A structural design and optimization method and system suitable for a multi-working-condition air building machine

Abstract

This invention relates to the field of structured design technology, and more specifically to a method and system for structural design and optimization of a multi-condition aerial building machine. The method includes: parameterizing the aerial building machine structure to obtain structural parameters; training an ensemble deep learning model using the structural parameters data based on a snapshot strategy, obtaining M neural network models after M training cycles, and calculating the average output of the M neural network models as the prediction result; based on the prediction result, using the NSGA-II multi-objective optimization algorithm, proposing a stochastic optimization method based on LHS, and simulating the uncertainty of external loads by repeatedly sampling the probability density distribution of external loads using the LHS method to obtain a steel platform structure for the aerial building machine adaptable to multiple operating conditions. This invention enables the rapid generation and flexible selection of structural design schemes for building machines under various operating conditions.

Description

Technical Field

[0001] This invention relates to the field of structured design technology, and more specifically to a structural design and optimization method and system for aerial building machines adapted to multiple working conditions. Background Technology

[0002] With the rapid urbanization of the past two decades, high-rise buildings have gained widespread popularity and use worldwide due to their aesthetic design, comfortable environment, and land-saving advantages. However, high-rise building construction presents a series of challenges, including high difficulty, high risk, and complex organization, urgently requiring advanced construction equipment to improve the on-site working environment and enhance construction efficiency. Based on these needs, aerial building construction machines (i.e., integrated construction equipment platforms) have played a significant role in the construction of numerous landmark buildings in China, thanks to their high degree of equipment integration, factory-like operation, fast construction speed, and guaranteed construction safety. This has greatly improved the mechanization and intelligence level of high-rise building construction, propelling my country's high-rise building construction technology to new heights. In the design process of the building construction machine, the power system is mainly provided by specialized manufacturers, while the support system, formwork system, and hanging system can be directly assembled on-site. However, the steel platform (truss system) requires designers to design specifically according to the actual site conditions. The steel platform of the building construction machine is mainly used for suspending formwork and hanging frames, and also serves as a storage area for upper construction machinery and materials. Therefore, the steel platform must possess sufficient rigidity and strength, making its design a major problem that technical personnel need to solve.

[0003] Traditional design methods for steel platforms used in building construction primarily rely on modular design rules. This involves dividing the structural plane into a grid and designing dimensions according to basic modules to achieve modular configuration of the truss system. However, traditional methods suffer from low efficiency, reliance on expert experience, and a conservative design tendency. Therefore, it is necessary to optimize the design process of steel platforms for building construction by combining advanced design methods and intelligent algorithms. A key issue is how to integrate advanced design methods and intelligent optimization algorithms with the design of steel platforms for building construction to achieve rule-based design, rapid iteration, and flexible selection of solutions.

[0004] Therefore, this invention proposes a structural design and optimization method and system for aerial building machines adapted to multiple working conditions, which can be used to realize the rapid generation and flexible selection of structural design schemes for building machines under various working conditions. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a structural design and optimization method and system for aerial building machines adapted to multiple operating conditions, enabling rapid generation and flexible selection of structural design schemes for building machines under various operating conditions.

[0006] In a first aspect, the present invention provides a structural design and optimization method for an aerial building machine adapted to multiple operating conditions, the method comprising the following steps:

[0007] Step S1: Parametric processing of the aerial building machine structure is performed in Grasshopper to obtain the parameters of the aerial building machine structure after parametric processing.

[0008] Step S2: Based on the snapshot strategy, train an ensemble deep learning model using the structural parameters of the aerial building machine. The ensemble deep learning model uses a convolutional neural network model as the base model. After M training cycles, M neural network models are obtained, and the average output of the M neural network models is used as the prediction result of the ensemble deep learning model.

[0009] Step S3: Based on the prediction results, the NSGA-II multi-objective optimization algorithm is used to propose a stochastic optimization method based on LHS. The constraints in the optimization problem are replaced with probabilistic constraints. By using the LHS method to perform multiple samplings in the probability density distribution of the external load, the uncertainty of the external load is simulated, and a steel platform structure for an aerial building machine that is adaptable to multiple working conditions is obtained.

[0010] As a preferred embodiment of the present invention, the parameterization process of the aerial building machine in step S1 includes the following steps:

[0011] Step S11: First, determine the position of the outer truss of the steel platform based on the position of the outer shear wall of the core tube. Second, determine the movement range of the supporting steel column of the steel platform based on the position of the inner shear wall of the core tube. Then, determine the support point of the outer steel column of the steel platform based on the truss cantilever length requirement. All other parameters except the above parameters are parameterized. Parameterization means using a certain method to replace fixed values ​​with parameters.

[0012] Step S12: According to the module design rules, the moving step of the supporting steel column is greater than or equal to the first module. By setting judgment statements through parametric programming, the corresponding type of Bailey panels are automatically generated on the upper part according to the moving distance of the supporting steel column, thereby forming a steel platform. Transverse supports are set between the outer truss and the main truss of the steel platform, and the number of struts is parameterized to complete the parametric modeling of the aerial building machine. The module refers to the selected standard scale unit of measurement.

[0013] As a preferred technical solution of the present invention, after completing the parametric modeling of the aerial building machine, the method further includes: converting the geometric model of the aerial building machine into a physical model by setting material properties, cross-section allocation, external load, boundary constraints and spatial mesh, realizing the data flow transfer between the geometric model and the physical model, and finally calculating and outputting structural response data, wherein the structural response refers to the stress, displacement, deformation and buckling caused by the structure after being subjected to external forces.

[0014] As a preferred technical solution of the present invention, the convolutional neural network (1D-CNN) model in step S2 is mainly composed of a convolutional feature extraction layer and a fully connected layer. The convolutional feature extraction layer is composed of a convolutional layer and a pooling layer, which is used to extract data features. The fully connected layer is used for regression output of feature data.

[0015] As a preferred embodiment of the present invention, the snapshot strategy employs a cyclic cosine annealing learning rate during the training of the ensemble deep learning model, causing the ensemble deep learning model to converge to multiple local minima. When the learning rate cycle restarts, the performance of the ensemble deep learning model temporarily decreases. After the learning rate annealing, the performance eventually surpasses that of the previous cycle. The training process is divided into M cycles, each starting with a high learning rate and gradually decreasing to a lower learning rate. The initial learning rate provides sufficient energy for the ensemble deep learning model to move away from the current gradient point, while the lowest learning rate guides the ensemble deep learning model to a well-performing local minimum. At the end of each training cycle, the ensemble deep learning model reaches a local minimum in terms of training loss. Before increasing the learning rate, the weights and biases of the ensemble deep learning model are saved once, where M is a positive integer greater than or equal to 1.

[0016] As a preferred technical solution of the present invention, the constraint condition in step S3 refers to the structural reliability under different working conditions. The structural reliability under different working conditions is sampled and calculated by the corresponding load distribution. The different working conditions include construction working conditions and abnormal working conditions.

[0017] As a preferred technical solution of the present invention, in step S3, after obtaining k samples from the probability density distribution of each variable during the optimization process, the output target value of the k samples is calculated by the optimization algorithm, where n represents the number of results that meet the constraints, and the probability is determined as p(x) = n / k.

[0018] As a preferred technical solution of the present invention, a scenario-based probabilistic method is adopted, assuming that the external loads follow different probability distributions under different working conditions, and the structural reliability under different working conditions is set as a probabilistic constraint in the optimization problem, thereby obtaining a steel platform structure for an aerial building machine that adapts to multiple working conditions.

[0019] Secondly, the present invention provides a structural design and optimization system for an aerial building machine adapted to multiple operating conditions, the system comprising:

[0020] The parametric design module is used to parametrically process the structure of the aerial building machine. The parametric processing is performed in Grasshopper to obtain the parameters of the aerial building machine structure after parametric processing.

[0021] The AI ​​agent model module is used to train an integrated deep learning model based on the data of the structural parameters of the aerial building machine, using the snapshot strategy. The integrated deep learning model uses a convolutional neural network model as the base model, and after M training cycles, it obtains M neural network models. The average output of the M neural network models is used as the prediction result of the integrated deep learning model.

[0022] The objective optimization module is used to propose a stochastic optimization method based on LHS using the NSGA-II multi-objective optimization algorithm based on the prediction results. The constraints in the optimization problem are replaced with probabilistic constraints. By using the LHS method to perform multiple samplings in the probability density distribution of the external load, the uncertainty of the external load is simulated, and a steel platform structure for an aerial building machine that can adapt to multiple working conditions is obtained.

[0023] Thirdly, the present invention also provides a storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described structural design and optimization method for an aerial building machine adapted to multiple operating conditions.

[0024] Compared with the prior art, the beneficial effects of the present invention are at least as follows:

[0025] 1. The technical solution of the present invention combines parametric design method and multi-objective optimization algorithm, and takes into account the uncertainty factors from the environment, to realize the rapid generation and flexible selection of building machine structural design schemes under various working conditions.

[0026] 2. The technical solution of this invention parameterizes the building machine structure, proposes an integrated deep learning model based on the snapshot strategy, and uses the NSGA-II multi-objective optimization algorithm considering uncertainty to quickly generate a steel platform structure for the building machine that meets usage requirements, achieving a structural reliability of 0.9999. Compared with 100 test cases generated using Optimal Space Filling (OSF), the optimal design scheme obtained by this invention achieves a comprehensive improvement rate of 16.59% under construction conditions and 16.91% under abnormal conditions. The results show that the LHS-based stochastic optimization method and integrated deep learning model used in this invention have good accuracy and computational efficiency. This invention effectively improves the visualization, interactivity, and intelligence level of the aerial building machine structure design process. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0028] Figure 1 This is a flowchart illustrating the steps of a structural design and optimization method for an aerial building machine adapted to multiple working conditions according to the present invention.

[0029] Figure 2 This is a flowchart illustrating the design process of the aerial building machine in this invention.

[0030] Figure 3 shows six different layout configurations of the main truss of the steel platform in this invention;

[0031] Figure 4 This is a diagram showing the lateral support arrangement area of ​​the steel platform in this invention;

[0032] Figure 5 This is a physical model diagram of the aerial building machine in this invention;

[0033] Figure 6 is a diagram showing the design results of this invention;

[0034] Figure 7 This is a structural diagram of the structural design and optimization system for an aerial building machine adapted to multiple working conditions, as described in this invention.

[0035] Figure 7 As shown: 100, Structural design and optimization system for multi-condition aerial building machine; 101, Parametric module; 102, Prediction module; 103, Target optimization module. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0037] It is understood that the terms "first," "second," etc., used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are used only to distinguish one element from another. For example, without departing from the scope of this application, a first script may be referred to as a second script, and similarly, a second script may be referred to as a first script.

[0038] Traditional steel platform design methods for building construction machines are mainly based on modular design rules. By dividing the structural plane into grids and designing dimensions according to basic modules, modular configuration of the truss system can be achieved. This traditional steel platform design method for building construction machines has problems such as low work efficiency, reliance on expert experience, and conservative design.

[0039] To address the aforementioned technical problems, the inventors proposed the following: Figure 1 The structural design and optimization method for an aerial building machine adapted to multiple operating conditions, as shown, includes the following steps:

[0040] Step S1: Parametrically process the aerial building machine structure. The parametric processing is performed in Grasshopper to obtain the parametrically processed aerial building machine structure parameters.

[0041] Step S2: Based on the snapshot strategy, use the data of the structural parameters of the aerial building machine to train an ensemble deep learning model. The ensemble deep learning model uses a convolutional neural network model as the base model. After M training cycles, M neural network models are obtained, and the average output of the M neural network models is used as the prediction result of the ensemble deep learning model.

[0042] Step S3: Based on the prediction results, the NSGA-II multi-objective optimization algorithm is used to propose a stochastic optimization method based on LHS. The constraints in the optimization problem are replaced with probabilistic constraints. By using the LHS method to perform multiple samplings in the probability density distribution of external loads, the uncertainty of external loads is simulated, and a steel platform structure for aerial building construction machines that can adapt to multiple working conditions is obtained.

[0043] Specifically, this embodiment provides a structural design and optimization method for an aerial building machine adapted to multiple working conditions, such as... Figure 2This example demonstrates the application of a lightweight building-building machine in the construction of a 400m high-rise building. The structure of the aerial building-building machine is parametrically processed to obtain its structural parameters. This parametric process is performed in Grasshopper, a plugin for generating models using algorithms within the Rhino environment, and is one of the mainstream software programs for data-driven design. The more specific parametric design of the aerial building-building machine structure is as follows: First, based on the location of the external shear walls of the core tube, the location of the outer truss can be determined. Second, based on the location of the internal shear walls of the core tube, the movement range of the supporting steel columns is determined. Then, according to the truss cantilever length requirements, the support points of the outer steel columns of the steel platform can be determined. Except for these parameters, all other parameters can be parametrically configured. In this example, the overall layout x1 of the main truss of the steel platform is set to 6 types, as shown in Figure 3. The transverse supports x2, x3, x4, and x5 of the steel platform are set in 4 areas, with the number of supports in each area set to 0-3, as shown in Figure 3. Figure 4 As shown. The upper and lower chords of the steel platform Bailey bridge use double-groove sections, while the web members use box sections. The cross-sectional dimensions vary as follows: x... 11 ,x 12 ,x 13 ,x 14 ,x 15 ,x 16 ,x 17 The building machine mainly bears loads from slabs, scaffolding, formwork, and wind. To assess the impact of external load uncertainties on the structural response, it is also necessary to consider loads x6, x7, x8, x9, and x... 10 As variables used in training the surrogate model. When encountering extreme wind loads, the building machine will stop construction, remove part of the platform load, and apply additional constraints to the supporting columns, such as... Figure 5 As shown. Since different constraints have a significant impact on the working state of the building machine, it is necessary to treat the constraints as variables x. 18 Based on the actual conditions of the construction site, the descriptions of the design variables and external loads, as well as the range of variable values, are shown in Table 1. Furthermore, the structural stress ratio y1, displacement y2, and steel consumption y3 are used as three objective functions to minimize the design in the optimization problem. The stress ratio y1 is used to evaluate the safety capacity of the structural materials from a strength perspective, the maximum displacement y2 is used to evaluate the deformation and stiffness of the entire structure, and the steel consumption y3 can be used to evaluate the cost-effectiveness of the structural design.

[0044] Table 1. Variable Value Table

[0045]

[0046] The optimal space filling (OSF) method based on clustering was used to generate 3000 samples according to the variables shown in Table 1. These samples were then imported into geometric design software and finite element analysis software to calculate stress ratios, displacements, and steel consumption, ultimately yielding samples for model training and testing. In this embodiment, an ensemble deep learning model using a snapshot strategy was employed to predict structural responses. This model network has seven layers, including three one-dimensional convolutional layers for advanced feature extraction and four fully connected layers for regression. Detailed information about the ensemble deep learning model is shown in Table 2. OSF is a clustering algorithm for high-dimensional datasets. It is a grid-based method that divides the data space into a series of small cubes and then fills these cubes to find the optimal clustering result. The advantage of the OSF algorithm is that it is an adaptive clustering method, capable of handling high-dimensional datasets, and can automatically determine the number of clusters.

[0047] Table 2. Parameter information of ensemble deep learning models

[0048]

[0049] In the field of structural reliability research, common loads follow a certain probability distribution. Therefore, platform loads and wind loads can be regarded as live loads, while scaffold loads and formwork loads can be regarded as dead loads. The specific distribution and parameter values ​​are shown in Table 4. The goal of the optimization problem in this invention is to minimize the structural stress ratio, displacement, and steel consumption. The structural reliability under different working conditions is regarded as a constraint. Through corresponding load distribution sampling and calculation, the structural stress reliability and displacement reliability under construction conditions, and the structural stress reliability and displacement reliability under high wind conditions are obtained.

[0050] The method of this invention selects the optimal design scheme from the Pareto front (finding the optimal solution in a multi-objective optimization problem), as shown in Figure 6, and its specific design parameters are shown in Table 5. After 10,000 Monte Carlo tests, the structural reliability can reach as low as 0.9999. The frequency distribution and cumulative distribution indicate that the optimal design scheme obtained by the generative design method of this invention can meet the reliability requirements under various working conditions. Through finite element simulation, it can be seen from the structural stress and displacement cloud diagrams that the optimal scheme has a reasonable structural design, and all indicators are within the safe range. By comparing with 100 test samples generated by the optimal space-filling design (OSF), the overall improvement rate of the three objective values ​​of the optimal design scheme under construction conditions is 16.59%, and the overall improvement rate under extreme wind conditions is 16.91%. The optimization rate of each objective value is shown in Table 6. The results show that the optimal design scheme has a significant optimization improvement effect compared with the test samples. The stochastic optimization method of LHS (Latin Hypercube Sampling) is a method for solving optimization problems. It obtains a set of representative sample points by randomly sampling in the solution space, and then performs optimization on this basis.

[0051] Among them, LHS (Latin Hypercube Sampling) is an efficient random sampling technique that ensures that sample points are uniformly distributed in the solution space, thereby reducing overlap and omission of sample points and improving the efficiency of random sampling. In LHS-based stochastic optimization methods, a representative set of sample points is first generated using LHS sampling, and then an optimization algorithm (such as gradient descent, genetic algorithm, etc.) is used to optimize the sample points to find the optimal solution. Additionally, parameterization refers to using a method to replace fixed values ​​with parameters to expand the scope of basic tests or components, thereby improving the functionality and flexibility of the tests or components.

[0052] Furthermore, in step S1 above, the parameterization process of the aerial building machine includes the following steps:

[0053] Step S11: First, determine the position of the outer truss of the steel platform based on the position of the outer shear wall of the core tube. Second, determine the movement range of the steel platform support steel column based on the position of the inner shear wall of the core tube. Then, determine the support point of the outer steel column of the steel platform based on the truss cantilever length requirement. All other parameters except the above parameters are parameterized. Parameterization means using a certain method to replace fixed values ​​with parameters.

[0054] Step S12: According to the module design rules, the movement step of the supporting steel column is greater than or equal to the first module. By setting judgment statements through parametric programming, the corresponding type of Bailey panels are automatically generated on the upper part according to the movement distance of the supporting steel column, thus forming a steel platform. Transverse supports are set between the outer truss and the main truss of the steel platform, and the number of struts is parameterized to complete the parametric modeling of the aerial building machine. The module refers to the selected standard scale unit of measurement.

[0055] Specifically, in the above embodiments, the parameterization process is performed in Grasshopper. After determining the values ​​of each parameter, judgment statements are set through parametric programming. Based on the moving distance of the supporting steel columns, the corresponding type of Bailey panels are automatically generated on the upper part to form a steel platform. Finally, transverse supports are set between the outer truss and the main truss of the steel platform, and the number of struts is parameterized to ensure that the truss system has appropriate out-of-plane stiffness. The first module is the standard dimensional unit selected for the building and its components (or assemblies), and serves as the increment unit in dimensional coordination, becoming the building module unit.

[0056] Furthermore, after completing the parametric modeling of the aerial building machine, the process also includes: converting the geometric model of the aerial building machine into a physical model by setting material properties, section allocation, external loads, boundary constraints, and spatial meshes, realizing data flow transfer between the geometric model and the physical model, and finally calculating and outputting structural response data. Structural response refers to the stress, displacement, deformation, and buckling caused by external forces acting on the structure.

[0057] Specifically, the geometric information of the parametric model is transferred from Grasshopper to the structural finite element analysis software YJK via the GAMA plugin. The material properties, section assignments, external loads, and boundary constraints of the geometric parametric model can be set visually in GAMA, and the physical information is mapped to YJK in real time to calculate and output structural response data. To improve the prediction accuracy of the surrogate model, a large number of physical data samples are needed from YJK. Through secondary development using the GH_CPython plugin, Grasshopper can automatically read different design parameters and transmit the design data stream to GAMA. Simultaneously, through the custom card function in GAMA, large-scale batch calculations jointly performed by Grasshopper, GAMA, and YJK are achieved. YJK-GAMA is a digital intelligent design software for the construction industry that integrates visual programming, parametric design, and computer-aided optimization.

[0058] Furthermore, in step S2 above, the convolutional neural network (1D-CNN) model mainly consists of a convolutional feature extraction layer and a fully connected layer. The convolutional feature extraction layer consists of a convolutional layer and a pooling layer, which is used to extract data features, while the fully connected layer is used for regression output of the feature data.

[0059] Specifically, the ensemble deep learning model network has 7 layers, including three one-dimensional convolutional layers for high-level feature extraction and four fully connected layers for regression. Detailed information about the ensemble deep learning model is shown in Table 2. Specifically, the total number of training epochs is 200, and the epoch duration M is 5, meaning that the ensemble deep learning model has 5 base models, each trained for 40 epochs. Furthermore, the training, validation, and test data are divided in a 6:2:2 ratio. As shown in Table 3, the test results show that the ensemble deep learning model has high prediction accuracy, particularly for structural stress ratio, displacement, and R0 of steel consumption. 2 The results, reaching 0.931, 0.981, and 0.998, indicate that the ensemble deep learning model used in this invention can effectively reflect the characteristics of structural response and can be used for predicting the response of complex structures. Here, epochs represents the number of training iterations of the model.

[0060] Table 3. Model Performance Evaluation

[0061]

[0062]

[0063] Furthermore, the aforementioned snapshot strategy employs a cyclic cosine annealing learning rate during the training of the ensemble deep learning model, causing the ensemble deep learning model to converge to multiple local minima. When the learning rate cycle restarts, the performance of the ensemble deep learning model temporarily decreases. After learning rate annealing, the performance eventually surpasses that of the previous cycle. The training process is divided into M cycles, each starting with a high learning rate and gradually decreasing to a lower learning rate. The initial learning rate provides the ensemble deep learning model with sufficient energy to move away from the current gradient point, while the lowest learning rate guides the ensemble deep learning model to a well-performing local minimum. At the end of each training cycle, the ensemble deep learning model reaches a local minimum in terms of training loss. Before increasing the learning rate, the weights and biases of the ensemble deep learning model are saved once, where M is a positive integer greater than or equal to 1.

[0064] Specifically, by employing an ensemble learning strategy, the model can avoid getting stuck in local optima, thereby reducing model variance and improving prediction accuracy.

[0065] Furthermore, the constraints in step S3 above refer to the structural reliability under different working conditions. The structural reliability under different working conditions is sampled and calculated through the corresponding load distribution. Different working conditions include construction working conditions and abnormal working conditions.

[0066] Specifically, the goal of the optimization problem in this embodiment is to minimize the structural stress ratio, displacement, and steel consumption. Structural reliability under different working conditions is considered a constraint, and sampling and calculation are performed using corresponding load distributions. Abnormal working conditions include situations such as strong winds. In the field of structural reliability research, common loads follow a certain probability distribution. Therefore, platform loads and wind loads can be considered as live loads, while scaffold loads and formwork loads can be considered as dead loads. The specific distribution and parameter values ​​are shown in Table 4.

[0067] Table 4. Load probability distribution and parameters (Unit: N / mm)

[0068]

[0069]

[0070] Furthermore, in step S3 above, after obtaining k samples from the probability density distribution of each variable during the optimization process, the output target value of the k samples is calculated through the optimization algorithm, where n represents the number of results that satisfy the constraints. Therefore, the probability is determined as p(x) = n / k.

[0071] Furthermore, since the construction of the building machine involves various working conditions, including construction conditions and extreme wind conditions, a scenario-based probabilistic method is adopted. This method assumes that the external loads follow different probability distributions under different working conditions, and sets the structural reliability under different working conditions as a probabilistic constraint in the optimization problem, thereby obtaining a steel platform structure for the aerial building machine that can adapt to multiple working conditions.

[0072] Specifically, in the field of structural reliability research, common loads follow a certain probability distribution. The goal of the optimization problem in this case is to minimize the structural stress ratio, displacement, and steel consumption. These are treated as three objective functions, and the optimization design aims to minimize them. The stress ratio is used to assess the structural material's safety capability from a strength perspective; the maximum displacement is used to assess the overall structural deformation and stiffness; and the steel consumption is used to evaluate the cost-effectiveness of the structural design. There is a certain relationship between the three objectives of structural stress ratio, displacement, and steel consumption, which need to be comprehensively considered and optimized during the design process. By optimizing the structural design, the goals of reducing the structural stress ratio, minimizing displacement, and reducing steel consumption can be achieved simultaneously.

[0073] Structural reliability under different operating conditions is considered as a constraint. The structural reliability under different operating conditions is sampled and calculated using the corresponding load distributions. The optimization problem can be summarized as follows:

[0074] min F(x)=(f1(x),f2(x),f3(x))

[0075] st

[0076]

[0077]

[0078]

[0079]

[0080] in, and These represent the structural stress reliability and displacement reliability under construction conditions, respectively. and λ represents the structural stress reliability and displacement reliability under high wind conditions, respectively, and λ is the reliability threshold, which is set to 0.95 in this case.

[0081] The NSGA-II dataset was set to 50 individuals, with 200 iterations, a crossover rate of 0.95, and a mutation rate of 0.05. The VIKOR multi-criteria decision-making method, with a decision coefficient of 0.5, was used to select the optimal design scheme. According to the "Steel Structure Design Standard" (GB 50017-2017), the stress ratio of a steel structure system cannot exceed 1, while the maximum vertical displacement cannot exceed 120 mm, and the lateral displacement cannot exceed 96 mm. In this case, the stress ratio and displacement control thresholds were set to 1 and 96 mm, respectively. Based on this, to calculate and verify the structural reliability of the optimal design scheme, 10,000 random samples were generated according to the defined load probability distribution. The prediction results of the 10,000 Monte Carlo samples were statistically analyzed using a trained ensemble deep learning model to obtain the structural reliability. Furthermore, to test the optimization effect of the optimal design scheme, 100 test cases were generated using the Optimal Space Filling Design (OSF) and compared with the optimal design scheme.

[0082] The results show that the method of this invention can select the optimal design scheme from the Pareto front (finding the optimal solution in a multi-objective optimization problem), as shown in Figure 6, and its specific design parameters are shown in Table 5. After 10,000 Monte Carlo tests, the structural reliability can reach as low as 0.9999. The frequency distribution and cumulative distribution indicate that the optimal design scheme obtained by the generative design method of this invention can meet the reliability requirements under various working conditions. Through finite element simulation, it can be seen from the structural stress and displacement cloud diagrams that the optimal scheme has a reasonable structural design, and all indicators are within the safe range. By comparing with 100 test samples generated by the optimal space-filling design (OSF), the overall improvement rate of the three objective values ​​of the optimal design scheme under construction conditions is 16.59%, and the overall improvement rate under extreme wind conditions is 16.91%. The optimization rate of each objective value is shown in Table 6. The results show that the optimal design scheme has a significant optimization improvement effect compared with the test samples.

[0083] Table 5. Optimal Design Scheme Design Parameters

[0084]

[0085]

[0086] Table 6. Comparison of Improvements from the Best Design Scheme

[0087]

[0088] The present invention also provides, for example Figure 7 The structural design and optimization system 100 for an aerial building machine adapted to multiple operating conditions, as shown, includes the following modules:

[0089] Parametric design module 101 is used to parametrically process the structure of the aerial building machine. The parametric processing is carried out in Grasshopper to obtain the parameters of the aerial building machine structure after parametric processing.

[0090] AI agent model module 102 is used to train an integrated deep learning model based on the data of the structural parameters of the aerial building machine on the basis of the snapshot strategy. The integrated deep learning model uses a convolutional neural network model as the base model. After M cycles of training, M neural network models are obtained, and the average output of the M neural network models is used as the prediction result of the integrated deep learning model.

[0091] The objective optimization module 103 is used to propose a stochastic optimization method based on LHS by using the NSGA-II multi-objective optimization algorithm on the basis of the prediction results. The constraints in the optimization problem are replaced with probabilistic constraints. By using the LHS method to perform multiple samplings in the probability density distribution of external loads, the uncertainty of external loads is simulated, and a steel platform structure for aerial building construction machines that can adapt to multiple working conditions is obtained.

[0092] The present invention also provides a computer storage medium storing program instructions, wherein, when the program instructions are executed, the device where the computer storage medium is located executes the above-described structural design and optimization method for an aerial building machine adapted to multiple working conditions.

[0093] In summary, by parameterizing the aerial building machine structure, the parameterized structural parameters were obtained. Based on a snapshot strategy, an ensemble deep learning model was trained using these structural parameters. The ensemble deep learning model used a convolutional neural network (CNN) model as its base model, and after M training cycles, M neural network models were obtained. The average output of these M neural network models was used as the prediction result of the ensemble deep learning model. Based on the prediction result, the NSGA-II multi-objective optimization algorithm was used, and a stochastic optimization method based on LHS was proposed. This method replaced the constraints in the optimization problem with probabilistic constraints. By employing the LHS method to perform multiple samplings in the probability density distribution of external loads, the uncertainty of external loads was simulated, resulting in a steel platform structure for the aerial building machine adaptable to various operating conditions. This enabled the rapid generation and flexible selection of structural design schemes for the building machine under multiple operating conditions. Through parameterization of the building machine structure and the proposal of an ensemble deep learning model based on a snapshot strategy, the NSGA-II multi-objective optimization algorithm, which considers uncertainty, was used to quickly generate a steel platform structure for the building machine that meets the usage requirements, achieving a structural reliability of 0.9999. Compared to 100 test cases generated using Optimal Space Filling (OSF), the optimal design scheme obtained by this invention achieves an overall improvement rate of 16.59% under construction conditions and 16.91% under extreme wind conditions. The results demonstrate that the LHS-based stochastic optimization method and ensemble deep learning model used in this invention possess good accuracy and computational efficiency. This invention effectively enhances the visualization, interactivity, and intelligence level of the aerial building machine structural design process.

[0094] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0095] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0096] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0097] The above-described embodiments are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0098] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A structural design and optimization method for an aerial building machine adaptable to multiple working conditions, characterized in that, The method includes the following steps: Step S1: Parametric processing of the aerial building machine structure is performed in Grasshopper to obtain the parameters of the aerial building machine structure after parametric processing. Step S2: Based on the snapshot strategy, train an ensemble deep learning model using the structural parameters of the aerial building machine. The ensemble deep learning model uses a convolutional neural network model as the base model. After M training cycles, M neural network models are obtained, and the average output of the M neural network models is used as the prediction result of the ensemble deep learning model. The snapshot strategy employs a cyclic cosine annealing learning rate during the training of the ensemble deep learning model, causing the model to converge to multiple local minima. When the learning rate cycle restarts, the performance of the ensemble deep learning model temporarily decreases. After the learning rate annealing, the performance eventually surpasses that of the previous cycle. The training process is divided into M cycles, each starting with a high learning rate and gradually decreasing to a lower learning rate. The initial learning rate provides sufficient energy for the ensemble deep learning model to move away from the current gradient point, while the lowest learning rate guides the model to a well-performing local minimum. At the end of each training cycle, the ensemble deep learning model reaches a local minimum in terms of training loss. Before increasing the learning rate, the weights and biases of the ensemble deep learning model are saved once, where M is a positive integer greater than or equal to 1. Step S3: Based on the prediction results, the NSGA-II multi-objective optimization algorithm is used to propose a stochastic optimization method based on LHS. The constraints in the optimization problem are replaced with probabilistic constraints. By using the LHS method to perform multiple samplings in the probability density distribution of the external load, the uncertainty of the external load is simulated, and a steel platform structure for an aerial building machine that is adaptable to multiple working conditions is obtained. The constraint conditions in step S3 refer to the structural reliability under different working conditions. The structural reliability under different working conditions is sampled and calculated by the corresponding load distribution. The different working conditions include construction working conditions and abnormal working conditions. In step S3, during the optimization process, after obtaining k samples from the probability density distribution of each variable, the output target value of the k samples is calculated by the optimization algorithm, where n represents the number of results within the constraint range, and the probability is determined as p(x) = n / k. By adopting a scenario-based probabilistic approach, assuming that the external loads follow different probability distributions under different working conditions, the structural reliability under different working conditions is set as a probabilistic constraint in the optimization problem, thereby obtaining a steel platform structure for aerial building construction machines that can adapt to multiple working conditions.

2. The structural design and optimization method for a multi-condition aerial building machine according to claim 1, characterized in that, In step S1, the parameterization process of the aerial building machine includes the following steps: Step S11: First, determine the position of the outer truss of the steel platform based on the position of the outer shear wall of the core tube. Second, determine the movement range of the supporting steel column of the steel platform based on the position of the inner shear wall of the core tube. Then, determine the support point of the outer steel column of the steel platform based on the truss cantilever length requirement. All other parameters except the above parameters are parameterized. Parameterization means using a certain method to replace fixed values ​​with parameters. Step S12: According to the module design rules, the moving step of the supporting steel column is greater than or equal to the first module. By setting judgment statements through parametric programming, the corresponding type of Bailey panels are automatically generated on the upper part according to the moving distance of the supporting steel column, thereby forming a steel platform. Transverse supports are set between the outer truss and the main truss of the steel platform, and the number of struts is parameterized to complete the parametric modeling of the aerial building machine. The module refers to the selected standard scale unit of measurement.

3. The structural design and optimization method for an aerial building machine adaptable to multiple working conditions as described in claim 2, characterized in that, After completing the parametric modeling of the aerial building machine, the process further includes: converting the geometric model of the aerial building machine into a physical model by setting material properties, section allocation, external loads, boundary constraints, and spatial mesh, realizing data flow transfer between the geometric model and the physical model, and finally calculating and outputting structural response data, wherein the structural response refers to the stress, displacement, deformation, and buckling caused by the structure being subjected to external forces.

4. The structural design and optimization method for a multi-condition aerial building machine according to claim 1, characterized in that, The convolutional neural network (1D-CNN) model described in step S2 mainly consists of a convolutional feature extraction layer and a fully connected layer. The convolutional feature extraction layer is composed of convolutional layers and pooling layers and is used to extract data features. The fully connected layer is used for regression output of feature data.

5. A structural design and optimization system for a multi-condition aerial building machine, used to implement the method described in any one of claims 1-4, characterized in that, The system includes the following modules: The parametric design module is used to parametrically process the structure of the aerial building machine. The parametric processing is performed in Grasshopper to obtain the parameters of the aerial building machine structure after parametric processing. The AI ​​agent model module is used to train an integrated deep learning model based on the data of the structural parameters of the aerial building machine, using the snapshot strategy. The integrated deep learning model uses a convolutional neural network model as the base model, and after M training cycles, it obtains M neural network models. The average output of the M neural network models is used as the prediction result of the integrated deep learning model. The objective optimization module is used to propose a stochastic optimization method based on LHS using the NSGA-II multi-objective optimization algorithm based on the prediction results. The constraints in the optimization problem are replaced with probabilistic constraints. By using the LHS method to perform multiple samplings in the probability density distribution of the external load, the uncertainty of the external load is simulated, and a steel platform structure for an aerial building machine that can adapt to multiple working conditions is obtained.

6. A computer storage medium, characterized in that, The storage medium stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to perform the structural design and optimization method for an aerial building machine adapted to multiple working conditions, as described in any one of claims 1-4.

Images