BIM and MARL fused complex structure steel bar intelligent arrangement system

Abstract

The invention, which relates to the technical field of information, discloses a BIM and MARL-fused complex structure steel bar intelligent arrangement system comprising a data acquisition module, a three-dimensional modeling module, a feature processing module, a scheme generation module, a scheme optimization module and a scheme output module. The data acquisition module extracts geometric and stress data of the BIM, and a final data set is obtained after preprocessing; the three-dimensional modeling module constructs a three-dimensional structure model through geometric analysis, optimization and standardization algorithms; the feature processing module calculates stress distribution and performs dimension reduction on a feature set; the scheme generation module generates an initial arrangement scheme set through a genetic algorithm; the scheme optimization module obtains a global coordination scheme set through multi-algorithm collaborative optimization; and the scheme output module generates a three-dimensional visual model and BIM compatible arrangement data. According to the method, the problem of arrangement of the reinforcing steel bars with complex structures is solved, and the accuracy and the engineering adaptability of reinforcing steel bar arrangement are improved.

Description

Technical Field

[0001] This invention relates to the field of information technology, specifically to an intelligent steel reinforcement layout system for complex structures that integrates BIM and MARL. Background Technology

[0002] Structural design occupies a central position in modern engineering, its importance lying in ensuring building safety, optimizing resource utilization, and improving construction efficiency. With the development of digital technology, intelligent design methods based on Building Information Modeling (BIM) have become a research hotspot; however, existing methods still have significant limitations when dealing with complex structures.

[0003] First, the quantitative representation of geometric and stress features is not precise enough, making it difficult to effectively capture the inherent laws of complex structures. Second, the dimensionality reduction and optimization process of high-dimensional feature data lacks efficient integration, resulting in high computational costs and unstable results. Finally, the local and global coordination in multi-agent collaborative optimization is difficult to balance, affecting the overall performance of the solution.

[0004] Therefore, the key issues of this study are how to accurately extract geometric and stress features from BIM models, construct an efficient feature dimensionality reduction and multi-agent optimization framework, and generate a high-precision, visualized rebar layout scheme through global search and consistency adjustment. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent reinforcement layout system for complex structures that integrates BIM and MARL, solving the problem of low efficiency in reinforcement layout for complex structures.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent reinforcement layout system for complex structures integrating BIM and MARL, comprising: The data acquisition module is used to extract geometric and stress data from the building information model, obtain the final dataset, and transmit the final dataset to the 3D modeling module. The 3D modeling module uses 3D geometric analysis algorithms to process the final dataset, obtain a 3D structural model, and then transmits the 3D structural model to the feature processing module. The feature processing module uses the finite element analysis algorithm to calculate the stress distribution and generate a feature set. If the dimension of the feature set exceeds a preset threshold, the principal component analysis algorithm is used to reduce the dimension to obtain a dimension-reduced feature matrix, and the dimension-reduced feature matrix is ​​transmitted to the scheme generation module. The scheme generation module generates an initial layout scheme set through a genetic algorithm and then transmits the initial layout scheme set to the scheme optimization module. The scheme optimization module uses the particle swarm optimization algorithm to locally optimize the initial layout scheme set to obtain a locally optimized scheme set. If there are stress concentration areas, the simulated annealing algorithm is used to adjust and generate a modified scheme set. The modified scheme set is then optimized using a consistency evaluation function and a differential evolution algorithm to obtain a globally coordinated scheme set, which is then transmitted to the scheme output module. The scheme output module is used to generate a 3D visualization model and output layout scheme data.

[0007] As a further aspect of the present invention, the specific steps for obtaining the final dataset are as follows: Extract node coordinates and boundary conditions from the building information model to generate an initial dataset containing node coordinates and boundary conditions; The initial dataset is preprocessed using a data integration algorithm to obtain a standardized initial dataset; A lossless data compression algorithm is used to compress the standardized initial dataset to generate a compressed initial dataset; The compressed initial dataset is converted into a standard output format using a format conversion algorithm to obtain the final dataset.

[0008] As a further aspect of the present invention, the steps for processing the final dataset using a three-dimensional geometric analytical algorithm to obtain the three-dimensional structural model are as follows: The stereomicroscopy algorithm is used to analyze the geometric data in the final dataset, identify the topological relationships between nodes, fill in the missing geometric information, and generate a preliminary three-dimensional structural model. If the geometric accuracy of the initial 3D structural model is lower than the preset accuracy threshold, the cross-sectional dimensions and node position parameters of the model are adjusted through a geometric optimization algorithm to obtain an optimized 3D structural model. The coordinate mapping algorithm is used to standardize the optimized 3D structural model, generating a standardized 3D structural model.

[0009] As a further aspect of the present invention, the step of calculating the stress distribution and generating a feature set using a finite element analysis algorithm includes: Based on the node coordinates and boundary conditions in the final dataset, the three-dimensional structural model is divided into finite element elements, and the stress distribution of each element is calculated using the finite element analysis algorithm to obtain stress distribution data. If the maximum value in the stress distribution data exceeds the preset stress threshold, the parameters of the three-dimensional structural model of the corresponding region are adjusted by the local geometric optimization algorithm, the stress distribution is recalculated, and optimized stress distribution data is generated. The data integration algorithm combines node coordinates, boundary conditions, and optimized stress distribution data according to a preset structure to form a feature set.

[0010] As a further aspect of the present invention, the specific operation of obtaining the dimensionality-reduced feature matrix through principal component analysis algorithm is as follows: Calculate the contribution rate of each principal component, retain the principal components whose cumulative contribution rate reaches a preset proportion, and obtain the dimensionality-reduced feature matrix; The elements of the dimensionality-reduced feature matrix are normalized to a preset numerical range using a standardization algorithm to obtain a standardized feature matrix. If the variance of the standardized feature matrix is ​​lower than the preset variance threshold, the correlation between each feature and the reinforcement layout optimization objective is calculated using a feature selection algorithm, key features are extracted, and a simplified feature matrix is ​​obtained.

[0011] As a further aspect of the present invention, the step of generating an initial arrangement scheme set using a genetic algorithm is as follows: Based on the dimensionality-reduced feature matrix, a genetic algorithm is used to initialize a multi-agent population, with each agent corresponding to a steel reinforcement layout scheme. Set a fitness function and iteratively optimize the rebar layout scheme through crossover and mutation operations to generate an optimized scheme set; The set of optimal solutions is sorted in descending order by using a sorting algorithm based on the fitness values ​​calculated by the fitness function, resulting in a sorted set of solutions. If there are schemes in the sorting scheme set with fitness values ​​lower than the preset fitness threshold, the matrix adjustment algorithm is used to correct the feature vector parameters corresponding to the schemes to generate an initial arrangement scheme set.

[0012] As a further aspect of the present invention, the step of generating an initial arrangement scheme set using a genetic algorithm is as follows: Based on the dimensionality-reduced feature matrix, a genetic algorithm is used to initialize a multi-agent population, with each agent corresponding to a steel reinforcement layout scheme. Set a fitness function and iteratively optimize the rebar layout scheme through crossover and mutation operations to generate an optimized scheme set; The set of optimal solutions is sorted in descending order by using a sorting algorithm based on the fitness values ​​calculated by the fitness function, resulting in a sorted set of solutions. If there are schemes in the sorting scheme set with fitness values ​​lower than the preset fitness threshold, the matrix adjustment algorithm is used to correct the feature vector parameters corresponding to the schemes to generate an initial arrangement scheme set.

[0013] As a further aspect of the present invention, the specific steps for adjusting and generating the modified scheme set using the simulated annealing algorithm are as follows: If there is a stress concentration region in the local optimization scheme set where the stress value exceeds the preset stress threshold, then the feature vector set corresponding to the region is obtained through feature extraction technology. The vectors in the feature vector set include the rebar coordinates and stress value parameters of the stress concentration region. The initial temperature and cooling rate of the simulated annealing algorithm are set, the position parameters of each vector in the feature vector set are iteratively adjusted, and the adjusted fitness value is calculated based on the stress distribution value and geometric constraints to obtain the adjusted feature vector set. If there are still vectors with stress values ​​exceeding the threshold in the adjusted feature vector set, then a new vector is generated between adjacent reasonable vectors using vector interpolation technology to correct the rebar arrangement parameters and obtain the corrected feature vector set. An adjusted set of layout schemes is generated based on the modified feature vector set, thus obtaining the modified scheme set.

[0014] As a further aspect of the present invention, the step of obtaining the global coordination scheme set is as follows: The global consistency value of each scheme in the modified scheme set is calculated by using the global consistency evaluation function, and the consistency distribution feature set is obtained. A multi-agent collaborative mechanism is used to group the vectors in the consistent distribution feature set according to feature similarity, resulting in a grouped feature set; If there are vectors in the grouped feature set with a consistency value lower than the preset consistency threshold, then the mutation factor and crossover probability of the differential evolution algorithm are preset to adjust the vector parameters and obtain an optimized feature set. The matrix transformation algorithm is used to perform coordinate mapping on the optimized feature set to generate a preliminary set of layout schemes. If there are schemes in the initial layout scheme set with a global consistency value lower than the threshold, then the feature vector parameters are recombined using vector recombination technology to adjust the global reinforcement layout and obtain a globally coordinated scheme set.

[0015] As a further aspect of the present invention, the specific operations for generating a three-dimensional visualization model and outputting layout scheme data are as follows: Extract the coordinates, dimensions, and spacing parameters of the reinforcing bars from the global coordination scheme set, and generate a parameter set containing these parameters; Geometric modeling of the parameter set is performed using a 3D visualization algorithm to construct a 3D relationship model between the reinforcing bars and the main structure, generating a preliminary visualization model. If there are elements in the initial visualization model whose coordinate deviation exceeds the preset deviation threshold, the element coordinates are corrected through a vector adjustment algorithm to generate an optimized visualization model. Arrangement data is extracted from the optimized visualization model, and the arrangement data is compressed using a data compression algorithm; The compressed layout data is converted into a BIM platform-compatible format using a format conversion algorithm to generate the final layout scheme data.

[0016] This invention provides an intelligent reinforcement layout system for complex structures that integrates BIM and MARL, which has the following advantages compared with existing technologies: (1) The present invention preprocesses the geometric and stress data of the BIM model through the data acquisition module, and combines lossless compression and format conversion to ensure data integrity. The three-dimensional modeling module uses stereomicroscope algorithm to identify node topological relationships and complete geometric information, and improves model accuracy through geometric optimization and coordinate standardization. (2) This invention achieves scheme optimization through multi-algorithm collaboration. The scheme generation module generates multiple initial schemes using a genetic algorithm. The scheme optimization module combines particle swarm optimization algorithm to achieve precise adjustment of local parameters, simulated annealing algorithm to eliminate stress concentration, and differential evolution algorithm to ensure global coordination. The multi-algorithm linkage takes into account both local details and global adaptability, effectively avoiding the limitation of a single algorithm easily getting trapped in local optima. (3) The present invention generates a three-dimensional visualization model through the scheme output module, which facilitates intuitive verification of the feasibility of the scheme. At the same time, it converts the layout data into a BIM platform compatible format, seamlessly connecting with downstream construction layout, collaborative design and other processes. Attached Figure Description

[0017] Figure 1 This is the system principle block diagram of the present invention; Figure 2 This is a flowchart illustrating the steps involved in obtaining a three-dimensional structural model according to the present invention. Figure 3 This is a flowchart illustrating the steps involved in generating an initial layout scheme set according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1 like Figure 1 This application provides an intelligent reinforcement layout system for complex structures that integrates BIM and MARL, including: The data acquisition module is used to extract geometric and stress data from the building information model, obtain the final dataset, and transmit the final dataset to the 3D modeling module. In the field of building information modeling, acquiring the geometric and stress data of the BIM model is the core of constructing the final dataset; Geometric data includes the node coordinates of beams, columns, and slabs. For example, the beam node coordinates of a building are (0,0,0) and (10,0,0), in meters. The stress data includes external loads, such as a concentrated force of 100kN at the end of the beam, and the boundary conditions are recorded at the fixed end or hinge point, such as one end of the beam being fixed. The specific steps to obtain the final dataset are as follows: Extract node coordinates and boundary conditions from the building information model to generate an initial dataset containing node coordinates and boundary conditions; The initial dataset is preprocessed using a data integration algorithm to obtain a standardized initial dataset; Understandably, combining node coordinates, boundary conditions, and optimized stress distributions into a standard dataset through data integration is primarily to provide standard input for compression and format conversion. A lossless data compression algorithm is used to compress the standardized initial dataset to generate a compressed initial dataset; By employing lossless compression algorithms, the size of the initial dataset can be significantly compressed while retaining all information. The compressed initial dataset is easier to transmit and store, and significantly reduces data management costs, especially in large projects. The compressed initial dataset is converted into a standard output format using a format conversion algorithm to obtain the final dataset; Specifically, the format conversion transforms the compressed initial dataset into a standard output format, such as JSON. The converted dataset contains node coordinates, boundary conditions, and stress data, and is compatible with various analysis software. For example, JSON files can be directly imported into the structure analysis platform, making it easier for engineers to perform secondary development, while the standard format improves data universality and enhances cross-platform collaboration efficiency.

[0020] The 3D modeling module uses 3D geometric analysis algorithms to process the final dataset, obtain a 3D structural model, and then transmits the 3D structural model to the feature processing module. The feature processing module uses the finite element analysis algorithm to calculate the stress distribution and generate a feature set. If the dimension of the feature set exceeds a preset threshold, the principal component analysis algorithm is used to reduce the dimension to obtain a dimension-reduced feature matrix, and the dimension-reduced feature matrix is ​​transmitted to the scheme generation module. The steps for calculating stress distribution and generating a feature set using the finite element analysis algorithm include: Based on the node coordinates and boundary conditions in the final dataset, the three-dimensional structural model is divided into finite element elements, and the stress distribution of each element is calculated using the finite element analysis algorithm to obtain stress distribution data. Finite element analysis breaks down a structure into hundreds or even thousands of simple small units. Each unit is used to calculate stress using basic mechanics formulas, and then the results are integrated into an overall stress distribution. This clearly shows the stress differences in different regions. At the same time, the system also needs to determine whether there are stress concentration areas. The "unit-level stress data" output by finite element analysis is the only quantitative basis. Without this data, relying solely on the engineer's experience will lead to misjudgment or omission. If the maximum value in the stress distribution data exceeds the preset stress threshold, the parameters of the three-dimensional structural model of the corresponding region are adjusted by the local geometric optimization algorithm, the stress distribution is recalculated, and optimized stress distribution data is generated. If the maximum stress calculated by the finite element method exceeds the preset threshold, it indicates that the geometric parameters of the 3D structural model itself do not meet the stress safety requirements, rather than that the steel reinforcement is insufficient. In this case, if geometric optimization is skipped and the excess stress is directly offset by increasing the amount of steel reinforcement, two major problems will occur: The amount of steel reinforcement used far exceeded the design specifications, significantly increasing the project cost; Too many steel bars can lead to insufficient spacing between bars, making it impossible to pour concrete, or too many layers of steel bars can exceed the height of the component's cross-section, causing the model to become disconnected from the construction. By using a local geometry optimization algorithm, the geometric parameters are adjusted only in the stress-exceeding-threshold area, rather than the entire structure is modified, thus balancing the reduction of stress levels at the source, economy, and construction feasibility. The data integration algorithm combines node coordinates, boundary conditions, and optimized stress distribution data according to a preset structure to form a feature set. Node coordinates are stored in the BIM geometry database (in CAD or IFC format), boundary conditions are recorded in the BIM attribute file (in XML format), and optimized stress data are stored in the finite element calculation result file (in INP or ODB format). The inconsistent formats prevent the data from being directly called by the subsequent feature processing module and scheme generation module. By combining the data according to the preset structure through the data integration algorithm, the scattered data can be transformed into a standardized and related dataset. The specific steps for obtaining the dimensionality-reduced feature matrix using principal component analysis are as follows: Calculate the contribution rate of each principal component, retain the principal components whose cumulative contribution rate reaches a preset proportion, and obtain the dimensionality-reduced feature matrix; The original feature set obtained by the system from BIM and finite element analysis is usually extremely high in dimensionality, and may contain dozens or even hundreds of features such as node coordinates, cross-sectional dimensions, stress distribution, and boundary conditions. If it is directly input into the MARL multi-agent optimization algorithm, it will lead to a doubling of the number of algorithm iterations, a significant decrease in system operating efficiency, and an inability to meet the timeliness requirements of engineering design. On the other hand, these high-dimensional feature sets contain a large number of highly correlated redundant features, such as the strong correlation between "the normal stress on the top surface of the column" and "the cross-sectional height of the column". However, these features essentially reflect the same information. Retaining all features will lead to repeated calculations by the algorithm, which will not only waste resources but may also amplify local errors. Principal Component Analysis (PCA) combines multiple related features into principal components through linear transformation. The contribution rate of each principal component (i.e., the proportion of variation in the original feature set that the principal component can explain) can be quantified. By retaining principal components whose cumulative contribution rate reaches a preset proportion, the core information of the original feature set can be preserved while significantly reducing the dimensionality of the original features. The elements of the dimensionality-reduced feature matrix are normalized to a preset numerical range using a standardization algorithm to obtain a standardized feature matrix. After obtaining the dimensionality-reduced feature matrix through the contribution rates of the principal components, although the dimensionality is reduced, the problem of inconsistent dimensions still exists. That is, the original features corresponding to different principal components have vastly different dimensions. Some principal components are associated with geometric features, such as nodal coordinates and beam length, with the dimension in meters (m) and a numerical range of 0-30m; Some principal components are associated with mechanical characteristics, such as stress distribution and load magnitude, with dimensions of megapascals (MPa) or kilonewtons (kN), and numerical ranges typically 20-50 MPa and 10-100 kN. Some principal components are related to material characteristics, such as concrete strength grade, which is a dimensionless grade, such as C30 and C40, and is 30 and 40 after numerical conversion; If features with large differences in dimensions are directly input into subsequent algorithms, such as fitness calculation in genetic algorithms or particle position update in particle swarm optimization, the features with large values ​​will dominate the optimization direction. For example, if the value of a beam length of 30m is much greater than the stress of 30MPa, the algorithm will prioritize adjusting around the beam length and ignore the core influencing factor of stress. The final reinforcement layout scheme may meet the geometric constraints but not the stress safety. Standardization algorithms can map feature values ​​of different dimensions to a unified preset range, such as Min-Max normalization and Z-Score standardization, so that all features have equal weight in the algorithm optimization and ensure that the optimization direction focuses on the arrangement of steel bars. If the variance of the standardized feature matrix is ​​lower than the preset variance threshold, the correlation between each feature and the reinforcement layout optimization objective is calculated by the feature selection algorithm, key features are extracted, and a simplified feature matrix is ​​obtained. The standardized feature matrix obtained after the first two steps may have the hidden danger of insufficient feature discrimination, that is, the variance of some features is too low. Such low variance features cannot provide effective discrimination information for subsequent algorithms. For example, if all beams have a concrete strength of C35, this feature cannot help the algorithm determine the differences in reinforcement between different beams. Instead, it increases the dimension of the feature matrix, causing the algorithm to lose focus and even introduce calculation errors due to small fluctuations in invalid features, thus reducing the optimization accuracy. The feature selection algorithm can select Pearson correlation coefficient or mutual information method. By calculating the correlation between each feature and the target of rebar layout optimization, only key features with high correlation and high discriminative power are retained, while invalid features with low variance and low correlation are eliminated. Based on dimensionality reduction, the features are further purified to ensure that the features input into subsequent algorithms are all core factors that have a real impact on rebar layout.

[0021] The scheme generation module generates an initial layout scheme set through a genetic algorithm and then transmits the initial layout scheme set to the scheme optimization module. The scheme optimization module uses the particle swarm optimization algorithm to locally optimize the initial layout scheme set to obtain a locally optimized scheme set. If there are stress concentration areas, the simulated annealing algorithm is used to adjust and generate a modified scheme set. The modified scheme set is then optimized using a consistency evaluation function and a differential evolution algorithm to obtain a globally coordinated scheme set, which is then transmitted to the scheme output module. The specific steps for local optimization of the initial layout scheme set using the particle swarm optimization algorithm are as follows: The local optimization objective is extracted from the initial layout scheme set, and the number of particles and the number of iterations of the particle swarm optimization algorithm are preset. The steel bar coordinates and diameter parameters of the layout scheme are iteratively updated to obtain the preliminary local optimization scheme set. Although the initial layout scheme set generated by the genetic algorithm meets global compliance, the local coordinates of the rebars may deviate from the node coordinates of the BIM model due to the randomness of the group coding, resulting in insufficient protective layer or collision with stirrups. Compared to the global search of genetic algorithms, PSO can quickly focus on the local parameter space through the collaborative mechanism of "particle memory individual optimality + group optimality", avoiding the waste of computing power in global iteration. Furthermore, the parameter iteration update of PSO can directly map the local parameters of the steel bars, adapting to the precise adjustment needs of local optimization objectives. Based on the uniformity of stress distribution, the initial local optimization scheme set is adjusted using a matrix transformation algorithm to generate an adjusted scheme set. If there are schemes in the adjustment scheme set where the stress distribution value is lower than the preset uniformity threshold, then reasonable parameters are inserted between the feature vectors using vector interpolation technology to correct the reinforcement arrangement position and generate a local optimization scheme set. Because matrix transformation is a batch scaling process, it may miss some special locations of reinforcement, such as irregular areas where beam ends meet columns, causing the stress in these areas to be lower than the uniformity threshold. In addition, some schemes may have blank local reinforcement parameters due to missing PSO particle coding, which need to be filled by interpolation to generate continuous parameters. Vector interpolation can insert new vectors that meet the constraints between adjacent reasonable feature vectors without overall adjustment, only correcting individual substandard areas, thereby improving stress uniformity and preserving the overall stability of the scheme. The specific steps for adjusting and generating the modified scheme set using the simulated annealing algorithm are as follows: If there is a stress concentration region in the local optimization scheme set where the stress value exceeds the preset stress threshold, then the feature vector set corresponding to the region is obtained through feature extraction technology. The vectors in the feature vector set include the rebar coordinates and stress value parameters of the stress concentration region. Stress concentration areas are usually small in size and may be adjacent to compliant areas, making them indistinguishable from global stress data alone. The root cause is often a mismatch between the steel reinforcement parameters and the stress in the area. Therefore, it is necessary to obtain both the stress value and the corresponding steel reinforcement parameters to make targeted adjustments. Feature extraction technology can establish a mapping relationship between stress-exceeding-threshold elements and corresponding steel reinforcement parameters based on the element stress data of finite element analysis and the geometric coordinates of the BIM model, thus binding the problem area with the corresponding parameters. The initial temperature and cooling rate of the simulated annealing algorithm are set, the position parameters of each vector in the feature vector set are iteratively adjusted, and the adjusted fitness value is calculated based on the stress distribution value and geometric constraints to obtain the adjusted feature vector set. Adjusting the reinforcement parameters in stress concentration areas can easily lead to local optima, for example: If stress is reduced simply by "reducing the spacing of the reinforcing bars", the spacing may be less than the minimum value specified in the code. Simply increasing the diameter may cause the reinforcing bars to collide with the stirrups; Simulated annealing algorithm simulates the process of "allowing disordered atomic movement at high temperatures and gradually ordering at low temperatures". At high temperatures, it allows for suboptimal solutions and avoids premature convergence to a local optimum where "stress is not up to standard but complies with regulations". As the temperature decreases, it gradually focuses on the optimal solution where "stress meets standards and geometry complies with regulations", and finally finds parameters that balance the two. If there are still vectors with stress values ​​exceeding the threshold in the adjusted feature vector set, a new vector is generated between adjacent reasonable vectors using vector interpolation technology to correct the rebar arrangement parameters and obtain the corrected feature vector set; an adjusted arrangement scheme set is generated based on the corrected feature vector set to obtain the corrected scheme set.

[0022] Simulated annealing adjusts the vectors independently, and parameter jumps may occur between the two. If the number of steel bars in the stress concentration area is small, the stress may still not be completely eliminated after simulated annealing due to the limited combination of parameters. By using vector interpolation technology, a "new vector" is inserted between the residual problem vector and the reasonable vector, and the stress gap is filled by parameter continuity, without restarting the large-scale iteration of simulated annealing. The steps to obtain the global coordination scheme set are as follows: The global consistency value of each scheme in the modified scheme set is calculated by using the global consistency evaluation function, and the consistency distribution feature set is obtained. The core value of the global consistency evaluation function lies in transforming qualitative requirements such as "cross-component collaboration, overall parameter matching, and global stress balance" into a single quantitative value, providing an objective basis for subsequent grouping and optimization. A multi-agent collaborative mechanism is used to group the vectors in the consistent distribution feature set according to feature similarity, resulting in a grouped feature set; The set of corrective solutions typically contains dozens to hundreds of solutions. If all solutions are optimized globally, it will lead to a waste of computing power and insufficient targeting. The multi-agent collaborative mechanism can improve efficiency while ensuring that solutions with similar features are grouped together through information sharing among agents. If there are vectors in the grouped feature set with a consistency value lower than the preset consistency threshold, then the mutation factor and crossover probability of the differential evolution algorithm are preset to adjust the vector parameters and obtain an optimized feature set. Low consistency vectors in a grouped feature set often suffer from global inconsistency due to contradictions in multi-parameter coupling. However, the differential evolution algorithm can adjust multiple coupling parameters simultaneously, such as beam diameter, column diameter, and node coordinates, through operations such as mutation and crossover, thus overcoming the limitations of adjusting a single parameter. The matrix transformation algorithm is used to perform coordinate mapping on the optimized feature set to generate a preliminary set of layout schemes. If there are schemes in the initial layout scheme set with a global consistency value lower than the threshold, then the feature vector parameters are recombined using vector recombination technology to adjust the global reinforcement layout and obtain a globally coordinated scheme set.

[0023] The scheme output module is used to generate a 3D visualization model and output layout scheme data; Extract the coordinates, dimensions, and spacing parameters of the reinforcing bars from the global coordination scheme set, and generate a parameter set containing these parameters; Geometric modeling of the parameter set is performed using a 3D visualization algorithm to construct a 3D relationship model between the reinforcing bars and the main structure, generating a preliminary visualization model. If there are elements in the initial visualization model whose coordinate deviation exceeds the preset deviation threshold, the element coordinates are corrected through a vector adjustment algorithm to generate an optimized visualization model. Arrangement data is extracted from the optimized visualization model, and the arrangement data is compressed using a data compression algorithm; The compressed layout data is converted into a BIM platform-compatible format using a format conversion algorithm to generate the final layout scheme data.

[0024] Example 2 This embodiment, based on Embodiment 1, further discloses a method for processing the final dataset using a three-dimensional geometric analytical algorithm to obtain a three-dimensional structural model, such as... Figure 2 As shown, the specific content includes: As described in Example 1, after obtaining the final dataset through the data acquisition module, the stereomicroscope algorithm is used to analyze the geometric data in the final dataset, identify the topological relationships between nodes, fill in the missing geometric information, and generate a preliminary three-dimensional structural model. The geometric data output by BIM models is mostly fragmented information, rather than a directly usable 3D model: These geometric data are scattered in form, containing only discrete node coordinates (such as the endpoint coordinates of beams and the rotation coordinates of columns) and cross-sectional parameters (such as beam width and column diameter), without specifying the topological relationships between nodes; In actual engineering projects, BIM data may contain missing geometric information due to modeling errors or omissions (such as the coordinates of the intermediate nodes of a secondary beam not being marked). If the model is generated solely by connecting nodes, it will result in model breakage or topological errors, failing to meet the accuracy requirements of subsequent stress analysis. The core capability of the stereomicroscope algorithm is to perform three-dimensional reconstruction based on discrete data, which is perfectly suited to the characteristics of the BIM geometric data mentioned above. It can identify the topological logic between nodes and fill in missing information to generate a continuous and complete preliminary model. If the geometric accuracy of the initial 3D structural model is lower than the preset accuracy threshold, the cross-sectional dimensions and node position parameters of the model are adjusted through a geometric optimization algorithm to obtain an optimized 3D structural model. While the preliminary 3D model meets the continuity requirements, its geometric accuracy may be insufficient due to the following reasons: Minor errors during BIM modeling (such as node coordinate deviation of 0.05m or cross-section dimension annotation deviation of 0.02m) will be retained in the initial model. The geometric parameters of the preliminary model (such as beam height and column cross-section dimensions) may only meet the requirements of appearance modeling, without considering the accuracy threshold of subsequent stress analysis; The system needs to set a geometric accuracy threshold to ensure that the stress calculation error in subsequent finite element analysis is within an acceptable range, while the preliminary model may not meet this threshold; The core of the geometric optimization algorithm is to adjust the geometric parameters in reverse based on the accuracy target, and to make the model accuracy meet the target by iteratively correcting the cross-sectional dimensions and node positions. For example, the initial model parameters of a simply supported beam are: beam length 6m, cross-section width 0.3m, and height 0.5m. The preset geometric accuracy thresholds are: cross-section size deviation ≤3%, node position deviation ≤0.02m. Upon inspection, the node position deviation is: the actual coordinates of the right-end node are (6.03,0,0), a deviation of 0.03m (exceeding the threshold by 0.02m). The cross-section height deviation is: the design specification requires a beam height of 0.51m, but the initial model is only 0.5m, a deviation of 1.96% (although not exceeding the threshold, subsequent stress calculations will result in a 5% higher stress value due to insufficient height). The geometric optimization algorithm will perform the following operations: correct the right-end node coordinates to (6.00,0,0) to eliminate the position deviation; adjust the beam height to 0.51m to match the design specification. The optimized 3D structural model is standardized using a coordinate mapping algorithm to generate a standardized 3D structural model. In architectural engineering, the coordinate system of a BIM model may not be uniform depending on the modeling scenario: BIM modeling may use a construction coordinate system, a design coordinate system, or a geodetic coordinate system. The system's feature processing module and scheme optimization module must both be based on a unified coordinate system; otherwise, geometric misalignment will occur. The core of the coordinate mapping algorithm is to convert the optimized model coordinates into a unified standard coordinate system, ensuring that the model can be seamlessly integrated with subsequent modules.

[0025] Example 3 This embodiment, based on Embodiments 1 and 2, further discloses a method for generating an initial arrangement scheme set using a genetic algorithm, such as... Figure 3 As shown, it specifically includes the following: Based on the dimensionality-reduced feature matrix, a genetic algorithm is used to initialize a multi-agent population, with each agent corresponding to a steel reinforcement layout scheme. The reinforcement layout of complex structures needs to meet the three objectives of uniform stress, geometric compliance, and economical usage. However, manual methods can only generate 1-2 initial schemes based on experience, which cannot cover the possible combinations of different reinforcement diameters, spacings, and layout directions. Subsequent optimization is prone to getting stuck in local optima due to the single initial point. The population iterative characteristic of genetic algorithms can be used to associate dimensionality reduction features with solution parameters through agent encoding, thereby achieving a direct transformation from features to solutions. For example, the dimensionality reduction feature matrix of a certain office building frame beam contains four core features, including mid-span stress of 32MPa, beam length of 8m, support shear stress of 18MPa, and concrete strength of C35. The preset size of the intelligent agent population is 50, that is, 50 initial schemes are generated. The genetic algorithm uses chromosome encoding, where the first two bits of the chromosome represent the diameter of the steel bar: 01=Φ20, 10=Φ22, the middle three bits represent the spacing: 001=25mm, 010=28mm, 100=30mm, and the last four bits represent the arrangement direction: 0001=parallel to beam length, 0010=45° diagonal. Among the 50 generated schemes, 20 schemes use Φ20 steel bars (suitable for medium stress) with a spacing of 25-30mm, and 30 schemes use Φ22 steel bars (suitable for higher mid-span stress) with a spacing of 25-28mm. Set a fitness function and iteratively optimize the rebar layout scheme through crossover and mutation operations to generate an optimized scheme set; The initial schemes generated in the first step only satisfy the basic feature constraints and have not yet taken into account the balance of multiple objectives. It is necessary to improve the basic quality through iterative optimization. Therefore, setting a fitness function can transform stress uniformity, geometric compliance, and quantity economy into a single quantitative index (i.e., fitness value) to judge the quality of the scheme. After multiple iterations, the excellent initial solution often gets stuck in a local optimum. At this point, the advantages of both can be combined through crossover operations. At the same time, in order to avoid the group from becoming homogeneous, mutation operations are needed to randomly adjust some parameters in order to explore new and better combinations. The set of optimal solutions is sorted in descending order by using a sorting algorithm based on the fitness values ​​calculated by the fitness function, resulting in a sorted set of solutions. The set of optimized solutions generated after crossover and mutation is still in an unordered state. If it is not sorted, subsequent algorithms may randomly select solutions for optimization, resulting in a waste of computing power. If there are schemes in the sorting scheme set with fitness values ​​lower than the preset fitness threshold, the feature vector parameters corresponding to the schemes are corrected by the matrix adjustment algorithm to generate an initial arrangement scheme set. Solutions that fall below the fitness threshold after sorting are not entirely ineffective. They are often caused by a deviation of a certain parameter from the constraint. If they are discarded directly, in addition to wasting the computing power in the early stage, it will also reduce the coverage of solutions, making it easy for subsequent optimizations to fall into local optima again. Since the core problem of substandard solutions is the mismatch between eigenvector parameters and constraints, an algorithm is needed to adjust the eigenvectors in reverse so that the solution meets the constraints again. The matrix adjustment algorithm can correct the mapping from the eigenvectors to the updated solution parameters, which precisely meets the above requirements.

[0026] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0027] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention 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 methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A smart reinforcement layout system for complex structures integrating BIM and MARL, characterized in that, include: The data acquisition module is used to extract geometric and stress data from the building information model to obtain the final dataset, and then transmit the final dataset to the 3D modeling module. The 3D modeling module uses 3D geometric analysis algorithms to process the final dataset, obtain a 3D structural model, and then transmits the 3D structural model to the feature processing module. The feature processing module uses the finite element analysis algorithm to calculate the stress distribution and generate a feature set. If the dimension of the feature set exceeds a preset threshold, the principal component analysis algorithm is used to reduce the dimension to obtain a dimension-reduced feature matrix, and the dimension-reduced feature matrix is ​​transmitted to the scheme generation module. The scheme generation module generates an initial layout scheme set through a genetic algorithm and then transmits the initial layout scheme set to the scheme optimization module. The scheme optimization module uses the particle swarm optimization algorithm to locally optimize the initial layout scheme set to obtain a locally optimized scheme set. If there are stress concentration areas, the simulated annealing algorithm is used to adjust and generate a modified scheme set. The modified scheme set is then optimized using a consistency evaluation function and a differential evolution algorithm to obtain a globally coordinated scheme set, which is then transmitted to the scheme output module. The scheme output module is used to generate a 3D visualization model and output layout scheme data.

2. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The specific steps to obtain the final dataset are as follows: Extract node coordinates and boundary conditions from the building information model to generate an initial dataset containing node coordinates and boundary conditions; The initial dataset is preprocessed using a data integration algorithm to obtain a standardized initial dataset; A lossless data compression algorithm is used to compress the standardized initial dataset to generate a compressed initial dataset; The compressed initial dataset is converted into a standard output format using a format conversion algorithm to obtain the final dataset.

3. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The steps to process the final dataset using a 3D geometric analytical algorithm to obtain the 3D structural model are as follows: The stereomicroscopy algorithm is used to analyze the geometric data in the final dataset, identify the topological relationships between nodes, fill in the missing geometric information, and generate a preliminary three-dimensional structural model. If the geometric accuracy of the initial 3D structural model is lower than the preset accuracy threshold, the cross-sectional dimensions and node position parameters of the model are adjusted through a geometric optimization algorithm to obtain an optimized 3D structural model. The coordinate mapping algorithm is used to standardize the optimized 3D structural model, generating a standardized 3D structural model.

4. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The steps for calculating stress distribution and generating a feature set using the finite element analysis algorithm include: Based on the node coordinates and boundary conditions in the final dataset, the three-dimensional structural model is divided into finite element elements, and the stress distribution of each element is calculated using the finite element analysis algorithm to obtain stress distribution data. If the maximum value in the stress distribution data exceeds the preset stress threshold, the parameters of the three-dimensional structural model of the corresponding region are adjusted by the local geometric optimization algorithm, the stress distribution is recalculated, and optimized stress distribution data is generated. The data integration algorithm combines node coordinates, boundary conditions, and optimized stress distribution data according to a preset structure to form a feature set.

5. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The specific steps for obtaining the dimensionality-reduced feature matrix using principal component analysis are as follows: Calculate the contribution rate of each principal component, retain the principal components whose cumulative contribution rate reaches a preset proportion, and obtain the dimensionality-reduced feature matrix; The elements of the dimensionality-reduced feature matrix are normalized to a preset numerical range using a standardization algorithm to obtain a standardized feature matrix. If the variance of the standardized feature matrix is ​​lower than the preset variance threshold, the correlation between each feature and the reinforcement layout optimization objective is calculated using a feature selection algorithm, key features are extracted, and a simplified feature matrix is ​​obtained.

6. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The steps for generating an initial set of arrangement schemes using a genetic algorithm are as follows: Based on the dimensionality-reduced feature matrix, a genetic algorithm is used to initialize a multi-agent population, with each agent corresponding to a steel reinforcement layout scheme. Set a fitness function and iteratively optimize the rebar layout scheme through crossover and mutation operations to generate an optimized scheme set; The set of optimal solutions is sorted in descending order by using a sorting algorithm based on the fitness values ​​calculated by the fitness function, resulting in a sorted set of solutions. If there are schemes in the sorting scheme set with fitness values ​​lower than the preset fitness threshold, the matrix adjustment algorithm is used to correct the feature vector parameters corresponding to the schemes to generate an initial arrangement scheme set.

7. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The specific steps for local optimization of the initial layout scheme set using the particle swarm optimization algorithm are as follows: The local optimization objective is extracted from the initial layout scheme set, and the number of particles and the number of iterations of the particle swarm optimization algorithm are preset. The steel bar coordinates and diameter parameters of the layout scheme are iteratively updated to obtain the preliminary local optimization scheme set. Based on the uniformity of stress distribution, the initial local optimization scheme set is adjusted using a matrix transformation algorithm to generate an adjusted scheme set. If there are schemes in the adjustment scheme set where the stress distribution value is lower than the preset uniformity threshold, then reasonable parameters are inserted between the feature vectors using vector interpolation technology to correct the reinforcement arrangement position and generate a local optimization scheme set.

8. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The specific steps for adjusting and generating the modified scheme set using the simulated annealing algorithm are as follows: If there is a stress concentration region in the local optimization scheme set where the stress value exceeds the preset stress threshold, then the feature vector set corresponding to the region is obtained through feature extraction technology. The vectors in the feature vector set include the rebar coordinates and stress value parameters of the stress concentration region. The initial temperature and cooling rate of the simulated annealing algorithm are set, the position parameters of each vector in the feature vector set are iteratively adjusted, and the adjusted fitness value is calculated based on the stress distribution value and geometric constraints to obtain the adjusted feature vector set. If there are still vectors with stress values ​​exceeding the threshold in the adjusted feature vector set, then a new vector is generated between adjacent reasonable vectors using vector interpolation technology to correct the rebar arrangement parameters and obtain the corrected feature vector set. An adjusted set of layout schemes is generated based on the modified feature vector set, thus obtaining the modified scheme set.

9. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The steps to obtain the global coordination scheme set are as follows: The global consistency value of each scheme in the modified scheme set is calculated by using the global consistency evaluation function, and the consistency distribution feature set is obtained. A multi-agent collaborative mechanism is used to group the vectors in the consistent distribution feature set according to feature similarity, resulting in a grouped feature set; If there are vectors in the grouped feature set with a consistency value lower than the preset consistency threshold, then the mutation factor and crossover probability of the differential evolution algorithm are preset to adjust the vector parameters and obtain an optimized feature set. The matrix transformation algorithm is used to perform coordinate mapping on the optimized feature set to generate a preliminary set of layout schemes. If there are schemes in the initial layout scheme set with a global consistency value lower than the threshold, then the feature vector parameters are recombined using vector recombination technology to adjust the global reinforcement layout and obtain a globally coordinated scheme set.

10. The intelligent reinforcement layout system for complex structures integrating BIM and MARL as described in claim 1, characterized in that, The specific steps for generating a 3D visualization model and outputting layout data are as follows: Extract the coordinates, dimensions, and spacing parameters of the reinforcing bars from the global coordination scheme set, and generate a parameter set containing these parameters; Geometric modeling of the parameter set is performed using a 3D visualization algorithm to construct a 3D relationship model between the reinforcing bars and the main structure, generating a preliminary visualization model. If there are elements in the initial visualization model whose coordinate deviation exceeds the preset deviation threshold, the element coordinates are corrected through a vector adjustment algorithm to generate an optimized visualization model. Arrangement data is extracted from the optimized visualization model, and the arrangement data is compressed using a data compression algorithm; The compressed layout data is converted into a BIM platform-compatible format using a format conversion algorithm to generate the final layout scheme data.

Images