An optimization method for buckling-resistance design of large-size corrugated steel webs

Abstract

The present invention relates to the field of design optimization, and in particular to a large-size corrugated steel web buckling prevention design optimization method. A large-size corrugated steel web buckling prevention design optimization method comprises the steps of corrugated steel web simulation individual setting, fitness calculation, finder set, follower set and sentinel set division and corrugated steel web simulation individual iteration. The present invention optimizes the corrugated steel web buckling prevention design through an improved sparrow search algorithm, and can achieve the optimal corrugated steel web buckling prevention design on the basis of considering the material utilization rate; and in the improved sparrow search algorithm, the corrugated steel web simulation individual X is optimized through a corrugated steel web design model established based on a generative adversarial network. n Global search can provide a fusion of high-quality individuals with the help of the learned favorable directions of design optimization, thereby achieving more valuable searches and improving the convergence speed in the early stages.

Description

Technical Field

[0001] The invention relates to the field of design optimization, and in particular to a buckling prevention design optimization method for a large-size corrugated steel web. Background Art

[0002] As a new type of structure, corrugated steel webs have been widely used in bridges, buildings and other large-scale structural projects due to their excellent mechanical properties and high material utilization efficiency. The corrugated steel webs can significantly enhance the overall stiffness and stability of the structure and reduce the risk of local buckling of the webs through their unique wavy geometric structure. However, with the continuous expansion of engineering applications, the traditional corrugated steel web design still faces many challenges when dealing with complex loads, long-span structures and extreme environmental conditions.

[0003] Buckling failure is one of the most common and critical failure modes of corrugated steel webs in actual engineering. Especially when subjected to large compressive stress or shear stress, the web is prone to local buckling, resulting in a decrease in the bearing capacity of the overall structure, and may even lead to catastrophic consequences. Therefore, how to optimize the anti-buckling design of corrugated steel webs to improve their anti-buckling performance and extend their service life has become an important issue that needs to be solved in the field of structural engineering. Summary of the invention

[0004] The present invention optimizes the buckling prevention design of corrugated steel webs by using an improved sparrow search algorithm, and can realize the optimal buckling prevention design of corrugated steel webs on the basis of considering the material utilization rate; and in the improved sparrow search algorithm, the corrugated steel web design model established based on the generative adversarial network is used to simulate the individual X-ray diffraction patterns of the corrugated steel webs. n Global search can provide a fusion of high-quality individuals with the help of the learned favorable directions of design optimization, thereby achieving more valuable searches and improving the convergence speed in the early stages.

[0005] A buckling-restrained design optimization method for large-size corrugated steel webs, comprising:

[0006] Step S1: individual setting of corrugated steel web simulation;

[0007] Construct several corrugated steel web simulation individuals X n , n=1, 2, 3, ..., N, corrugated steel web simulation individual X n The corrugated steel web design parameters are included; all corrugated steel webs are simulated as individual X n Form a population set and set the maximum number of iterations W;

[0008] Step S2: Let w=1, where w is used to record the number of iterations;

[0009] Step S3: constructing a first simulation intermediate set and a second simulation intermediate set;

[0010] Calculate the number of corrugated steel web simulation individuals X in the population set n The corresponding fitness δ n , and simulate all the corrugated steel webs in the population set as individuals X n According to the corresponding fitness δ n Arrange from large to small and select the first M corrugated steel webs to simulate individual X n Form the first simulation intermediate set; set the fitness δ in the population set n The largest corrugated steel web simulation unit X n Note: X best , the fitness δ in the population set n Minimum corrugated steel web simulates individual X n Note: X worst , and the corrugated steel web is simulated as an individual X best Stored in the optimal individual library; output H corrugated steel web simulation individuals X through the trained corrugated steel web design model n , and satisfy H+M=N, N is the corrugated steel web simulation individual X in the population set n The corrugated steel web design model is set based on the generative adversarial network, and the corrugated steel web design model is trained in each iteration. The H corrugated steel web simulation individuals X output by the corrugated steel web design model are n The second simulation intermediate set is formed; the H value is based on all the corrugated steel web simulation individuals X in the population set n The diversity and number of iterations are set;

[0011] Step S4: dividing the discoverer set, follower set and sentinel set;

[0012] Select the fitness δ in the second simulation intermediate set n The largest corrugated steel web simulation unit X n Note: X p , traverse the first simulation intermediate set, and simulate individual X for each selected corrugated steel web n , the selected corrugated steel web is simulated as individual X n The corresponding fitness δ n Simulate individual X with corrugated steel webs p The corresponding fitness δ n For comparison, if the corrugated steel web is selected to simulate the individual X n The corresponding fitness δ n Higher than the corrugated steel web simulation individual X p The corresponding fitness δ n , the corrugated steel web will be selected to simulate the individual Xn Stored in the finder collection; if the selected corrugated steel web simulation individual X n The corresponding fitness δ n Not higher than the corrugated steel web simulation unit X p The corresponding fitness δ n , the corrugated steel web will be selected to simulate the individual X n Store it in the follower set; record the second simulation intermediate set as the vigilant set;

[0013] Step S5: individual iteration of corrugated steel web simulation;

[0014] Iteratively update the discoverer set, follower set and vigilant set, and combine the updated discoverer set, follower set and vigilant set to construct a population set;

[0015] Step S6: Determine whether w<W is true. If so, assign w+1 to w and return to step S3. If w<W is not true, select the fitness δ in the optimal individual library. n The largest corrugated steel web simulation unit X n The result is output as a design parameter set of corrugated steel web to optimize the buckling-resistance design of corrugated steel web.

[0016] As a preferred aspect, based on all the corrugated steel webs in the population set, individual X is simulated. n The diversity and number of iterations set the H value, which specifically includes the following steps:

[0017] Calculate all corrugated steel web simulation individuals X in the population set n The diversity value R;

[0018] The H value is calculated by the following formula:

[0019] ;

[0020] Where floor is the rounding function; H0 is the initial value; μ is the adjustment parameter; R0 is the diversity threshold.

[0021] As a preferred aspect, the trained corrugated steel web design model is used to output H corrugated steel web simulation individuals X n , specifically including the following steps:

[0022] H noise vectors are collected from the standard normal distribution, all noise vectors are traversed, and for each selected noise vector, the selected noise vector is spliced ​​after the standard corrugated steel web design parameter vector to construct a corrugated steel web simulation individual generation vector; until all noise vectors are traversed, H corrugated steel web simulation individual generation vectors are output; the standard corrugated steel web design parameter vector includes the corrugated steel web design parameters;

[0023] The vectors generated by H corrugated steel web simulation individuals are input into the trained corrugated steel web design model for processing, and H corrugated steel web simulation individuals X are output. n .

[0024] As a preferred aspect, training is performed on the corrugated steel web design model, specifically comprising the following steps:

[0025] Step E1: obtaining the first simulation intermediate set as a training set for corrugated steel web simulation individuals;

[0026] Step E2: Initialize the discriminator and the generator, wherein the discriminator and the generator are both established based on the BP neural network;

[0027] Step E3: Randomly select a part of the corrugated steel web simulation individuals from the corrugated steel web simulation individual generation training set as positive training samples, and annotate the positive training samples with positive labels, and form all the annotated positive training samples into a positive training set; splice a random noise vector after the standard corrugated steel web design parameter vector to construct a first corrugated steel web simulation individual generation training vector, input each first corrugated steel web simulation individual generation training vector into the generator for processing, output negative training samples, and annotate the negative training samples with negative labels, and form all the annotated negative training samples into a negative training set; train the discriminator with the positive training set and the negative training set;

[0028] Step E4: connect the generator and the discriminator, fix the parameters of the discriminator unchanged, splice the random noise vector after the standard corrugated steel web design parameter vector to construct the second corrugated steel web simulation individual generation training vector, annotate the second corrugated steel web simulation individual generation training vector through the forward label, form the generator training set with all the annotated second corrugated steel web simulation individual generation training vectors, and send the generator training set to the generator and the discriminator for training;

[0029] Step E5: Repeat steps E3 and E4 until the training conditions are met, and output the generator as a corrugated steel web design model.

[0030] As a preferred aspect, the calculation of each corrugated steel web simulation individual X n The corresponding fitness δ n , specifically including the following steps:

[0031] Simulation of individual X based on corrugated steel web n All corrugated steel web design parameters are analyzed by finite element analysis, and the maximum stress T is output. n The corrugated steel web simulation individual X is calculated by the following formula nThe corresponding fitness δ n :

[0032] ;

[0033] Where α1 and α2 are the first weight coefficient and the second weight coefficient respectively, satisfying α1+α2=1, and T0 is the critical buckling stress.

[0034] As a preferred aspect, the corrugated steel web simulation individual X in the finder set is simulated by the following formula: n To update:

[0035] ;

[0036] in, For the corrugated steel web simulation individual X at the wth iteration n The j-th corrugated steel web design parameter, j = 1, 2, 3, ..., J, J is the total number of corrugated steel web design parameters in the corrugated steel web simulation individual; η is any random number between 0 and 1; R2 is the warning value, which is in the interval [0,1]; ST is the safety value, which is in the interval [0.5,1]; Q is a random number that obeys the normal distribution; L is a 1×d matrix with all elements being 1, and d is the dimension of the corrugated steel web design parameter.

[0037] As a preferred aspect, the corrugated steel web simulation individual X in the follower set is simulated by the following formula: n To update:

[0038] ;

[0039] in, For the corrugated steel web simulation individual X at the wth iteration p The jth corrugated steel web design parameter, x worst，j Model individual X for corrugated steel web worst The jth corrugated steel web design parameters in A + is a 1×d matrix, containing two random elements -1 and 1; k is all the corrugated steel web simulation individuals X in the follower set n According to the fitness δ n The corresponding serial number value after the order from large to small, K is all the corrugated steel web simulation individuals X in the follower set n The total number of .

[0040] As a preferred aspect, for the corrugated steel web simulation individual X in the alerter set n To update, the specific steps include:

[0041] Select the corrugated steel web simulation individual X one by one from the vigilante collection n , for the selected corrugated steel web, simulate individual X n , execute the following content, generate a random number ζ between 0 and 1 through a random function, and determine whether ζ>Pc is true, P c is the transformation threshold. If ζ>Pc does not hold, no operation is performed. Otherwise, the selected corrugated steel web is simulated by the following formula: n To update:

[0042] ;

[0043] Where U is a random number between -1 and 1; β is the step size control parameter; δ best Model individual X for corrugated steel web best The corresponding fitness, δ worst Model individual X for corrugated steel web worst The corresponding fitness, x best，j Model individual X for corrugated steel web best Design parameters of the jth corrugated steel web in .

[0044] The present invention has the following advantages:

[0045] The present invention optimizes the buckling prevention design of corrugated steel webs by using an improved sparrow search algorithm, and can realize the optimal buckling prevention design of corrugated steel webs on the basis of considering the material utilization rate; and in the improved sparrow search algorithm, the corrugated steel web design model established based on the generative adversarial network is used to simulate the individual X-ray diffraction patterns of the corrugated steel webs. n Global search can provide a fusion of high-quality individuals with the help of the learned favorable directions of design optimization, thereby achieving more valuable searches and improving the convergence speed in the early stages. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic flow chart of the anti-buckling design optimization method of the corrugated steel web adopted in the embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0048] Example 1, a large-size corrugated steel web buckling protection design optimization method, such as Figure 1 As shown, including:

[0049] Step S1: individual setting of corrugated steel web simulation;

[0050] Construct several corrugated steel web simulation individuals X n , n=1, 2, 3, ..., N, corrugated steel web simulation individual X n The corrugated steel web design parameters are included, such as the corrugated steel web material, wave height, wave length and wave thickness, etc. These corrugated steel web design parameters have a direct impact on the buckling bearing capacity of the corrugated steel web; when constructing several corrugated steel web simulation individual X n When the design parameters of each corrugated steel web are set, random values ​​will be generated within the corresponding range, and all the generated random values ​​will be combined into several corrugated steel web simulation individual X n ; Simulate all corrugated steel webs as individuals X n Form a population set and set the maximum number of iterations W;

[0051] Step S2: Let w=1, where w is used to record the number of iterations;

[0052] Step S3: constructing a first simulation intermediate set and a second simulation intermediate set;

[0053] Calculate the number of corrugated steel web simulation individuals X in the population set n The corresponding fitness δ n , and simulate all the corrugated steel webs in the population set as individuals X n According to the corresponding fitness δ n Arrange from large to small and select the first M corrugated steel webs to simulate individual X n The first simulation intermediate set is composed of M corrugated steel webs simulating individual X n It represents a part with higher fitness, which also indicates that the corresponding corrugated steel web design parameters have better anti-buckling ability; the fitness δ in the population set n The largest corrugated steel web simulation unit X n Note: X best , the fitness δ in the population set n Minimum corrugated steel web simulates individual X n Note: X worst , and the corrugated steel web is simulated as an individual X best Stored in the optimal individual library; output H corrugated steel web simulation individuals X through the trained corrugated steel web design model n , and satisfy H+M=N, N is the corrugated steel web simulation individual X in the population set n The corrugated steel web design model is set based on the generative adversarial network, and the corrugated steel web design model is trained in each iteration. The H corrugated steel web simulation individuals X output by the corrugated steel web design model are nThe second simulation intermediate set is composed of the corrugated steel web design model. It should be noted that the corrugated steel web design model can learn the favorable direction of corrugated steel web design optimization, and generate corrugated steel web design individuals based on the learned information to realize the global search of corrugated steel web design individuals. Different from moving directly to the optimal solution, the traditional optimal solution movement is to iterate the optimal solution for each parameter separately without considering the coordination between the parameters. The corrugated steel web simulation individuals generated by the corrugated steel web design model can achieve more valuable search under the guidance of favorable directions and improve the convergence speed in the early stage. The H value is based on all the corrugated steel web simulation individuals X in the population set. n The diversity and number of iterations are set;

[0054] Simulate individual X based on all corrugated steel webs in the population n The diversity and number of iterations set the H value, which specifically includes the following steps:

[0055] Calculate all corrugated steel web simulation individuals X in the population set n The diversity value R is the sum of all corrugated steel web simulation individuals X in the population set. n The corresponding fitness δ n The standard deviation or variance of

[0056] The H value is calculated by the following formula:

[0057] ;

[0058] Where floor is the rounding function; H0 is the initial value, which is generally half of the N value; μ is an adjustment parameter determined by the designer and is used to map the value of R0-R to (-1, 1); R0 is the diversity threshold determined by the designer; By calculating the H value by the above formula, when the diversity is large in the early stage of iteration, more corrugated steel web simulation individuals output by the corrugated steel web design model can be integrated to increase the search space, and when the diversity is small in the later stage of iteration, the corrugated steel web simulation individuals output by the corrugated steel web design model can be reduced to improve the iteration speed;

[0059] The trained corrugated steel web design model is used to output H corrugated steel web simulation individuals X n , specifically including the following steps:

[0060] H noise vectors are collected from the standard normal distribution. It should be noted that the present embodiment collects from the standard normal distribution, and can also collect from uniform distribution and multi-peak distribution, etc., without limitation. All noise vectors are traversed, and for each selected noise vector, the selected noise vector is spliced ​​after the standard corrugated steel web design parameter vector to construct a corrugated steel web simulation individual generation vector; until all noise vectors are traversed, H corrugated steel web simulation individual generation vectors are output; the standard corrugated steel web design parameter vector includes the corrugated steel web design parameters, which are determined by the designer, and generally the median value within the corresponding range of the corrugated steel web design parameters is selected to form the standard corrugated steel web design parameter vector;

[0061] The vectors generated by H corrugated steel web simulation individuals are input into the trained corrugated steel web design model for processing, and H corrugated steel web simulation individuals X are output. n ;

[0062] Training the corrugated steel web design model includes the following steps:

[0063] Step E1: obtaining the first simulation intermediate set as a training set for generating corrugated steel web simulation individuals. The first simulation intermediate set stores corrugated steel web simulation individuals with high fitness, which can reflect the favorable direction of design optimization;

[0064] Step E2: Initialize the discriminator and the generator, wherein the discriminator and the generator are both established based on the BP neural network;

[0065] Step E3: randomly select a part of the corrugated steel web simulation individuals from the corrugated steel web simulation individual generation training set as positive training samples, and annotate the positive training samples with positive labels, the positive labels are generally annotated as 1, and all the annotated positive training samples form a positive training set; splice a random noise vector after the standard corrugated steel web design parameter vector to construct a first corrugated steel web simulation individual generation training vector, input each first corrugated steel web simulation individual generation training vector into the generator for processing, output negative training samples, and annotate the negative training samples with negative labels, the negative labels are generally annotated as 0, and all the annotated negative training samples form a negative training set; train the discriminator with the positive training set and the negative training set;

[0066] Step E4: connect the generator and the discriminator, fix the parameters of the discriminator unchanged, splice the random noise vector after the standard corrugated steel web design parameter vector to construct the second corrugated steel web simulation individual generation training vector, mark the second corrugated steel web simulation individual generation training vector by the positive label, the positive label is generally marked as 1, and all the marked second corrugated steel web simulation individual generation training vectors form a generator training set, and send the generator training set to the generator and the discriminator for training. During the training, take the positive label as the target, adjust the parameters in the generator, so that the corrugated steel web simulation individual generated by the generator is more in line with the favorable direction of design optimization;

[0067] Step E5: Repeat steps E3 to E4 until the training condition is met, and output the generator as the corrugated steel web design model; the training condition may be that the loss of the generator is minimal, and the loss of the generator is the loss between the output of the generator after being processed by the discriminator and 1;

[0068] Calculate each corrugated steel web simulation individual X n The corresponding fitness δ n , specifically including the following steps:

[0069] Simulation of individual X based on corrugated steel web n Finite element analysis is performed on all the corrugated steel web design parameters in the experiment. The finite element analysis can be performed using the corresponding finite element analysis software, such as ANSYS software. First, the individual X-shaped corrugated steel web is simulated based on the corrugated steel web. n The corresponding corrugated steel web finite element model is constructed based on all the corrugated steel web design parameters in the formula, and then the material properties and loads are set. Finally, the stress of different units is solved by the finite element solver, and the maximum stress T is output. n The corrugated steel web simulation individual X is calculated by the following formula n The corresponding fitness δ n :

[0070] ;

[0071] Where α1 and α2 are the first weight coefficient and the second weight coefficient respectively, satisfying α1+α2=1, T0 is the critical buckling stress, which is determined by the designer based on the elastic modulus of the material and the geometric characteristics of the structure; the above formula can be used to describe the individual X of each corrugated steel web simulation. n The degree of approximation between the maximum stress that the corresponding corrugated steel web can withstand and the critical buckling stress;

[0072] Step S4: dividing the discoverer set, follower set and sentinel set;

[0073] Select the fitness δ in the second simulation intermediate set nThe largest corrugated steel web simulation unit X n Note: X p , traverse the first simulation intermediate set, and simulate individual X for each selected corrugated steel web n , the selected corrugated steel web is simulated as individual X n The corresponding fitness δ n Simulate individual X with corrugated steel webs p The corresponding fitness δ n For comparison, if the corrugated steel web is selected to simulate the individual X n The corresponding fitness δ n Higher than the corrugated steel web simulation individual X p The corresponding fitness δ n , the corrugated steel web will be selected to simulate the individual X n Stored in the finder set, the finder set is set in advance and initially empty; if the selected corrugated steel web simulation individual X n The corresponding fitness δ n Not higher than the corrugated steel web simulation unit X p The corresponding fitness δ n , the corrugated steel web will be selected to simulate the individual X n Store in the follower set, the follower set is set in advance and is initially empty; record the second simulation intermediate set as the vigilant set;

[0074] Step S5: individual iteration of corrugated steel web simulation;

[0075] Iteratively update the discoverer set, follower set and vigilant set, and combine the updated discoverer set, follower set and vigilant set to construct a population set;

[0076] The corrugated steel web in the finder set is simulated by the following formula: n To update:

[0077] ;

[0078] in, For the corrugated steel web simulation individual X at the wth iteration n The jth corrugated steel web design parameter, j = 1, 2, 3, ..., J, J is the total number of corrugated steel web design parameters in the corrugated steel web simulation individual; η is any random number between 0 and 1; R2 is the warning value, which is in the interval [0,1]; ST is the safety value, which is in the interval [0.5,1]; Q is a random number that obeys the normal distribution; L is a 1×d matrix with all elements being 1, and d is the dimension of the corrugated steel web design parameter;

[0079] The corrugated steel web in the follower set is simulated by the following formulan To update:

[0080] ;

[0081] in, For the corrugated steel web simulation individual X at the wth iteration p The jth corrugated steel web design parameter, x worst，j Model individual X for corrugated steel web worst The jth corrugated steel web design parameters in A + is a 1×d matrix, containing two random elements -1 and 1; k is all the corrugated steel web simulation individuals X in the follower set n According to the fitness δ n The corresponding serial number value after the order from large to small, K is all the corrugated steel web simulation individuals X in the follower set n Total number of;

[0082] Modeling individual X for the corrugated steel web in the Vigilante set n To update, the specific steps include:

[0083] Select the corrugated steel web simulation individual X one by one from the vigilante collection n , for the selected corrugated steel web, simulate individual X n , execute the following content, generate a random number ζ between 0 and 1 through a random function, and determine whether ζ>Pc is true, P c is the transformation threshold, which is set by the designer and is generally 0.7. If ζ>Pc does not hold, no operation is performed; otherwise, the selected corrugated steel web is simulated by the following formula for the individual X n To update:

[0084] ;

[0085] Where U is a random number between -1 and 1; β is the step size control parameter; δ best Model individual X for corrugated steel web best The corresponding fitness, δ worst Model individual X for corrugated steel web worst The corresponding fitness, x best，j Model individual X for corrugated steel web best Design parameters of the jth corrugated steel web in ;

[0086] Step S6: Determine whether w<W is true. If so, it means that the iteration has not been completed. Assign w+1 to w and return to step S3. If w<W is not true, it means that the iteration has been completed. Select the fitness δ in the optimal individual library. n The largest corrugated steel web simulation unit Xn The result is output as a design parameter set of corrugated steel web to optimize the buckling-resistance design of corrugated steel web.

[0087] The present invention optimizes the buckling prevention design of corrugated steel webs by using an improved sparrow search algorithm, and can achieve the optimal buckling prevention design of corrugated steel webs on the basis of considering the material utilization rate; and in the improved sparrow search algorithm, the corrugated steel web design model established based on the generative adversarial network is used to simulate the individual X-ray diffraction patterns of the corrugated steel webs. n Global search can provide a fusion of high-quality individuals with the help of the learned favorable directions of design optimization, thereby achieving more valuable searches and improving the convergence speed in the early stages.

[0088] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A buckling-resistance design optimization method for large-size corrugated steel webs, characterized in that: include: Step S1: individual setting of corrugated steel web simulation; Construct several corrugated steel web simulation individuals X n , n=1, 2, 3, ..., N, corrugated steel web simulation individual X n The corrugated steel web design parameters are included; all corrugated steel webs are simulated as individual X n Form a population set and set the maximum number of iterations W; Step S2: Let w=1, where w is used to record the number of iterations; Step S3: constructing a first simulation intermediate set and a second simulation intermediate set; Calculate the number of corrugated steel web simulation individuals X in the population set n The corresponding fitness δ n , and simulate all the corrugated steel webs in the population set as individuals X n According to the corresponding fitness δ n Arrange from large to small and select the first M corrugated steel webs to simulate individual X n Form the first simulation intermediate set; set the fitness δ in the population set n The largest corrugated steel web simulation unit X n Note: X best , the fitness δ in the population set n Minimum corrugated steel web simulates individual X n Note: X worst , and the corrugated steel web is simulated as an individual X best Stored in the optimal individual library; output H corrugated steel web simulation individuals X through the trained corrugated steel web design model n , and satisfy H+M=N, N is the corrugated steel web simulation individual X in the population set n The corrugated steel web design model is set based on the generative adversarial network, and the corrugated steel web design model is trained in each iteration. The H corrugated steel web simulation individuals X output by the corrugated steel web design model are n The second simulation intermediate set is formed; the H value is based on all the corrugated steel web simulation individuals X in the population set n The diversity and number of iterations are set; Step S4: dividing the discoverer set, follower set and sentinel set; Select the fitness δ in the second simulation intermediate set n The largest corrugated steel web simulation unit X n Note: X p , traverse the first simulation intermediate set, and simulate individual X for each selected corrugated steel web n , the selected corrugated steel web is simulated as individual X n The corresponding fitness δ n Simulate individual X with corrugated steel webs p The corresponding fitness δ n For comparison, if the corrugated steel web is selected to simulate the individual X n The corresponding fitness δ n Higher than the corrugated steel web simulation individual X p The corresponding fitness δ n , the corrugated steel web will be selected to simulate the individual X n Stored in the finder collection; if the selected corrugated steel web simulation individual X n The corresponding fitness δ n Not higher than the corrugated steel web simulation unit X p The corresponding fitness δ n , the corrugated steel web will be selected to simulate the individual X n Store it in the follower set; record the second simulation intermediate set as the vigilant set; Step S5: individual iteration of corrugated steel web simulation; Iteratively update the discoverer set, follower set and vigilant set, and combine the updated discoverer set, follower set and vigilant set to construct a population set; Step S6: Determine whether w<W is true. If w<W is true, assign w+1 to w and return to step S3; If w<W does not hold, select the fitness δ in the optimal individual pool n The largest corrugated steel web simulation unit X n Output as a design parameter set of corrugated steel web to optimize the buckling-resistance design of corrugated steel web; The trained corrugated steel web design model is used to output H corrugated steel web simulation individuals X n , specifically including the following steps: H noise vectors are collected from the standard normal distribution, all noise vectors are traversed, and for each selected noise vector, the selected noise vector is spliced ​​after the standard corrugated steel web design parameter vector to construct a corrugated steel web simulation individual generation vector; until all noise vectors are traversed, H corrugated steel web simulation individual generation vectors are output; the standard corrugated steel web design parameter vector includes the corrugated steel web design parameters; The vectors generated by H corrugated steel web simulation individuals are input into the trained corrugated steel web design model for processing, and H corrugated steel web simulation individuals X are output. n ; Training the corrugated steel web design model includes the following steps: Step E1: obtaining the first simulation intermediate set as a training set for corrugated steel web simulation individuals; Step E2: Initialize the discriminator and the generator, wherein the discriminator and the generator are both established based on the BP neural network; Step E3: Randomly select a part of the corrugated steel web simulation individuals from the corrugated steel web simulation individual generation training set as positive training samples, and annotate the positive training samples with positive labels, and form all the annotated positive training samples into a positive training set; splice a random noise vector after the standard corrugated steel web design parameter vector to construct a first corrugated steel web simulation individual generation training vector, input each first corrugated steel web simulation individual generation training vector into the generator for processing, output negative training samples, and annotate the negative training samples with negative labels, and form all the annotated negative training samples into a negative training set; train the discriminator with the positive training set and the negative training set; Step E4: connect the generator and the discriminator, fix the parameters of the discriminator unchanged, splice the random noise vector after the standard corrugated steel web design parameter vector to construct the second corrugated steel web simulation individual generation training vector, annotate the second corrugated steel web simulation individual generation training vector through the forward label, form the generator training set with all the annotated second corrugated steel web simulation individual generation training vectors, and send the generator training set to the generator and the discriminator for training; Step E5: Repeat steps E3 and E4 until the training conditions are met, and output the generator as a corrugated steel web design model.

2. The buckling-resistance design optimization method for large-size corrugated steel webs according to claim 1 is characterized in that: Simulate individual X based on all corrugated steel webs in the population n The diversity and number of iterations set the H value, which specifically includes the following steps: Calculate all corrugated steel web simulation individuals X in the population set n The diversity value R; The H value is calculated by the following formula: ； Where floor is the rounding function; H0 is the initial value; μ is the adjustment parameter; R0 is the diversity threshold.

3. The buckling-resistance design optimization method for large-size corrugated steel webs according to claim 2 is characterized in that: Calculate each corrugated steel web simulation individual X n The corresponding fitness δ n , specifically including the following steps: Simulation of individual X based on corrugated steel web n All corrugated steel web design parameters are analyzed by finite element analysis, and the maximum stress T is output. n The corrugated steel web simulation individual X is calculated by the following formula n The corresponding fitness δ n : ； Where α1 and α2 are the first weight coefficient and the second weight coefficient respectively, satisfying α1+α2=1, and T0 is the critical buckling stress.

4. The buckling-resistance design optimization method for large-size corrugated steel webs according to claim 3 is characterized in that: The corrugated steel web in the finder set is simulated by the following formula: n To update: ； in, For the corrugated steel web simulation individual X at the wth iteration n The j-th corrugated steel web design parameter, j = 1, 2, 3, ..., J, J is the total number of corrugated steel web design parameters in the corrugated steel web simulation individual; η is any random number between 0 and 1; R2 is the warning value, which is in the interval [0,1]; ST is the safety value, which is in the interval [0.5,1]; Q is a random number that obeys the normal distribution; L is a 1×d matrix with all elements being 1, and d is the dimension of the corrugated steel web design parameter.

5. The buckling-resistance design optimization method for large-size corrugated steel webs according to claim 4 is characterized in that: The corrugated steel web in the follower set is simulated by the following formula n To update: ； in, For the corrugated steel web simulation individual X at the wth iteration p The jth corrugated steel web design parameter, x worst，j Model individual X for corrugated steel web worst The jth corrugated steel web design parameters in A + is a 1×d matrix, containing two random elements -1 and 1; k is all the corrugated steel web simulation individuals X in the follower set n According to the fitness δ n The corresponding serial number value after the order from large to small, K is all the corrugated steel web simulation individuals X in the follower set n The total number of .

6. The buckling-resistance design optimization method for large-size corrugated steel webs according to claim 5 is characterized in that: Modeling individual X for the corrugated steel web in the Vigilante set n To update, the specific steps include: Select the corrugated steel web simulation individual X one by one from the vigilante collection n , for the selected corrugated steel web, simulate individual X n , execute the following content, generate a random number ζ between 0 and 1 through a random function, and determine whether ζ>Pc is true, P c is the transformation threshold. If ζ>Pc does not hold, no operation is performed. Otherwise, the selected corrugated steel web is simulated by the following formula: n To update: ； Where U is a random number between -1 and 1; β is the step size control parameter; δ best Model individual X for corrugated steel web best The corresponding fitness, δ worst Model individual X for corrugated steel web worst The corresponding fitness, x best，j Model individual X for corrugated steel web best Design parameters of the jth corrugated steel web in .

Images