability by simulating the mechanism of ant foraging path selection, ensuring that the final scheme not only meets the technical requirements but also maximizes the economic benefits.

[0047] The following is a specific embodiment: In a large bridge construction project, engineers need to ensure that the bridge can maintain safety and stability under extreme weather conditions while controlling the construction cost. They first established a structural response prediction model and analyzed the behavior patterns of the bridge under different loads. Then, a fitness function including safety, stability, and economy was defined, with the highest proportion for safety, followed by stability, and finally economy. Next, the team used the particle swarm optimization algorithm and the differential evolution algorithm for joint optimization. In each iteration, the fitness function value corresponding to each particle was calculated, and an adaptive inertia weight adjustment strategy was applied to dynamically adjust the velocity update rule of the particle to generate an optimized particle swarm. After multiple iterations, a set of optimal design parameters was obtained, and the design schemes with fitness function values higher than the preset threshold were selected from it to form the optimal set of design schemes. To further improve the quality of the scheme, the team also introduced the ant colony optimization algorithm to fine-tune the optimal design scheme, and finally generated a set of bridge support design schemes that are both safe and economical. This set of schemes not only meets the requirements of all technical specifications but also achieves the best performance within the budget, providing strong support for the successful implementation of the project.

[0048] To solve the problem that traditional design methods are difficult to effectively handle complex structures and variable load conditions, in some embodiments, based on the structural response prediction model in step 103, the initial support is selected according to the behavior characteristics of the key stress points and the expected load changes, and the genetic algorithm is used to encode the design parameters of the initial support to generate the initial population. The differential evolution algorithm is used to analyze the initial population to generate the optimized set of design parameters, and further includes: According to the structural response prediction model, based on the behavioral characteristic description of the key stress points and the expected load change situation, select the initial support members applicable to each of the key stress points; use the genetic algorithm to encode the design parameters of the initial support members to obtain a set of initial support members, and based on the set of initial support members, construct an initial population, and apply the differential evolution algorithm to construct and mutate the differential vectors of the individuals in the initial population to obtain an optimized population; based on the optimized population, generate an optimized set of design parameters.

[0049] In this embodiment, the structural response prediction model is a mathematical model that is based on a multi-level mechanical model and actual monitoring data and is used to predict the dynamic behavior of a building under different working conditions. The description of the behavioral characteristics of the key stress points includes information such as the stress distribution, deformation mode, and potential risk points that these positions may exhibit under various loading conditions. The expected load change covers various types of loads (such as static loads, dynamic loads, seismic loads, etc.) that the building may encounter during its life cycle and their change trends over time. The initial support member refers to the type and size of the support member that is most suitable for each key stress point according to the above prediction model and behavioral characteristic description in combination with the expected load change situation. The genetic algorithm is a global search algorithm that simulates the process of natural selection, forms "chromosomes" by encoding design parameters, and generates new solutions through operations such as crossover and mutation. The differential evolution algorithm is another evolutionary algorithm that enhances population diversity by constructing differential vectors and introducing mutation operations, thereby improving the search efficiency and the ability to find better solutions.

[0050] In the embodiment of the present application, first, according to the structural response prediction model, engineers select the initial support members applicable to each key stress point based on the behavioral characteristic description of the key stress points and the expected load change situation. Then, use the genetic algorithm to encode the design parameters of these initial support members to obtain a set of initial support members. Based on this set, construct an initial population, that is, a set of candidate solutions. Next, apply the differential evolution algorithm to construct and mutate the differential vectors of the individuals in the initial population to generate an optimized population. In this process, the differential evolution algorithm increases population diversity by introducing a mutation mechanism and avoids premature convergence to a local optimal solution. Finally, based on the optimized population, generate an optimized set of design parameters, and these parameters can better adapt to the actual working condition requirements, ensuring that the design scheme reaches the optimum in terms of safety, stability, and economy.

[0051] The following is a specific embodiment: In a high-rise office building construction project, engineers need to ensure the safety and stability of the building under various loads. First, they used a structural response prediction model to analyze the dynamic behavior of the building under different load conditions, especially the performance of key stress points under extreme conditions. Based on these analysis results, the team selected initial support members suitable for each key stress point. For example, high-strength steel was used in some parts, while prestressed concrete was used in other places. Then, they encoded the design parameters of these initial support members using a genetic algorithm to form an initial set of support members. Based on this set, an initial population containing multiple candidate solutions was constructed. Subsequently, the team applied a differential evolution algorithm to construct and mutate the differential vectors of each individual in the initial population to generate an optimized population. In this way, not only was the population diversity improved, but also a series of better combinations of design parameters were found. Finally, based on the optimized population, a set of optimized design parameter sets was generated. These parameters not only met the requirements of all technical specifications but also achieved the best performance within the budget, providing strong support for the successful implementation of the project. In addition, the optimized support design solution also considered long-term maintenance costs and construction convenience, further enhancing the economic benefits of the overall project.

[0052] To solve the problem that traditional modeling methods are difficult to comprehensively capture the behavior characteristics of reinforced concrete materials at different levels, in some embodiments, in step 102, based on the key stress points, a hybrid modeling technique integrating deep learning and physical rules is used to simulate the behavior characteristics of reinforced concrete materials at different building structural levels and generate a multi-level mechanical model, including: Based on the key stress points, a hybrid modeling technique integrating deep learning and physical rules is used to collect key data for each key stress point to generate a basic data set. Using the basic data set, a hybrid modeling framework is constructed. The key data includes relevant experimental data, historical monitoring data, and theoretical calculation results. Based on the hybrid modeling framework, a deep learning model is used to simulate the behavior of reinforced concrete materials at the detailed level and predict the performance changes of the reinforced concrete materials under different stress states to obtain a detailed-level material behavior model. According to the detailed-level material behavior model, combined with the structural characteristics of the reinforced concrete members at the member level, the behavior of the reinforced concrete members at the structural detail level is simulated. By introducing physical rules to guide model training, a member-level behavior model is generated. The behavior in the member-level behavior model is extended to the overall structural level to simulate the response of the entire building structure under different loading conditions. Using physical rules to constrain the output of the deep learning model, an overall structural-level response model is generated. Based on the detailed-level material behavior model, the member-level behavior model, and the overall structural-level response model, a multi-level mechanical model is established.

[0053] In this embodiment, the key data includes relevant experimental data, historical monitoring data, and theoretical calculation results. These data are used to describe and predict the performance changes of reinforced concrete materials under various stress states. Experimental data usually comes from laboratory tests, such as compressive strength, elastic modulus, etc.; historical monitoring data is sourced from the long-term monitoring systems of existing buildings, providing the material performance under actual working conditions; theoretical calculation results are mathematical formulas and analytical solutions derived based on classical mechanics principles. The basic dataset refers to integrating the above three types of data to form a comprehensive database as the basis for subsequent modeling. The hybrid modeling framework combines the powerful pattern recognition ability of deep learning and the precise description ability of physical rules, which can not only handle complex non-linear problems but also ensure that the model output conforms to physical laws. The material behavior model at the detailed level focuses on micro-level simulations, such as the cement hydration reaction process, the interface effect between aggregates and the matrix, etc., to predict the performance changes of materials under different stress states. The component-level behavior model considers the overall structural characteristics of reinforced concrete components, such as size, shape, reinforcement configuration, etc., and guides model training by introducing physical rules to ensure the accuracy of simulation results. The overall structural-level response model extends the behavior at the component level to the entire building structure, evaluates its response under different loading conditions, and uses physical rules to constrain the output of the deep learning model to ensure the reliability of the model.

[0054] In the embodiment of this application, first, based on the key stress points, a hybrid modeling technique integrating deep learning and physical rules is adopted to collect the key data of each key stress point and generate a basic dataset. These data cover information at all levels from micro to macro, forming a comprehensive data resource library. Next, a hybrid modeling framework is constructed using the basic dataset, where the deep learning model is used to capture the implicit relationship between the internal structure and external performance of the material, while physical rules provide an accurate description of the system behavior. Then, based on the hybrid modeling framework, the deep learning model is used to simulate the behavior of reinforced concrete materials at the detailed level and predict their performance changes under different stress states, obtaining the material behavior model at the detailed level. According to this model, combined with the structural characteristics of reinforced concrete materials at the component level, the behavior of reinforced concrete components at the structural detail level is simulated, and the component-level behavior model is generated by introducing physical rules to guide model training. Finally, the behavior in the component-level behavior model is extended to the overall structural level, the response of the entire building structure under different loading conditions is simulated, and the output of the deep learning model is constrained by physical rules to generate the overall structural-level response model. Finally, based on the material behavior model at the detailed level, the component-level behavior model, and the overall structural-level response model, a complete multi-level mechanical model is established, which can comprehensively reflect the true performance of reinforced concrete materials.

[0055] The following is a specific embodiment: In a high-rise residential project, engineers need to accurately predict the performance of reinforced concrete materials at different levels to ensure the safety and durability of the building. They first identified several key stress points, such as the columns at the bottom layer, the connections between beams and slabs, etc., and collected relevant experimental data (such as compressive strength, elastic modulus), historical monitoring data (such as strain gauge readings, displacement sensor outputs), and theoretical calculation results (such as finite element analysis). These data were integrated into a basic dataset for constructing a hybrid modeling framework. Then, the team developed a hybrid model that includes a deep learning module and a physical rule module. The former is responsible for learning the implicit relationship between the internal structure and external performance of the material, and the latter ensures that the prediction results conform to the basic physical laws. Through multiple iterations of optimization, the model successfully simulated the nonlinear response of reinforced concrete under different loading conditions. Especially at the microscopic level, it accurately captured the process of cement hydration reaction and the interface effect between aggregates and matrix. Based on this, they further simulated the behavior of reinforced concrete components at the structural detail level, generated a component-level behavior model, and extended it to the overall structural level to simulate the response of the entire building under extreme conditions such as earthquakes and wind loads. Finally, based on the detailed-level material behavior model, component-level behavior model, and overall structural-level response model, a multi-level mechanical model was established, providing reliable decision-making support for the design team. This method not only improves the design accuracy but also provides valuable reference for future maintenance and renovation.

[0056] To solve the problems that traditional evaluation methods are difficult to comprehensively quantify the risks and uncertainties of the reinforced concrete support design scheme and lack an effective feedback mechanism in the optimization process, in some embodiments, in step 104, the comprehensive evaluation system is established by using Bayesian network and grey system theory to quantitatively analyze the reinforced concrete support design scheme to obtain feedback information, and the feedback information is used to optimize the reinforced concrete support design scheme to generate an optimized reinforced concrete support design scheme, including: According to the key factors of the reinforced concrete support design scheme, evaluation indicators are defined to construct a comprehensive evaluation index system. The key factors include safety, stability, and economy. Using a Bayesian network, risk assessment is carried out on the comprehensive evaluation index system, and the probability relationship between different risk factors in the reinforced concrete support design scheme is dynamically adjusted to obtain a risk assessment result. The grey system theory is applied to analyze the uncertainty of the reinforced concrete support design scheme, and the grey relational analysis method is used to generate an uncertainty analysis report. The risk assessment result and the uncertainty analysis report are sorted out to form a comprehensive evaluation system. According to the comprehensive evaluation system, a quantitative score is given to the reinforced concrete support design scheme to obtain a quantitative scoring result. The data obtained from monitoring the building structure, the expert review opinions, and the feedback information from relevant parties are analyzed to obtain a summary of the feedback information. Based on the summary of the feedback information and the quantitative scoring result, the genetic algorithm, particle swarm optimization, and reinforcement learning framework are used to adjust the reinforced concrete support design scheme to generate an optimized reinforced concrete support design scheme.

[0057] In this embodiment, the comprehensive evaluation index system refers to a series of evaluation indicators defined according to the key factors (such as safety, stability, economy) of the reinforced concrete support design scheme. These indicators are used to measure the performance of the design scheme in different aspects and form a comprehensive evaluation framework. The Bayesian network is a probabilistic graphical model used to represent the causal relationship between variables and dynamically update the probability distribution of these relationships according to new observed data. It can help identify and evaluate the mutual influence between different risk factors and provide an evidence-based risk assessment tool. The grey system theory is applicable to dealing with uncertainty and fuzzy information, especially when the data is incomplete or has large fluctuations, and can quantify the uncertainty through the grey relational analysis method. The risk assessment result includes the change of the probability relationship between various risk factors and reflects the potential risk level of the design scheme. The uncertainty analysis report provides a detailed description of the uncertainty of the scheme to help decision-makers understand the possible deviations and unknown factors. The quantitative scoring result is a comprehensive score obtained by scoring the performance of the design scheme on each evaluation indicator and is used to compare the advantages and disadvantages of different schemes.

[0058] In the embodiments of the present application, first, according to the key factors of the reinforced concrete support design scheme (such as safety, stability, and economy), evaluation indicators are defined, and a comprehensive evaluation index system is constructed. Then, the Bayesian network is used to conduct risk assessment on this comprehensive evaluation index system, dynamically adjust the probability relationship between different risk factors in the design scheme, and obtain the risk assessment result. Next, the grey system theory is applied to analyze the uncertainty of the design scheme, and the grey relational analysis method is used to generate an uncertainty analysis report. The above risk assessment results and uncertainty analysis reports are sorted out to form a comprehensive evaluation system, and the design scheme is quantitatively scored according to this system to obtain the quantitative scoring result. In addition, the team also analyzes the data obtained from monitoring the building structure, the expert review opinions, and the feedback information from relevant parties to obtain a summary of the feedback information. Finally, based on the summary of the feedback information and the quantitative scoring result, the genetic algorithm, particle swarm optimization, and reinforcement learning framework are used to adjust the reinforced concrete support design scheme to generate an optimized design scheme.

[0059] The following is a specific example: In a design project of a large commercial complex, engineers need to ensure that the support structure is both safe and economical. They first defined a series of evaluation indicators according to the key factors (such as compressive strength, seismic performance, material cost, etc.) and constructed a comprehensive evaluation index system. Next, the Bayesian network was used to conduct risk assessment on this system, and it was found that the layout method of the support members at certain positions significantly affected the safety of the overall structure. At the same time, the grey system theory was applied to analyze the uncertainty of the design scheme, especially some fuzzy factors in the long-term performance prediction, and a detailed uncertainty analysis report was generated. After sorting out this information, the team formed a complete comprehensive evaluation system and quantitatively scored multiple candidate schemes. In addition, they also collected on-site monitoring data, expert review opinions, and feedback information from the owner and other relevant parties to obtain a detailed summary of the feedback information. Based on this information, the team used the genetic algorithm, particle swarm optimization, and reinforcement learning framework to adjust the original design scheme. For example, the design parameters are encoded by the genetic algorithm to generate an initial population, the particle swarm optimization algorithm iteratively searches for the best combination of design parameters, and the reinforcement learning framework continuously adjusts the strategy parameters according to the feedback information to ensure that the final scheme is optimal in terms of safety, stability, and economy. This method not only improves the quality of the design scheme but also enhances the feasibility and reliability of the project, laying a solid foundation for the success of the project.

[0060] Considering that the traditional particle swarm optimization algorithm is prone to falling into local optimal solutions when dealing with complex multi-objective optimization problems and its parameter settings have a great impact on the algorithm performance, the present application proposes a new alternative solution, which includes: Based on the fitness function, an improved particle swarm optimization algorithm combined with a differential evolution algorithm is used to iteratively search for the best combination of design parameters, calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm, and apply an adaptive inertia weight to obtain the best set of design parameters. From the best set of design parameters, design schemes with fitness function values higher than a preset threshold are screened out to generate an optimal set of design schemes, including: Using the fitness function, according to the improved particle swarm optimization algorithm, calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm to evaluate the performance of each particle, and apply an adaptive inertia weight adjustment strategy for each particle , calculate the fitness value of the particle according to the fitness function, and introduce an adaptive inertia weight , and automatically adjust according to the current iteration number and the change of the global historical best position. The inertia weight formula is defined as: ; Among them, represents the inertia weight at the -th iteration, is the basic weight, and are the maximum and minimum values of the inertia weight respectively, is the middle iteration number, is the coefficient controlling the weight change rate, is the amplitude, is the angular frequency, is the phase difference; The following is a detailed explanation of each parameter: represents the basic weight, which represents the minimum value of the inertia weight without additional adjustment. Usually set through experiments or experience. A relatively conservative basic value can be selected based on the results of previous tests to ensure sufficient exploration ability in the initial stage of the algorithm.

[0061] and represent the maximum and minimum values of the inertia weight respectively, which limit the change range of the inertia weight. Set according to the experience of specific problems. The maximum value should be large enough to ensure strong global search ability in the initial stage, and the minimum value ensures fine local search in the later stage.

[0062] represents the middle iteration number, which represents the key point of the transition from global search to local search. It can be set according to half of the expected iteration number, or the best conversion point can be determined through previous experiments.

[0063] The coefficient representing the rate of change of the control weight determines the speed at which the inertia weight changes over time. It is set through experiments or experience. A smaller value means a slower rate of change, while a larger value accelerates the change.

[0064] Represents the amplitude, which is the magnitude of the sine wave. It is set according to the intensity of the random perturbation to be added as needed. The appropriate amplitude value can be determined through experiments to maintain a certain exploration ability.

[0065] Represents the angular frequency, which determines the period length of the sine wave. It is set according to the desired frequency of the periodic change to be introduced. Usually, a smaller value is chosen to avoid overly frequent perturbations.

[0066] Represents the phase difference, which determines the initial position of the sine wave. It can be set randomly or adjusted according to specific circumstances to ensure a certain difference between different runs.

[0067] The reasons for each sub - design are introduced as follows: It is to provide a stable base value to ensure that the inertia weight does not become too low, thus maintaining a certain exploration ability.

[0068] By introducing the Sigmoid function, the inertia weight gradually decreases from a larger value to a smaller value as the number of iterations increases. This change helps to conduct extensive global searches in the initial stage and turn to more refined local searches in the later stage.

[0069] By introducing the sine wave, it provides periodic random perturbations to prevent the algorithm from falling into local optima and enhance the global search ability.

[0070] By adding each sub - item, each sub - item represents a different regulation mechanism, and they work together to achieve better optimization effects.

[0071] Based on the result of the inertia weight, the velocity update rule of the particle is dynamically adjusted. According to the particle's own historical best position and the global historical best position gbest of the group, the velocity and position of the particle are updated to generate an optimized particle swarm. Among them, the velocity update formula is: ; Among them, represents the velocity of particle at the th iteration, represents the velocity of particle at the The velocity, and is the acceleration constant, and is a random number within the interval [0, 1], denotes the particle at the current position at the and is the exponential factor, gbest represents the historical best position in the population, denotes the particle at the new position at the Among them, the acceleration constant adjusts and The calculation formula is: ; Among them, and are respectively the minimum and maximum values of is the coefficient controlling the rate of change, and are respectively the minimum and maximum values of is the coefficient controlling the rate of change; among them, the exponential factor and The calculation formula is: ; Among them, and are respectively the minimum and maximum values of and are respectively the minimum and maximum values of and are the coefficients controlling the rate of change of the acceleration constant and the exponential factor, and are the amplitudes, and are the angular frequencies, and is the phase difference.

[0072] The following is a detailed explanation of each parameter: Denotes the particle at the velocity at the

[0073] Denotes the The velocity of the particle at the next iteration. The acquisition method is based on the result of the previous iteration and can be set randomly or to zero during initialization. The velocity of the particle at the next iteration. The acquisition method is based on the result of the previous iteration and can be set randomly or to zero during initialization.

[0074] represents the inertia weight, which is dynamically adjusted in each iteration. It is calculated through the adaptive inertia weight formula.

[0075] and represent the acceleration constants, which are used to adjust the velocity of the particle moving towards its own historical best position and the global historical best position. They are dynamically calculated through the acceleration constant adjustment formula to ensure appropriate search intensity at different stages.

[0076] and represent random numbers within the interval [0, 1], which introduce randomness to avoid the algorithm falling into a local optimal solution. The acquisition method is to randomly generate them in each iteration.

[0077] represents the historical best position of the particle itself. The acquisition method is to update it according to the fitness value of the particle in each iteration.

[0078] represents the current position of the particle at the th iteration. The acquisition method is based on the result of the previous iteration and can be set randomly during initialization.

[0079] and represent the exponential factor, which is used to adjust the acceleration of the particle moving towards and gbest. The acquisition method is to dynamically calculate it through the exponential factor adjustment formula to enhance the control of the search process.

[0080] gbest represents the global historical best position in the population. The acquisition method is to update it according to the fitness values of all particles in the entire population.

[0081] represents the new position of the particle at the th iteration. The acquisition method is to calculate it through the velocity update formula.

[0082] represents the minimum and maximum values of the acceleration constant The acquisition method is to set them according to the experience of specific problems.

[0083] and A coefficient representing the rate of change control. It is obtained by experimental or empirical settings. A smaller value means a slower rate of change, while a larger value speeds up the change.

[0084] and and represent the exponential factor and are the minimum and maximum values. They are obtained by empirical settings according to specific problems.

[0085] and A coefficient representing the rate of change control. It is obtained by experimental or empirical settings. A smaller value means a slower rate of change, while a larger value speeds up the change.

[0086] and represents the amplitude, indicating the magnitude of the sine wave and cosine wave. It is obtained by setting according to the intensity of the random perturbation to be added as needed.

[0087] and represents the angular frequency, which determines the period length of the sine wave and cosine wave. It is obtained by setting according to the frequency of the periodic change to be introduced.

[0088] and represents the phase difference, which determines the initial position of the sine wave and cosine wave. It can be randomly set or adjusted according to specific circumstances.

[0089] The design reasons for each sub-item are introduced as follows: The design reason is to maintain the original movement trend of the particles, prevent the particles from changing direction too quickly, and help maintain the global search ability.

[0090] The design reason is to guide the particles to move towards their own historical best positions and enhance the individual learning ability. The introduction of random numbers and the exponential factor provide flexibility and diversity.

[0091] The design reason is to guide the particles to move towards the group historical best position and enhance the group cooperation ability. The introduction of random numbers and the exponential factor provide flexibility and diversity.

[0092] By adding each sub-item, each sub-item represents a different regulation mechanism, and they work together to achieve a better optimization effect: The inertial part retains the original movement trend of the particle, ensuring that the particle does not suddenly change direction. The individual learning part guides the particle to move towards its own best position, enhancing the individual's learning ability. The group collaboration part guides the particle to move towards the best position of the group, enhancing the group's collaboration ability. By adding, the velocity update of the particle not only considers the original movement trend but also combines the historical best information of the individual and the group, thus improving the exploration and exploitation ability of the algorithm.

[0093] The following is a specific example: Suppose a high-rise residential project is being designed, and the most suitable arrangement of support members needs to be selected to ensure the safety and economic efficiency of the building. Some key stress points have been determined, and a multi-level mechanical model has been constructed to simulate the structural response under different loading conditions.

[0094] Parameter setting: Initial population size , velocity range , position range , basic value of inertia weight , maximum value of inertia weight , minimum value of inertia weight , number of intermediate iterations , coefficient for controlling the rate of change of weight , amplitude Angular frequency , phase difference , minimum value of acceleration constant , maximum value of acceleration constant , coefficient for controlling the rate of change , minimum value of exponential factor , maximum value of exponential S factor , coefficient for controlling the rate of change , amplitude , angular frequency , phase difference Calculation process: Randomly generate 50 particles, and each particle represents a possible arrangement plan of support members. Calculate the fitness value of each particle according to the fitness function . Suppose in the 1st iteration: ; Suppose particle 's current position and velocity are 0.3 and 0.2 respectively, the group historical best position gbest is 0.7, and the particle's own historical best position Is 0.5. At the second iteration: ; Assume

[0095]

[0096] (If it exceeds the position range, limit it within the range) After multiple iterations, design solutions with fitness function values higher than the preset threshold are screened out from the obtained set of optimal design parameters to generate a set of optimal design solutions.

[0097] It can be seen from the calculation results that the improved particle swarm optimization algorithm combined with the differential evolution algorithm not only speeds up the search speed but also increases the possibility of finding the global optimal solution. In particular, the application of the adaptive inertia weight, dynamically adjusted acceleration constants, and the exponential factor enables the algorithm to achieve a good balance between exploration and exploitation. The finally generated optimal design solutions not only meet the requirements of safety and stability but also achieve the goal of cost control. This method provides strong support for the successful implementation of high-rise residential projects and valuable reference for future maintenance and renovation.

[0098] Figure 2 The following is a schematic structural diagram of an optimization system for reinforced concrete support design provided by an embodiment of the present application, as Figure 2 shown. The system includes: An analysis module 21 for performing multi-dimensional analysis on various loads that the building is expected to bear to determine the key stress points in the building structure; A simulation module 22 for simulating the behavior characteristics of reinforced concrete materials at different building structure levels based on the key stress points by using a hybrid modeling technology integrating deep learning and physical rules to generate a multi-level mechanical model; A generation module 23 for generating a reinforced concrete support design solution based on the multi-level mechanical model combined with the data obtained by monitoring the building structure and using a genetic algorithm, particle swarm optimization, and a reinforcement learning framework to establish an intelligent decision-making process; A quantification module 24 for establishing a comprehensive evaluation system by using a Bayesian network and grey system theory, quantitatively analyzing the reinforced concrete support design solution to obtain feedback information, and optimizing the reinforced concrete support design solution by using the feedback information to generate an optimized reinforced concrete support design solution.

[0099] Figure 2 The described optimization system for reinforced concrete support design can execute Figure 1For an optimization method of a reinforced concrete support design described in the illustrated embodiment, its implementation principle and technical effects will not be elaborated further. For an optimization system of a reinforced concrete support design in the above embodiment, the specific manners in which each module and unit perform operations have been described in detail in the embodiment related to this method, and will not be elaborated here.

[0100] In a possible design, Figure 2 The optimization system of a reinforced concrete support design in the illustrated embodiment can be implemented as a computing device, such as Figure 3 shown, this computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32.

[0101] The processing component 32 is configured to: analyze the loads expected to be borne by a building in multiple dimensions to determine the key stress points in the building structure; based on the key stress points, adopt a hybrid modeling technology integrating deep learning and physical rules to simulate the behavior characteristics of reinforced concrete materials at different building structure levels and generate a multi-level mechanical model; based on the multi-level mechanical model and the data obtained by monitoring the building structure, use genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to establish an intelligent decision-making process to generate a reinforced concrete support design scheme; use Bayesian networks and grey system theory to establish a comprehensive evaluation system, quantitatively analyze the reinforced concrete support design scheme to obtain feedback information, and use the feedback information to optimize the reinforced concrete support design scheme to generate an optimized reinforced concrete support design scheme.

[0102] Among them, the processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components for executing the above method.

[0103] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0104] Of course, the computing device necessarily may also include other components, such as input / output interfaces, display components, communication components, etc.

[0105] The input / output interface provides an interface between the processing component and the peripheral interface module, and the above-mentioned peripheral interface module may be an output device, an input device, etc.

[0106] The communication component is configured to facilitate communication between the computing device and other devices in a wired or wireless manner, etc.

[0107] Among them, the computing device may be a physical device or an elastic computing host provided by a cloud computing platform, etc. At this time, the computing device may refer to a cloud server, and the above-mentioned processing component, storage component, etc. may be basic server resources leased or purchased from a cloud computing platform.

[0108] The embodiment of the present application also provides a computer storage medium storing a computer program, and when the computer program is executed by a computer, it can implement the above-mentioned Figure 1 optimization method for the design of a reinforced concrete support shown in the above-mentioned embodiment.

[0109] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0110] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement without creative efforts.

[0111] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An optimization method for the design of reinforced concrete supports, characterized in that, Including: Conduct multi-dimensional analysis on the loads expected to be borne by the building to determine the key stress points in the building structure; Based on the key stress points, adopt a hybrid modeling technology integrating deep learning and physical rules to simulate the behavioral characteristics of reinforced concrete materials at different building structure levels and generate a multi-level mechanical model; Based on the multi-level mechanical model and the data obtained by monitoring the building structure, establish an intelligent decision-making process using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to generate a reinforced concrete support design plan; Establish a comprehensive evaluation system using Bayesian networks and grey system theory, quantitatively analyze the reinforced concrete support design plan to obtain feedback information, and use the feedback information to optimize the reinforced concrete support design plan to generate an optimized reinforced concrete support design plan.

2. The method according to claim 1, wherein The step of establishing an intelligent decision-making process using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks based on the multi-level mechanical model and the data obtained by monitoring the building structure to generate a reinforced concrete support design plan includes: Use the finite element analysis method to simulate the multi-level mechanical model to generate simulation data. Based on the simulation data, verify the behavioral patterns and potential risk points of the key stress points under different loading conditions, and use the random forest algorithm to classify and regression predict the simulation data to identify the key parameters and change trends affecting the building structure performance, and obtain the behavioral characteristics of the key stress points; According to the behavioral characteristics, the data obtained by monitoring the building structure, and the simulation data, apply the principal component analysis dimensionality reduction technology to extract the key feature vectors, and use the time series modeling method to analyze the change trend of the building structure response under preset working conditions to identify the dynamic stress conditions, and generate a structure response prediction model according to the dynamic stress conditions and the expected load change of the building structure; Based on the structure response prediction model, select the initial support members according to the behavioral characteristics of the key stress points and the expected load change, encode the design parameters of the initial support members using the genetic algorithm to generate an initial population, and use the differential evolution algorithm to analyze the initial population to generate an optimized design parameter set; According to the change trend and dynamic stress conditions of the structure response prediction model, define the fitness function, and use the particle swarm optimization algorithm to iteratively search for the best design parameter combination. Based on the best design parameter combination and the optimized design parameter set, analyze the optimal design plan set corresponding to the fitness function, and generate the best combination of optimized design parameters based on the optimal design plan set combined with the ant colony optimization algorithm; According to the reinforcement learning framework, use the best combination as the action space of the intelligent agent in the building structure, and use the safety, stability, and economy of the building structure as the reward signal to adjust the strategy parameters of the intelligent agent to generate a reinforced concrete support design plan.

3. The method according to claim 2, wherein Define a fitness function according to the changing trend and dynamic force-bearing conditions of the structural response prediction model, and use the particle swarm optimization algorithm to iteratively search for the optimal combination of design parameters. Based on the optimal combination of design parameters and the set of optimized design parameters, analyze the set of optimal design schemes of the fitness function. Based on the set of optimal design schemes and combining the ant colony optimization algorithm, generate the optimal combination of optimized design parameters, including: Define a fitness function according to the changing trend and dynamic force-bearing conditions of the structural response prediction model, and combine different factors of the building structure, where the different factors of the building structure include safety, stability, and economy; Based on the fitness function, use the particle swarm optimization algorithm and the differential evolution algorithm to iteratively search for the optimal combination of design parameters, calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm, and apply an adaptive inertia weight to obtain the set of optimal design parameters. Screen out the design schemes with fitness function values higher than the preset threshold from the set of optimal design parameters to generate the set of optimal design schemes; Based on the set of optimal design schemes, combine the ant colony optimization algorithm to generate the optimal combination of optimized design parameters.

4. The method according to claim 3, characterized in that, The method of using the particle swarm optimization algorithm combined with the differential evolution algorithm based on the fitness function to iteratively search for the optimal combination of design parameters, calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm, and apply an adaptive inertia weight to obtain the set of optimal design parameters, and screen out the design schemes with fitness function values higher than the preset threshold from the set of optimal design parameters to generate the set of optimal design schemes, including: Calculate the fitness function value corresponding to each particle in the particle swarm optimization algorithm; Apply an adaptive inertia weight adjustment strategy to dynamically adjust the velocity update rule of the particle to generate an optimized particle swarm. Based on the optimized particle swarm, obtain the set of optimal design parameters; Screen out the design schemes with fitness function values higher than the preset threshold from the set of optimal design parameters, and based on the design schemes, generate the set of optimal design schemes.

5. The method according to claim 2, characterized in that, Based on the structural response prediction model, select the initial support members according to the behavior characteristics of the key force-bearing points and the expected load changes, and use the genetic algorithm to encode the design parameters of the initial support members to generate the initial population. Use the differential evolution algorithm to analyze the initial population to generate the set of optimized design parameters, including: According to the structural response prediction model, select the initial support members suitable for each key force-bearing point according to the description of the behavior characteristics of the key force-bearing points and the expected load changes; Use the genetic algorithm to encode the design parameters of the initial support members to obtain the set of initial support members. Based on the set of initial support members, construct the initial population, and apply the differential evolution algorithm to perform the construction and mutation operations of the difference vectors for the individuals of the initial population to obtain the optimized population; Based on the optimized population, generate the set of optimized design parameters.

6. The method according to claim 1, wherein Based on the key stress points, a hybrid modeling technique integrating deep learning and physical rules is used to simulate the behavioral characteristics of reinforced concrete materials at different building structural levels, generating a multi-level mechanical model, including: Based on the key stress points, a hybrid modeling technique integrating deep learning and physical rules is adopted to collect the key data of each key stress point, generating a basic data set. Using the basic data set, a hybrid modeling framework is constructed. The key data includes relevant experimental data, historical monitoring data, and theoretical calculation results; Based on the hybrid modeling framework, a deep learning model is used to simulate the behavior of reinforced concrete materials at the detailed level, predicting the performance changes of the reinforced concrete materials under different stress states, and obtaining a material behavior model at the detailed level; According to the material behavior model at the detailed level, combined with the structural characteristics of reinforced concrete components at the component level, the behavior of reinforced concrete components at the structural detail level is simulated. By introducing physical rules to guide model training, a component-level behavior model is generated; The behavior in the component-level behavior model is extended to the overall structural level, simulating the response of the entire building structure under different loading conditions, and using physical rules to constrain the output of the deep learning model, generating an overall structural-level response model; Based on the material behavior model at the detailed level, the component-level behavior model, and the overall structural-level response model, a multi-level mechanical model is established.

7. The method according to claim 1, characterized in that, Using the Bayesian network and grey system theory to establish a comprehensive evaluation system, quantitatively analyzing the design scheme of the reinforced concrete support, obtaining feedback information, and using the feedback information to optimize the design scheme of the reinforced concrete support, generating an optimized design scheme for the reinforced concrete support, including: According to the key factors of the design scheme of the reinforced concrete support, evaluation indicators are defined to construct a comprehensive evaluation index system. The key factors include safety, stability, and economy; Using the Bayesian network, risk assessment is carried out on the comprehensive evaluation index system, dynamically adjusting the probability relationship between different risk factors in the design scheme of the reinforced concrete support, and obtaining a risk assessment result; Applying the grey system theory to analyze the uncertainty of the design scheme of the reinforced concrete support, using the grey relational analysis method, and generating an uncertainty analysis report; Sorting out the risk assessment result and the uncertainty analysis report to form a comprehensive evaluation system. According to the comprehensive evaluation system, a quantitative score is given to the design scheme of the reinforced concrete support, and a quantitative scoring result is obtained; Analyzing the data obtained from monitoring the building structure, the expert review opinions, and the feedback information from relevant parties, and obtaining a summary of the feedback information; Based on the summary of the feedback information and the quantitative scoring result, using genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to adjust the design scheme of the reinforced concrete support, generating an optimized design scheme for the reinforced concrete support.

8. An optimization system for the design of reinforced concrete supports, characterized in that, Including: An analysis module for multi-dimensional analysis of various loads expected to be borne by the building to determine the key stress points in the building structure; A simulation module, configured to simulate the behavioral characteristics of reinforced concrete materials at different building structural levels based on the key stress points, using a hybrid modeling technique that integrates deep learning and physical rules, and generate a multi-level mechanical model; A generation module, configured to establish an intelligent decision-making process based on the multi-level mechanical model in combination with the data obtained by monitoring the building structure, and use genetic algorithms, particle swarm optimization, and reinforcement learning frameworks to generate a design solution for reinforced concrete supports; A quantification module, configured to establish a comprehensive evaluation system using Bayesian networks and grey system theory, quantitatively analyze the design solution for reinforced concrete supports, obtain feedback information, and use the feedback information to optimize the design solution for reinforced concrete supports to generate an optimized design solution for reinforced concrete supports.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an optimization method for the design of reinforced concrete supports as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, A computer program is stored, and when the computer program is executed by a computer, it implements an optimization method for the design of reinforced concrete supports as described in any one of claims 1 to 7.

Images