An ai-driven multi-objective optimization method for building structures

Abstract

The application relates to the technical field of building structure optimization, in particular to an AI-driven building structure multi-objective optimization method. Building characteristic parameters are input, a generative artificial intelligence model embedded with structure specification constraints is used to generate multiple initial structure arrangement paths; the initial structure arrangement paths are screened based on the number of conversion beams, the size and section size of a column grid and the supply cycle of high-strength concrete to obtain to-be-optimized paths; a multi-objective evolutionary algorithm is used to iteratively optimize the to-be-optimized paths and receive path preference operations to obtain to-be-selected paths; the to-be-selected paths are subjected to seismic performance verification and reliability verification, and the paths passing the double verification are marked as optimal paths. The generative exploration, multi-objective evolutionary optimization and double verification are deeply integrated, and the path availability, optimization efficiency and engineering reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of building structure optimization technology, and in particular to an AI-driven multi-objective optimization method for building structures. Background Technology

[0002] In the field of building structural design and optimization technology, the quality of path selection and the efficiency of multi-objective trade-offs during the conceptual design phase are the core quality assurance units of the system. Traditional paths usually rely on empirical selection or local fine-tuning using finite element parameterization, lacking quantitative optimization methods based on the global performance of the structure, making it difficult to evaluate the merits of the macroscopic structural layout in a timely manner. While using generative AI alone for path exploration can expand the search breadth, the output results lack physical consistency, making it impossible to distinguish between reasonable structural stress and random algorithmic artifacts, resulting in a high misuse rate. When existing methods encounter cost overruns or insufficient seismic performance, they repeatedly adjust the column grid or shear wall positions, resulting in high trial-and-error costs and low efficiency in the adjustment process, and the adjustment range is difficult to quantify, leading to long-term insufficient accuracy in path design for multiple projects and batches.

[0003] Chinese Patent Publication No. CN121328329A discloses an AI-based automatic optimization system for BIM parametric design. The method is characterized by calculating indices for structure, energy consumption, constructability, and regulatory compliance, and triggering parameter fine-tuning, algorithm iteration, or path regeneration based on the severity index of the problems. This method avoids the resource waste caused by global recalculation. However, its optimization objective is only a single comprehensive index, failing to achieve a Pareto trade-off between multiple conflicting objectives such as cost, seismic resistance, and carbon emissions. Furthermore, the hierarchical feedback mechanism still starts from a single initial path, unable to escape local optima for path-level global exploration. Summary of the Invention

[0004] To address these issues, this invention provides an AI-driven multi-objective optimization method for building structures. This method overcomes the technical problems in existing technologies where, in the multi-objective path selection scenario during the conceptual design phase of building structures, optimization intervention is too late, the paths output by generative AI lack physical rationality and cannot be used for multi-objective trade-offs, and the large solution set of multi-objective optimization outputs leads to a heavy cognitive burden on decision-making. These problems result in low optimization efficiency, low availability of generated paths, and low efficiency in final selection and long-term detection accuracy.

[0005] To achieve the above objectives, this invention provides an AI-driven multi-objective optimization method for building structures, comprising: Generative artificial intelligence models are constructed by embedding architectural feature parameters from historical periods into constraints of different structural codes. Based on building feature parameters, a generative artificial intelligence model embedding various structural specification constraints generates several corresponding initial structural layout paths; Each of the initial structural layout paths is filtered based on path constraint information, and a predetermined number of initial structural layout paths are selected as paths to be optimized. The path to be optimized is iteratively optimized based on a multi-objective evolutionary algorithm to determine the candidate path corresponding to each initial structure layout path. For each of the candidate paths, a high-precision finite element model is called to verify the seismic performance. The path that passes the seismic performance verification is marked as the optimal candidate path. The reliability of each of the optimal candidate paths is verified using Monte Carlo simulation, and the paths that pass the reliability verification are marked as the optimal paths.

[0006] Furthermore, the building characteristic parameters include building outline, area ratio of each functional area, load, seismic fortification level, and site category.

[0007] Furthermore, the process of constructing the embedding structure specifications and constraints of the generative artificial intelligence model includes: In the iterative denoising process of generative artificial intelligence models, gradient projection correction with normalization constraints is applied to the intermediate results of each iterative denoising step. The aforementioned constraints include constraints on the economic range of column spacing, the ratio of the total area of ​​shear walls to the floor area, the ratio of the core tube area to the standard floor area, and the constraint on the vertical continuity of columns and shear walls between floors.

[0008] Furthermore, the initial structural layout path includes column grid spacing, shear wall placement, and core tube outline.

[0009] Furthermore, the path constraint information includes the number of transfer beams, column grid dimensions and cross-sectional dimensions, and the high-strength concrete supply cycle.

[0010] Furthermore, the multi-objective evolutionary algorithm is the NSGA-III algorithm; during the iterative optimization process, after each predetermined number of optimizations, a preset path preference is received to further limit the optimization range.

[0011] Furthermore, the process of iteratively optimizing the path to be optimized includes: The cost, seismic performance, and carbon emission objective function values ​​of each path are calculated based on a single-iteration path optimization. The cost objective function includes concrete volume, total concrete cost, steel reinforcement weight, total steel reinforcement cost, formwork area, total formwork cost, and estimated foundation cost. The seismic performance objective function includes the inter-story drift angle and the stress ratio of each component; The carbon emission objective function includes concrete volume, concrete carbon emissions, steel reinforcement weight, and steel reinforcement carbon emissions.

[0012] Furthermore, the process of verifying the seismic performance of each candidate path includes: The seismic performance index of each candidate path under rare earthquake action is calculated by calling a high-precision finite element model. The seismic performance index includes inter-story drift angle and component stress ratio. If the inter-story drift angle is greater than the predetermined inter-story drift angle threshold or the stress ratio of any component is greater than the predetermined stress ratio threshold, then the seismic performance verification of the candidate path is deemed unsuccessful and the path is discarded. If the inter-story drift angle is less than or equal to the predetermined inter-story drift angle threshold and the stress ratio of all components is less than or equal to the predetermined stress ratio threshold, then the seismic performance verification of the candidate path is deemed successful, the candidate path is marked as the optimal candidate path and output.

[0013] Furthermore, the process of using Monte Carlo simulation to verify the reliability of each optimal candidate path includes: The Monte Carlo simulation considers the random variability of concrete strength, steel yield strength, and variable load to calculate the reliability index of the optimal candidate path. If the reliability index is less than the predetermined reliability index threshold, the reliability verification of the optimal candidate path is deemed to have failed. If the reliability index is greater than or equal to the reliability index threshold, the reliability verification of the optimal candidate path is deemed to have passed, and the optimal candidate path is marked as the optimal path.

[0014] Furthermore, the optimal path includes the design path of the complete architectural design drawings, as well as the cost reduction rate, carbon emission reduction rate, and the reduction in inter-story drift angle.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides an AI-driven multi-objective optimization method for building structures. Through a phased architecture involving artificial intelligence to generate multiple initial paths, select paths to be optimized, perform multi-objective evolutionary iterative optimization, finite element verification, and Monte Carlo reliability verification, it deeply integrates the global exploration capabilities of generative artificial intelligence, the conflict trade-off capabilities of multi-objective evolutionary algorithms, and high-fidelity numerical verification methods. Under multiple conflicting objectives such as cost, seismic performance, and carbon emissions, it efficiently outputs building structure design paths that combine global optimality and engineering feasibility. This method elevates the optimization intervention stage from traditional local fine-tuning of construction drawings to global selection in conceptual design, effectively improving optimization efficiency, path availability, the completeness of multi-objective trade-offs, and the engineering reliability of the final design.

[0016] In particular, this invention uses a generative artificial intelligence model with embedded structural specification constraints to apply gradient projection corrections to the intermediate results of each iteration of denoising, including the economic range of column spacing, shear wall area ratio, core tube ratio, and vertical continuity constraints. This automatically generates multiple initial structural layout paths with reasonable structures and different styles, ensuring the physical feasibility of the paths from the source and improving the availability of the generated paths.

[0017] In particular, by setting up a screening process, the present invention converts the number of beams, column grid size and cross-sectional size and high-strength concrete supply cycle, and quickly selects a predetermined number of paths to be optimized from the initial path. This can efficiently inject difficult-to-code engineering intuition into the optimization process with extremely low cognitive cost, thereby improving the quality of the optimization starting point.

[0018] In particular, this invention employs a third-generation multi-objective evolutionary algorithm based on a non-dominated sorting genetic algorithm to iteratively optimize the path to be optimized. In each generation, the objective function values ​​of cost, seismic performance, and carbon emissions for each path are calculated. This allows for the simultaneous handling of multiple conflicting optimization objectives, overcoming the limitations of traditional single-objective optimization or local fine-tuning in achieving comprehensive improvement. It also enhances the global optimality and completeness of decision information in multi-objective optimization.

[0019] In particular, by introducing a preference guidance mechanism, the present invention receives a preset path preference to guide the optimization direction after each predetermined number of optimizations in the multi-objective evolutionary algorithm. This avoids the dilemma of difficult selection caused by a large number of solution sets output by traditional multi-objective optimization, and improves the focusing efficiency and path acceptability of the final decision.

[0020] In particular, by sequentially verifying the seismic performance and reliability of the selected paths, this invention ensures that the final output path not only meets the seismic code requirements in nominal terms, but also has sufficient safety margin under the uncertainties of construction and actual use, thereby improving the engineering reliability of the optimization results.

[0021] In particular, by outputting the final optimal path, which includes the cost reduction rate, carbon emission reduction rate, and inter-story drift angle reduction compared to the path to be optimized, this invention can intuitively demonstrate the optimization effect in a quantitative manner, providing a clear basis for comparing the economy, low carbon emissions, and safety for design decisions, and improving the interpretability of optimization results and the efficiency of engineering transformation. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating the steps of the AI-driven multi-objective optimization method for building structures in this embodiment. Figure 2 A flowchart illustrating the steps for generating the initial structural layout path in this embodiment; Figure 3 This is a flowchart illustrating the steps of multi-objective optimization in this embodiment; Figure 4 This is a flowchart of the dual verification steps in this embodiment. Detailed Implementation

[0023] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0024] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0025] This embodiment provides an AI-driven multi-objective optimization system for building structures, including: The initial structural layout path generation module includes a generative artificial intelligence model with embedded structural specification constraints, which is used to automatically generate multiple initial structural layout paths based on input building feature parameters.

[0026] The filtering module, which is connected to the initial structural layout path generation module, is used to present the generated initial structural layout path in a visual manner. It performs a filtering operation on the path based on the number of conversion beams, column grid size and cross-sectional size and high-strength concrete supply cycle, and selects a predetermined number of initial structural layout paths as paths to be optimized.

[0027] A multi-objective evolutionary optimization module, which is connected to the screening module, is used to iteratively optimize the path to be optimized using a multi-objective evolutionary algorithm to determine several candidate paths.

[0028] The seismic performance verification module, which is connected to the multi-objective evolutionary optimization module, is used to call a high-precision finite element model for the candidate paths, determine whether the inter-story drift angle and component stress ratio of each candidate path meet the standard threshold, mark the candidate paths that have passed the seismic performance verification as the optimal candidate paths and output them.

[0029] The reliability verification module, which is connected to the seismic performance verification module, is used to consider the random variability of concrete strength, steel yield strength and variable load for the optimal candidate path. It calculates the reliability index through Monte Carlo simulation, determines whether the reliability index reaches the predetermined threshold, marks the optimal candidate path that has passed the reliability verification as the optimal path and outputs it.

[0030] This embodiment constructs a system architecture that integrates an initial structural layout path generation module, a screening module, a multi-objective evolutionary optimization module, a seismic performance verification module, and a reliability verification module. It deeply integrates the global exploration capability of generative AI, the conflict trade-off capability of multi-objective evolutionary algorithms, and a dual verification mechanism, realizing a closed-loop automated optimization process from the conceptual design stage to the final path output. This effectively improves the efficiency, path availability, and engineering reliability of multi-objective optimization of building structures.

[0031] Please see Figure 1 This is a flowchart illustrating the AI-driven multi-objective optimization method for building structures in this embodiment; the embodiment provides an AI-driven multi-objective optimization method for building structures, including: Step S1: Construct a generative artificial intelligence model by embedding the building feature parameters of the historical cycle into different structural specification constraints; Step S2: Based on the building feature parameters, generate several corresponding initial structural layout paths through the generative artificial intelligence model embedded with the constraints of each structural specification. Step S3: Filter each of the initial structural layout paths based on path constraint information, and select a predetermined number of initial structural layout paths as paths to be optimized. Step S4: Iteratively optimize the path to be optimized based on the multi-objective evolutionary algorithm to determine the candidate paths corresponding to each initial structure layout path; Step S5: For each of the candidate paths, call the high-precision finite element model to verify the seismic performance, and mark the path that passes the seismic performance verification as the optimal candidate path; Step S6: Perform reliability verification on each of the optimal candidate paths using Monte Carlo simulation, and mark the paths that pass the reliability verification as the optimal paths.

[0032] In this embodiment, the building characteristic parameters include building outline, area ratio of each functional area, load, seismic fortification level, and site category; wherein, building outline is data describing the geometric shape and vertical dimensions of the building; area ratio of each functional area refers to the area ratio of different areas inside the building classified according to their use; load refers to the quantitative value of various forces borne by the building during construction and use; seismic fortification level refers to the seismic intensity level on which the building is based when conducting seismic design; site category refers to the category of the building site divided according to the thickness of the overburden layer and the equivalent shear wave velocity of the soil layer, usually divided into Class I, Class II, Class III, and Class IV.

[0033] This embodiment employs a phased architecture that utilizes artificial intelligence to generate multiple initial paths, filter paths to be optimized, perform multi-objective evolutionary iterative optimization, verify seismic performance, and verify Monte Carlo reliability. This architecture enables the efficient output of building structure design paths that combine global optimality and engineering feasibility under multiple conflicting objectives such as cost, seismic performance, and carbon emissions.

[0034] Please see Figure 2 This is a flowchart illustrating the steps involved in generating the initial structural layout path in this embodiment; the process of generating the initial structural layout path in this embodiment includes: Input building feature parameters into the generative artificial intelligence model; Generative artificial intelligence models undergo iterative denoising to obtain interim results; Generative AI models perform gradient correction on interim results; Determine whether the preset number of iterations for denoising has been reached: If the number of iterations for denoising does not reach the preset number of iterations, the generative artificial intelligence model will continue to iterate and denoise the intermediate results after gradient correction. If the number of iterations for denoising reaches the preset number of iterations, the initial structure layout path will be output.

[0035] Specifically, the generative artificial intelligence model starts with an initial noise tensor and iterates to denoise according to a preset total number of iterations. In each iteration, the model predicts the noise component contained in the current step based on the noise tensor and building feature parameters of the current step, and subtracts the predicted noise from the current tensor to obtain an intermediate result that is closer to the real data distribution.

[0036] Understandably, generative artificial intelligence models can output different initial structural layout paths based on different initial noise tensors, even with the same building feature parameters.

[0037] Specifically, in the iterative denoising process of the generative artificial intelligence model, gradient projection correction with normative constraints is applied to the intermediate results of each iterative denoising step; the normative constraints include the economic range constraint of column spacing, the ratio constraint of the total area of ​​shear walls to the floor area, the ratio constraint of the core tube area to the standard floor area, and the vertical continuity constraint of columns and shear walls between floors.

[0038] Understandably, the preset number of iteration steps is determined based on a balance between the generation quality and efficiency of the initial structure layout path. The more iteration steps, the more refined the generation process and the higher the physical rationality of the path, but the computation time increases. The fewer iteration steps, the faster the generation speed, but it may lead to insufficient path details or violation of specification constraints.

[0039] Specifically, the initial structural layout path includes column grid spacing, shear wall placement, core tube outline, and also includes a rapidly estimated total cost, estimated steel consumption, and structural natural vibration period.

[0040] This embodiment uses a generative artificial intelligence model to iteratively denoise and obtain the initial structural layout path. The structural design specification constraints are embedded in each denoising iteration by gradient projection correction. Under the premise of ensuring the rationality of the generated path, multiple initial structural layout paths that meet the specification requirements can be automatically generated, thereby improving the efficiency of initial structural design.

[0041] Please see Figure 3 This is a flowchart illustrating the iterative optimization steps in this embodiment; the iterative optimization process in this embodiment includes: Filter to obtain paths to be optimized; A multi-objective evolutionary algorithm is used to perform the first generation of evolution on the path to be optimized. Multi-objective evolutionary algorithms accept pre-defined path preferences; A multi-objective evolutionary algorithm is used to evolve the path to be optimized after the first generation of evolution a predetermined number of times; The multi-objective evolutionary algorithm again receives the preset path preference; A multi-objective evolutionary algorithm is used to evolve the path to be optimized after a predetermined number of evolutions to determine several candidate paths.

[0042] Specifically, the process of selecting and obtaining paths to be optimized includes: The initial structural layout path is screened based on three aspects: the number of transfer beams, the size and cross-sectional dimensions of the column grid, and the supply cycle of high-strength concrete, in order to obtain the path to be optimized.

[0043] Specifically, the multi-objective evolutionary algorithm is the NSGA-III algorithm.

[0044] Specifically, during the iterative optimization process of the multi-objective evolutionary algorithm, each generation calculates the cost, seismic performance, and carbon emission objective function values ​​for each path. The cost objective function includes concrete volume, total concrete price, steel reinforcement weight, total steel reinforcement price, formwork area, total formwork price, and estimated foundation cost. The seismic performance objective function includes inter-story drift angle and stress ratio of each component. The carbon emission objective function includes concrete volume, concrete carbon emissions, steel reinforcement weight, and steel reinforcement carbon emissions.

[0045] This embodiment employs the NSGA-III multi-objective evolutionary algorithm and simultaneously calculates the three conflicting objective function values ​​of cost, seismic performance, and carbon emissions during the iterative optimization process, thereby achieving a multi-objective trade-off for the structural path.

[0046] In this embodiment, after each predetermined number of optimizations, the multi-objective evolutionary algorithm receives a preset path preference to further limit the optimization range. The predetermined number of optimizations is determined based on the actual engineering situation. For high-dimensional and complex optimization problems, a smaller predetermined number of optimizations is preferable to increase the interaction frequency and correct the search direction in a timely manner. For simple problems, a larger predetermined number of optimizations can be used to reduce the interaction burden.

[0047] This embodiment receives a preset path preference after each predetermined number of evolutions, which can guide the optimization to converge to a region that meets the actual needs, effectively solving the problem of the large output solution set of traditional multi-objective optimization leading to decision-making difficulties.

[0048] Please see Figure 4 This is a flowchart illustrating the steps of dual verification in this embodiment; the dual verification process in this embodiment includes: Get the candidate paths; The inter-layer displacement angle and stress ratio of each component of the candidate path are calculated using a high-precision finite element model. Determine whether the inter-story drift angle and the stress ratio of each component are greater than the threshold: If the inter-story drift angle is greater than the predetermined inter-story drift angle threshold or the stress ratio of any component is greater than the predetermined stress ratio threshold, then the seismic performance verification of the candidate path is deemed unsuccessful and the path is discarded. If the inter-story drift angle is less than or equal to the predetermined inter-story drift angle threshold and the stress ratio of all components is less than or equal to the predetermined stress ratio threshold, then the seismic performance verification of the candidate path is deemed to be passed, the candidate path is marked as the optimal candidate path and output. Reliability indicators were obtained by performing Monte Carlo simulations on the optimal candidate path; Determine if the reliability index is greater than the threshold: If the reliability index is less than the reliability index threshold, the reliability verification of the optimal candidate path is deemed to have failed, and the path is discarded. If the reliability index is greater than or equal to the reliability index threshold, the reliability verification of the optimal candidate path is deemed to have passed, and the optimal candidate path is marked as the optimal path and output.

[0049] Specifically, the high-precision finite element model in this embodiment includes the following modeling details: Rigid domain modeling of beam-column joints; by setting the length of the rigid domain, the true stiffness of the joint area is represented, avoiding incorrect calculations of beam end bending moment and column side shear force; it can be understood that the length of the rigid domain is determined by taking the smaller value of half the column section height and half the beam section height.

[0050] Modeling of openings in shear walls: For coupling beam openings that are common in actual engineering, the model refines the mesh around the opening, uses nonlinear beam elements for the coupling beam above the opening, uses wall elements for the wall segments on both sides of the opening, and sets stress concentration monitoring points at the corners of the opening.

[0051] Modeling out-of-plane stiffness of floor slabs: By using membrane elements or composite shell elements, the coupling effect of in-plane and out-of-plane stiffness of floor slabs is considered, avoiding the overestimation of stiffness and distortion of internal forces caused by the assumption of rigid floor slabs in traditional elastic analysis.

[0052] This embodiment constructs a high-precision finite element model that includes rigid domain modeling of beam-column joints, shear wall opening modeling, and out-of-plane stiffness modeling of floor slabs. This model can more realistically simulate the nonlinear mechanical behavior of structures under rare earthquakes, avoiding stiffness distortion and misjudgment of internal forces caused by traditional simplified models, thereby improving the accuracy and reliability of seismic performance index verification.

[0053] Specifically, the inter-story drift angle is the ratio of the maximum horizontal displacement of a floor under rare earthquake conditions to the floor height. It represents the overall deformation capacity and collapse resistance reserve of the structure under strong earthquake conditions. The larger the value, the more severe the structural deformation and the worse the earthquake resistance. The specific calculation process includes: extracting the absolute horizontal displacement of all nodes in each floor; calculating the difference between the maximum and minimum displacements in the same floor to obtain the relative inter-story displacement; removing the relative displacement and dividing it by the floor height to obtain the inter-story drift angle of that floor; and taking the largest inter-story drift angle among all floors as the inter-story drift angle index of the structure.

[0054] Understandably, the inter-story drift angle threshold is determined based on the inter-story drift angle limit under rare earthquake action in the current national seismic design code, and the specific value varies depending on the structural system and structural type.

[0055] Specifically, the stress ratio of a component is the ratio of the actual internal force of the component under the most unfavorable load combination to the design value of the component's bearing capacity. It represents the strength reserve level of the component. The larger the value, the closer the component is to its ultimate bearing capacity and the smaller the safety redundancy.

[0056] It is understandable that setting the stress ratio threshold to one means that the load-bearing capacity requirement of a component is not greater than the load-bearing capacity supply. In practice, for critical components or important parts, a more stringent threshold can be set to provide additional safety redundancy.

[0057] This embodiment establishes a seismic performance evaluation system based on two indicators by clearly defining and calculating the inter-story drift angle and component stress ratio. This system can comprehensively evaluate the candidate paths from two dimensions: the overall structural collapse resistance and the strength of local components, ensuring the safety and reliability of the final output path under rare earthquakes.

[0058] Specifically, when conducting seismic performance verification, if the verification fails, the specific cause classification will be output to pinpoint the root cause of the problem: Inter-story drift angle exceeding the limit: This indicates that the overall stiffness of the structure is too low and the deformation is too large under rare earthquakes. The reasons are insufficient number of shear walls, small core tube size, or excessive column grid spacing.

[0059] The stress ratio of a component exceeds the limit: This indicates that the actual internal force of the component under rare earthquake conditions has exceeded its bearing capacity design value, that is, the component strength is insufficient. The reason is that the load effect is too large or the bearing capacity design value is too low.

[0060] This embodiment classifies and outputs the reasons for non-compliance during the seismic performance index determination process, transforming high-precision finite element verification results into interpretable engineering diagnostic information. This enables rapid identification of the root causes of insufficient seismic performance of structures, thereby allowing for targeted adjustments to the design path and improving the feedback efficiency and accuracy of design corrections in the verification process.

[0061] Specifically, the process of obtaining reliable indicators by performing Monte Carlo simulation calculations on the optimal candidate path in this embodiment includes: Determine the random variables and their probability distributions, including concrete strength, steel yield strength, and variable load; Perform a predetermined number of independent samplings, each time randomly generating several sample values ​​according to the probability distribution of each random variable; For each set of sampled values, the structural analysis model is called to calculate the inter-story drift angle; The inter-story drift angle is compared with the inter-story drift angle threshold, and the number of samplings for drift angle exceeding the limit is counted. The failure probability is obtained by dividing the number of times the limit is exceeded by the total number of samplings, and then converted into a reliability index by the inverse function of the standard normal distribution.

[0062] Specifically, the reliability index is a quantitative value that measures the structure's ability to maintain a predetermined level of safety under the combined influence of multiple uncertainties. It represents the scale of the gap between the actual resistance of the structure and the load effect relative to the uncertainty. The larger the value, the more sufficient the safety reserve of the structure and the lower the expected probability of failure.

[0063] Understandably, the reliability index threshold is determined based on the reliability index limit in the current unified standard for building structure reliability design, and the specific value varies depending on the safety level and the type of failure.

[0064] This invention uses Monte Carlo simulation to perform probabilistic reliability analysis on the optimal candidate path, incorporates the random variability of material strength and load into the evaluation system, and uses the reliability index converted from failure probability as the judgment basis to ensure that the final output optimal path still has sufficient safety margin in uncertain environments, thereby improving the robustness of structural design and engineering reliability.

[0065] Specifically, during reliability verification, if a verification fails, the system will output a detailed classification of the reasons to help pinpoint the root cause of the problem: Material strength fluctuation sensitivity: If the simulation results show that the failure rate is significantly higher in the sampling with low strength values, it can be determined that the structure is sensitive to material strength fluctuations.

[0066] High load uncertainty: If the simulation results show that the failure rate is significantly higher in the sample with larger load values, it can be determined that the structure is sensitive to load uncertainty.

[0067] Insufficient safety margin: This refers to a situation where the structural response is close to or has reached the failure threshold when both material strength and load are within the normal fluctuation range.

[0068] This embodiment transforms the statistical results of Monte Carlo simulation into interpretable engineering diagnostic information by finely classifying and outputting the reasons for non-compliance during the reliability index determination process. This enables the rapid identification of the root causes of structural reliability deficiencies, thereby allowing for targeted adjustments to the design path and improving the closed-loop efficiency and engineering operability of the optimization process.

[0069] Specifically, the optimal output path includes complete architectural design drawings, as well as the cost reduction rate, carbon emission reduction rate, and inter-story drift angle reduction compared to the path to be optimized.

[0070] The embodiments described above are for illustrative purposes only and are not intended to limit the scope of this application. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this application without departing from the spirit and scope of the technical solutions of this application, and all such modifications or substitutions should be covered within the scope of the claims of this application.

Claims

1. An AI-driven multi-objective optimization method for building structures, characterized in that, include, Generative artificial intelligence models are constructed by embedding architectural feature parameters from historical periods into constraints of different structural codes. Based on building feature parameters, a generative artificial intelligence model embedding various structural specification constraints generates several corresponding initial structural layout paths; Each of the initial structural layout paths is filtered based on path constraint information, and a predetermined number of initial structural layout paths are selected as paths to be optimized. The path to be optimized is iteratively optimized based on a multi-objective evolutionary algorithm to determine the candidate path corresponding to each initial structure layout path. For each of the candidate paths, a high-precision finite element model is called to verify the seismic performance. The path that passes the seismic performance verification is marked as the optimal candidate path. The reliability of each of the optimal candidate paths is verified using Monte Carlo simulation, and the paths that pass the reliability verification are marked as the optimal paths.

2. The AI-driven multi-objective optimization method for building structures according to claim 1, characterized in that, The building characteristic parameters include the building outline, the area ratio of each functional area, the load, the seismic fortification level, and the site category.

3. The AI-driven multi-objective optimization method for building structures according to claim 1, characterized in that, The process of constructing the embedding structure specifications and constraints of the generative artificial intelligence model includes: In the iterative denoising process of generative artificial intelligence models, gradient projection correction with normalization constraints is applied to the intermediate results of each iterative denoising step. The aforementioned constraints include constraints on the economic range of column spacing, the ratio of the total area of ​​shear walls to the floor area, the ratio of the core tube area to the standard floor area, and the constraint on the vertical continuity of columns and shear walls between floors.

4. The AI-driven multi-objective optimization method for building structures according to claim 1, characterized in that, The initial structural layout path includes column grid spacing, shear wall placement, and core tube outline.

5. The AI-driven multi-objective optimization method for building structures according to claim 1, characterized in that, The path constraint information includes the number of transfer beams, column grid dimensions and cross-sectional dimensions, and the high-strength concrete supply cycle.

6. The AI-driven multi-objective optimization method for building structures according to claim 1, characterized in that, The multi-objective evolutionary algorithm is the NSGA-III algorithm; during the iterative optimization process, after each predetermined number of optimizations, a preset path preference is received to further limit the optimization range.

7. The AI-driven multi-objective optimization method for building structures according to claim 6, characterized in that, The process of iteratively optimizing the path to be optimized includes: The cost, seismic performance, and carbon emission objective function values ​​of each path are calculated based on a single-iteration path optimization. The cost objective function includes concrete volume, total concrete cost, steel reinforcement weight, total steel reinforcement cost, formwork area, total formwork cost, and estimated foundation cost. The seismic performance objective function includes the inter-story drift angle and the stress ratio of each component; The carbon emission objective function includes concrete volume, concrete carbon emissions, steel reinforcement weight, and steel reinforcement carbon emissions.

8. The AI-driven multi-objective optimization method for building structures according to claim 1, characterized in that, The process of verifying the seismic performance of each candidate path includes: The seismic performance index of each candidate path under rare earthquake action is calculated by calling a high-precision finite element model. The seismic performance index includes inter-story drift angle and component stress ratio. If the inter-story drift angle is greater than the predetermined inter-story drift angle threshold or the stress ratio of any component is greater than the predetermined stress ratio threshold, then the seismic performance verification of the candidate path is deemed unsuccessful and the path is discarded. If the inter-story drift angle is less than or equal to the predetermined inter-story drift angle threshold and the stress ratio of all components is less than or equal to the predetermined stress ratio threshold, then the seismic performance verification of the candidate path is deemed successful, the candidate path is marked as the optimal candidate path and output.

9. The AI-driven multi-objective optimization method for building structures according to claim 1, characterized in that, The process of using Monte Carlo simulation to verify the reliability of each optimal candidate path includes: The Monte Carlo simulation considers the random variability of concrete strength, steel yield strength, and variable load to calculate the reliability index of the optimal candidate path. If the reliability index is less than the predetermined reliability index threshold, the reliability verification of the optimal candidate path is deemed to have failed. If the reliability index is greater than or equal to the reliability index threshold, the reliability verification of the optimal candidate path is deemed to have passed, and the optimal candidate path is marked as the optimal path.

10. The AI-driven multi-objective optimization method for building structures according to claim 9, characterized in that, The optimal path includes the design path of the complete architectural design drawings, as well as the cost reduction rate, carbon emission reduction rate, and the reduction in inter-story drift angle.

Images