A method for designing a reinforced steel plate shear wall based on a generative adversarial network and fusing image-text features

Abstract

The application discloses a kind of based on the design method of steel plate shear wall of fusion image-text feature of generative adversarial network, which comprises the following steps: step S1, the S.Mises diagram and ESEDEN diagram of steel plate shear wall needing to be designed are obtained, and its classification group is determined;Step S2, the S.Mises diagram and ESEDEN diagram of steel plate shear wall needing to be designed are input into pre-trained BraceGAN-TXT, and BraceGAN-TXT intelligent generation design is obtained;Step S3, according to intelligent generation design drawing, it is deepened and designed, and the final stiffening rib design is obtained.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of civil engineering and computer deep learning application, and particularly relates to a stiffened steel plate shear wall design method based on image-text feature fusion of a generative adversarial network. BACKGROUND

[0002] Although computers can learn abstract semantic information contained in images, they cannot learn abstract semantic information implied in structural design condition texts. Existing studies have shown that in seismic design, in addition to the input of building images, seismic design condition texts are often needed to guide structural design to meet seismic design requirements. At present, existing generative adversarial networks cannot simultaneously learn image data and text data. Therefore, it is necessary to conduct more in-depth research on the stiffened steel plate shear wall design method to achieve the structural design goal of simultaneously meeting the structural image feature and text requirements. SUMMARY

[0003] The technical problem to be solved by the application is to overcome the deficiencies of the existing stiffened steel plate shear wall design method based on a generative adversarial network and to propose a stiffened steel plate shear wall design method based on image-text feature fusion of a generative adversarial network. This method can generate a stiffened rib layout that meets the target design requirements by inputting steel plate shear wall training images and design texts. Compared with the existing stiffened steel plate shear wall design method based on a generative adversarial network, this method exhibits more excellent design generation capability and can effectively generate a structural scheme that meets the design requirements according to text instructions. The design results are comparable to those of the classic topology optimization design method in terms of planar design indicators and overall structural mechanical properties.

[0004] The technical solution is that in order to solve the aforementioned technical problem, the application proposes a stiffened steel plate shear wall design method based on image-text feature fusion of a generative adversarial network. This method includes the following steps:

[0005] Step S1: Take the S.Mises diagram and ESEDEN diagram of the steel plate shear wall that needs to be designed, and determine its classification group;

[0006] Step S2: Input the S.Mises diagram and ESEDEN diagram of the steel plate shear wall that needs to be designed into the pre-trained BraceGAN-TXT to obtain a BraceGAN-TXT intelligent generated design diagram;

[0007] Step S3: Deepen the design according to the intelligent generated design diagram to obtain the final stiffened rib design.

[0008] Further, in step S1, the S.Mises diagram and ESEDEN diagram production process is as follows:

[0009] (2.1) According to the steel plate shear wall structure design drawing, the corresponding steel plate shear wall model is established in the finite element analysis software ABAQUS;

[0010] (2.2) Create field output in the analysis step module, the scope is the middle steel plate part of the steel plate shear wall, and the output variable is set to MISES, Mises equivalent stress and ELEDEN, all energy density components;

[0011] (2.3) Create and submit the Job file, run and open the Job.odb file, draw the cloud chart on the undeformed chart in the visualization module, and get the S.Mises chart and ESEDEN chart of the middle steel plate.

[0012] Further, the modeling method in step (2.1) is:

[0013] (3.1) In terms of finite element modeling unit selection, the middle embedded steel plate of the steel plate shear wall belongs to the plate shell structure, and the linear reduced thin shell element is selected;

[0014] (3.2) The initial defect distribution adopts the deformation distribution pattern of the first-order elastic modal of the component, and the amplitude of the defect is determined according to the actual measurement.

[0015] Further, the specific method for determining the grouping category in step S1 is as follows:

[0016] According to the height and width-height ratio of the steel plate shear wall, it is determined whether it belongs to the N class or the W class in the classification group. "N" and "M" are the classification of the design text conditions, "N" represents the steel plate with a height lower than 2600mm and a width-height ratio less than 1.5 or a height higher than 2600mm and a width-height ratio less than 1.2; "W" represents the steel plate with a height lower than 2600mm and a width-height ratio greater than 1.5 or a height higher than 2600mm and a width-height ratio greater than 1.2.

[0017] Further, the following processing is required before inputting the pre-trained BraceGAN-TXT in step S2:

[0018] (5.1) Use the PIL (Python Imaging Library) library in python to convert the S.Mises chart and ESEDEN chart to "PNG" format, and set the pixel to 2048p x 1024p;

[0019] (5.2) Take ESEDEN chart as train_A; take S.Mises chart as train_H.

[0020] Further, the network architecture of BraceGAN-TXT algorithm in step S2 is as follows:

[0021] (6.1) The generator accepts image and text data as input, extracts high-dimensional features of the encoded image using a convolutional neural network, and extracts high-dimensional feature representation of the text information using FastText;

[0022] (6.2) The extracted image feature data and text feature data are fused in a residual network (ResNet) to form a fused image-text feature data, and the text feature data is fused with the fused image-text feature data again to enhance the expression effect of the features;

[0023] (6.3) The target image is generated using a deconvolutional neural network, and the discriminator fuses the text feature data with the image and text data input in the generator to judge the authenticity of the generated image.

[0024] Further, the text encoding and feature extraction process in the BraceGAN-TXT algorithm in step S2 is as follows:

[0025] (7.1) Extract the keywords in the train_A and train_H picture file name;

[0026] (7.2) According to the keywords, if the picture name contains "N", a sentence list containing "N" description is generated; if the picture name contains "W", a sentence list containing "W" description is generated; the sentence list is converted into a word vector using the embedding function of FastText, and the word vector and the description length are saved to a.pth file, the file name is associated with the corresponding picture name, and the folder name is train_C;

[0027] (7.3) In the encoder, the average hidden state of the text sequence is calculated according to the text data and the text sequence length, the latent variable is predicted using a neural network, and the text feature data is obtained; the text feature data is normalized, and the dimension of the text feature data is matched with the dimension of the image feature data; the image feature data and the text feature data are spliced together and passed to the residual block network "Residual blocks" to obtain the final merged feature merge;

[0028] (7.4) Add the output feature of the encoder and the merged feature merge to get a temporary feature representation; splice the temporary feature and the text feature data together to get a spliced feature tensor; use the decoder to decode the spliced feature tensor as input to get the final output result;

[0029] (7.5) The text feature data is converted in dimension to match the input image and text data in the generator in dimension; the input image and text data in the generator and the text feature data are spliced together to form comprehensive input features, and the final output result is obtained through the discriminator model.

[0030] Further, the evaluation method and index for the BraceGAN-TXT intelligent generation design in step S2 are:

[0031] (8.1) An evaluation method and index suitable for BraceGAN-TXT, Score IoU , are proposed. IoU Score is composed of three indexes, WIoU, SIoU, as a reduction coefficient to control the number of entity units, a control coefficient representing the proportion of stiffening ribs; WIoU measures the weighted generation of the four elements in the image, evaluating the pixel-level accuracy of the generated image; SIoU is used to measure the consistency between the generated design and the target design in the design of the stiffened steel plate shear wall structure, Score IoU The evaluation index evaluates the design from two aspects of the number of stiffening ribs and the rationality of the arrangement position, Score IoU greater than 0.5 is a good design performance.

[0032] (8.2) Score IoU is calculated using formula (1), and

[0033] Score IoU =η SRratio ×(η SIoU ×SIoU+η WIoU ×WIoU) (1)

[0034]

[0035] Wherein, WIoU, SIoU are the weight intersection ratio and the stiffening rib contour intersection ratio respectively; η SIoU and η WIoU represent the weight coefficients of SIoU and WIoU respectively, both of which are 0.5; represents the total amount of steel plate shear wall stiffening ribs in the BraceGAN design drawing, represents the total amount of steel plate shear wall stiffening ribs in the target design.

[0036] (8.3) WIoU is calculated using formula (3);

[0037]

[0038] where k represents the category, i.e. 0 is background, 1 is solid unit, 2 is virtual unit, and 3 is black grid line; i represents the true value, and j represents the predicted value; P ij represents predicting i as j, which is false negative, i.e. FN; P ji represents predicting j as i, which is false positive, i.e. FP; P ii represents predicting i as i, which is true positive, i.e. TP; w i is a weight factor, i.e. w0=0, w1=0.5, w2=0.4, and w3=0.1;

[0039] (8.4) SIoU is calculated using formula (4);

[0040]

[0041] where A inter represents the intersection area of solid units in the BraceGAN design and the target design, A union represents the union area of solid units in the BraceGAN design and the target design, A union =A target +A GAN -A inter , A target represents the area of solid units in the target design, A GAN represents the area of solid units in the BraceGAN design.

[0042] Further, the specific implementation process of the optimization scheme in step S3 is as follows:

[0043] (9.1) Based on BraceGAN-TXT, the stiffening rib layout of steel plate shear wall is designed: the S.Mises diagram and ESEDEN diagram of the steel plate filled in the steel plate shear wall required for design are obtained, the resolution of the picture is converted to 2048*1024 by calling the PIL library (Python Imaging Library) in python, and the category is judged according to the classification standard. If it belongs to N category, prefix "N_" is added before the picture file name, if it belongs to W category, prefix "W_" is added before the picture file name, and the picture with modified file name is input into BraceGAN-TXT, and the stiffening rib layout is obtained by output.

[0044] (9.2) Deepening design based on stiffening rib layout: the equivalent bending stiffness method is used for deepening design of stiffening rib, considering the post-buckling strength design of stiffened steel plate shear wall and the stiffening rib is steel plate strip, the width-thickness ratio of stiffening rib meets formula (5), where λ sis the width-thickness ratio of the stiffening rib, which is the ratio of the overhanging width of the stiffening rib plate member to the thickness, the size of the stiffening rib is adjusted by using the equivalent bending stiffness method to make the cross-sectional moment of inertia of the stiffening rib consistent before and after design, and the cross-sectional moment of inertia I is calculated by formula (6) s ;

[0045] 6≤λ s ≤12 (5)

[0046]

[0047] In the formula, h s represents the overhanging width of the stiffening rib plate member, t s represents the thickness of the stiffening rib plate member;

[0048] (9.3) Construct the corresponding stiffened steel plate shear wall analysis model: model the structure by using ABAQUS finite element analysis software, and carry out performance evaluation on buckling mode, bearing capacity, stiffness, out-of-plane deformation and energy dissipation capacity of the model.

[0049] Advantages, compared with the prior art, the above technical scheme has the following beneficial technical effects:

[0050] (1) The stiffened steel plate shear wall design method based on the fusion of image-text features of the generated adversarial network provided by the present application applies the deep learning method in artificial intelligence to structure design, designs the text coding and feature extraction process, constructs a network structure capable of processing high-dimensional abstract features of images and texts, and solves the problem that the existing generated adversarial network is difficult to learn image data and text data at the same time.

[0051] (2) The improved and optimized generated adversarial network model can effectively generate a structure scheme meeting the design requirements according to the text instructions, and the design result is equivalent to the classic topological optimization design method in terms of plane design indicators and overall structure mechanical properties, which has strong practical significance. BRIEF DESCRIPTION OF DRAWINGS

[0052] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

[0053] Figure 1 It is a schematic diagram of the BraceGAN-TXT method framework in the method of the present application;

[0054] Figure 2 It is a schematic diagram of the network architecture of the BraceGAN-TXT in the method of the present application;

[0055] Figure 3 It is a standard layer plane arrangement diagram of an example structure for verifying the effectiveness of the method of the present application;

[0056] Figure 4 A steel plate shear wall planar layout diagram of an example for verifying the effectiveness of the method of the present application;

[0057] Figure 5 A steel plate shear wall elevation view of an example for verifying the effectiveness of the method of the present application;

[0058] Figure 6 A stiffening rib deepening design result schematic diagram of the BraceGAN-TXT generated design of the example for verifying the effectiveness of the method of the present application;

[0059] Figure 7 A stiffening rib deepening design result schematic diagram of the topology optimization design of the example for verifying the effectiveness of the method of the present application;

[0060] Figure 8 A flowchart of the method of the present application. DETAILED DESCRIPTION

[0061] In order to make the technical solutions, advantages and ultimate purposes of the present application clearer, the present application will be further explained and described below in combination with the accompanying drawings and specific embodiments.

[0062] As shown in Figure 8 The present application proposes a stiffened steel plate shear wall design method based on a generative adversarial network fused image-text features, which comprises the following steps:

[0063] Step S1, taking the S.Mises diagram and ESEDEN diagram of the steel plate shear wall to be designed, and determining the classification group thereof;

[0064] Step S2, inputting the S.Mises diagram and ESEDEN diagram of the steel plate shear wall to be designed into the pre-trained BraceGAN-TXT to obtain a BraceGAN-TXT intelligent generated design diagram;

[0065] Step S3, deepening design according to the intelligent generated design diagram to obtain the final stiffening rib design.

[0066] Further, in step S1, the S.Mises diagram and ESEDEN diagram manufacturing process is as follows:

[0067] (2.1) According to the steel plate shear wall structure design drawing, a corresponding steel plate shear wall model is established in the finite element analysis software ABAQUS;

[0068] (2.2) Create a field output in the analysis step module, the scope is the middle steel plate part of the steel plate shear wall, and the output variable is set to MISES, Mises equivalent stress and ELEDEN, all energy density components;

[0069] (2.3) Create and submit the Job file, run the completion and open the Job.odb file, draw the cloud chart on the non-deformed graph in the visualization module, and get the S.Mises chart and ESEDEN chart of the intermediate steel plate.

[0070] Further, the modeling method in step (2.1) is:

[0071] (3.1) In terms of finite element modeling unit selection, the intermediate embedded steel plate in the steel plate shear wall belongs to the plate shell structure, and the linear reduced thin shell element is selected;

[0072] (3.2) The initial defect distribution adopts the deformation distribution pattern of the first-order elastic modal of the component, and the amplitude of the defect is determined according to the actual measurement.

[0073] Further, the specific method for determining the grouping category in step S1 is as follows:

[0074] According to the height and width-height ratio of the steel plate shear wall, it is determined whether it belongs to the N class or the W class in the classification group. "N" and "M" are the classification of the design text conditions, "N" represents the steel plate with a height lower than 2600mm and a width-height ratio less than 1.5 or a height higher than 2600mm and a width-height ratio less than 1.2; "W" represents the steel plate with a height lower than 2600mm and a width-height ratio greater than 1.5 or a height higher than 2600mm and a width-height ratio greater than 1.2.

[0075] Further, the following processing is required before inputting the pre-trained BraceGAN-TXT in step S2:

[0076] (5.1) Use the PIL (Python Imaging Library) library in python to convert the S.Mises chart and the ESEDEN chart into "PNG" format, and set the pixel to 2048p x 1024p;

[0077] (5.2) Take the ESEDEN chart as train_A; take the S.Mises chart as train_H.

[0078] Further, the network architecture of BraceGAN-TXT algorithm in step S2 is as follows:

[0079] (6.1) The generator accepts image and text data as input, uses convolutional neural network to extract high-dimensional features of the encoded image, and uses FastText to extract high-dimensional feature representation of text information;

[0080] (6.2) The image feature data and the text feature data after extraction are fused in a residual network (ResNet) to form fused image-text feature data, and the text feature data is fused with the fused image-text feature data again to enhance the expression effect of the features;

[0081] (6.3) A target image is generated by using a deconvolutional neural network, and a discriminator fuses the text feature data with the image and text data input in a generator to judge the authenticity of the generated image.

[0082] Further, the text encoding and feature extraction process in the BraceGAN-TXT algorithm in step S2 is as follows:

[0083] (7.1) Extract the keywords in the train_A and train_H picture file names;

[0084] (7.2) According to the keywords, if the picture name contains “N”, a sentence list containing “N” description is generated; if the picture name contains “W”, a sentence list containing “W” description is generated; the sentence list is converted into a word vector by using the embedding function of FastText, and the word vector and the description length are saved to a.pth file, the file name is associated with the corresponding picture name, and the folder name is train_C;

[0085] (7.3) In the encoder, the average hidden state of the text sequence is calculated according to the text data and the text sequence length, the latent variable is predicted by using a neural network, and the text feature data is obtained; the text feature data is normalized, and the dimension of the text feature data is matched with the dimension of the image feature data; the image feature data and the text feature data are spliced together and passed to the residual block network “Residual blocks” to obtain the final merged feature merge;

[0086] (7.4) The output feature of the encoder and the merged feature merge are added to obtain a temporary feature representation; the temporary feature and the text feature data are spliced together to obtain a spliced feature tensor; the spliced feature tensor is used as input to decode by using the decoder to obtain the final output result;

[0087] (7.5) The text feature data is converted in dimension to match the image and text data input in the generator in dimension; the image and text data input in the generator and the text feature data are spliced together to form a comprehensive input feature, and the last output result is obtained through the discriminator model.

[0088] Further, the evaluation method and index for the BraceGAN-TXT intelligent generation design in step S2 are:

[0089] (8.1) The evaluation method and index suitable for BraceGAN-TXT are proposed, i.e. Score IoU , and Score IoU is composed of three indexes η SRratio , WIoU, SIoU, η SRratio is a reduction coefficient to control the number of entity units, and is a control coefficient representing the proportion of stiffening ribs; WIoU measures the weighted generation of four elements in the image, and evaluates the pixel-level accuracy of the generated image; SIoU is used to measure the consistency between the generated design and the target design in the design of the stiffened steel plate shear wall structure, and Score IoU The evaluation index evaluates the design from two aspects of the number of stiffening ribs and the rationality of the arrangement position, and Score IoU is greater than 0.5, which is a good design performance.

[0090] (8.2) Score IoU is calculated using formula (1), and Score

[0091] Score IoU = η SRratio ×(η SIoU × SIoU + η WIoU × WIoU) (1)

[0092]

[0093] wherein, WIoU and SIoU are the weight intersection ratio and the stiffening rib contour intersection ratio, respectively; η SIoU and η WIoU represent the weight coefficients of SIoU and WIoU, both of which are 0.5; represents the total amount of stiffening ribs in the BraceGAN design drawing, represents the total amount of stiffening ribs in the target design;

[0094] (8.3) WIoU is calculated using formula (3);

[0095]

[0096] wherein, k represents the category, i.e. 0 is the background, 1 is the entity unit, 2 is the virtual unit, and 3 is the black grid line; i represents the true value, and j represents the predicted value; P ij represents that i is predicted as j, which is false negative, i.e. FN; P ji represents that j is predicted as i, which is false positive, i.e. FP; P ii represents that i is predicted as i, which is true, i.e. TP; wi is a weight factor, i.e., w0=0, w1=0.5, w2=0.4, w3=0.1;

[0097] (8.4) SIoU is calculated using formula (4);

[0098]

[0099] In the formula, A inter represents the intersection area of the entity unit in the BraceGAN design drawing and the target design drawing, A union represents the union area of the entity unit in the BraceGAN design drawing and the target design drawing, A union =A target +A GAN -A inter , A target represents the area of the entity unit in the target design drawing, A GAN represents the area of the entity unit in the BraceGAN design drawing.

[0100] Further, the specific implementation process of the optimization scheme in step S3 is as follows:

[0101] (9.1) Based on BraceGAN-TXT, the stiffening rib plane layout of the steel plate shear wall is designed: the S.Mises and ESEDEN diagrams of the steel plate filled in the steel plate shear wall required by the design are obtained, the resolution of the picture is converted to 2048*1024 by calling the PIL library (Python Imaging Library) in python, and the category is judged according to the classification standard. If it belongs to N category, prefix "N_" is added before the picture file name, if it belongs to W category, prefix "W_" is added before the picture file name, and the picture with modified file name is input into BraceGAN-TXT, and the stiffening rib layout diagram is obtained by output.

[0102] (9.2) Deepening design based on stiffening rib layout diagram: the equivalent bending stiffness method is used for deepening design of stiffening rib, considering the post-buckling strength design of stiffened steel plate shear wall and the stiffening rib is steel plate, the width-thickness ratio of stiffening rib meets the formula (5), in which, λ s is the width-thickness ratio of stiffening rib, the size of stiffening rib is adjusted by using the equivalent bending stiffness method, so that the sectional moment of inertia of stiffening rib before and after design remains unchanged, and the sectional moment of inertia I s is calculated by formula (6).

[0103] 6≤λ s ≤12 (5)

[0104]

[0105] wherein h s represents the overhanging width of the stiffened rib plate member, t s represents the thickness of the stiffened rib plate member;

[0106] (9.3) Constructing the corresponding stiffened steel plate shear wall analysis model: using ABAQUS finite element analysis software to model the structure, and carrying out performance evaluation on buckling mode, bearing capacity, stiffness, out-of-plane deformation and energy dissipation capacity of the model.

[0107] In order to verify the performance of the present application, the examples in the steel plate shear wall structure design in the national building standard design drawing set are selected, and the stiffened ribs are designed by applying the present application to the actual engineering. Figure 3 is a structural standard layer plan layout, Figure 4 is a steel plate shear wall plan layout, Figure 5 is a steel plate shear wall elevation, Figure 6 is the stiffened rib deepening design result generated by BraceGAN-TXT, Figure 7 is the stiffened rib deepening design result of topological optimization design, Table 1 is the performance quantitative analysis of the example BraceGAN-TXT intelligent generation design, and Table 2 is the comparison of the mechanical properties of BraceGAN-TXT generation design and topological optimization design. According to Table 1 and Table 2, the mechanical performance difference of both aspects is within 10%. Through intelligent plane and overall structure design analysis of the design case of the stiffened steel plate shear wall, it is found that the algorithm can effectively generate a structure scheme according to the text instructions. The design result is equivalent to the classic topological optimization design method in terms of plane design index and overall structure mechanical performance, which further verifies the effectiveness of the stiffened steel plate shear wall design method based on the generation of adversarial network and the fusion of image-text features proposed in the present application.

[0108] Table 1 Performance quantitative analysis of example BraceGAN-TXT intelligent generation design

[0109]

[0110] Table 2 Comparison of mechanical properties of BraceGAN-TXT generation design and topological optimization design

[0111]

[0112]

[0113] The specific embodiments described above are only a preferred scheme of the present application, but they are not used to limit the present application. Any skilled person in the art can easily think of alternative or equivalent conversion methods within the technical scope disclosed by the present application, which should be covered by the protection scope of the present application.

Claims

1. A method for designing a reinforced steel plate shear wall based on a fusion of image-text features using a generative adversarial network, the method comprising: The method comprises the following steps: Step S1, taking the S.Mises diagram and ESEDEN diagram of the steel plate shear wall to be designed, and determining the classification group; Step S2, inputting the S.Mises diagram and ESEDEN diagram of the steel plate shear wall to be designed into the pre-trained BraceGAN-TXT to obtain an intelligent design diagram generated by BraceGAN-TXT; Step S3, deepening the design according to the intelligent design diagram to obtain the final stiffening rib design; The specific method for determining the classification group in step S1 is as follows: According to the height and width-height ratio of the steel plate shear wall, it is determined whether it belongs to the N class or the W class in the classification group, and "N" and "M" are classifications of design text conditions, "N" represents a steel plate with a height lower than 2600 mm and a width-height ratio less than 1.5 or a height higher than 2600 mm and a width-height ratio less than 1.2; "W" represents a steel plate with a height lower than 2600 mm and a width-height ratio greater than 1.5 or a height higher than 2600 mm and a width-height ratio greater than 1.2; In step S2, the following processing is required before inputting into the pre-trained BraceGAN-TXT: (5.1) using the PIL (Python Imaging Library) library in python, converting the S.Mises diagram and ESEDEN diagram into "PNG" format, and setting the pixel to 2048p x 1024p; (5.2) taking the ESEDEN diagram as train_A; taking the S.Mises diagram as train_H; The network architecture of the BraceGAN-TXT algorithm in step S2 is as follows: (6.1) the generator accepts image and text data as input, adopts convolutional neural network to extract high-dimensional features of the encoded image, and uses FastText to extract high-dimensional feature representation of the text information; (6.2) the extracted image feature data and text feature data are fused in the residual network (ResNet) to form a fused image-text feature data, and the text feature data and the fused image-text feature data are fused again to enhance the expression effect of the features; (6.3) using the deconvolutional neural network to generate the target image, the discriminator fuses the text feature data with the image and text data input into the generator to judge the true or false of the generated image; The text encoding and feature extraction process in the BraceGAN-TXT algorithm in step S2 is as follows: (7.1) extracting the keywords in the train_A and train_H picture file name; (7.2) according to the keywords, if the picture name contains "N", a sentence list containing "N" description is generated; if the picture name contains "W", a sentence list containing "W" description is generated; the embedding function of FastText is used to convert the sentence list into a word vector, and the word vector and the description length are saved to a.pth file, the file name is associated with the corresponding picture name, and the folder name is train_C; (7.3) In the encoder, the average hidden state of the text sequence is calculated according to the text data and the length of the text sequence, the latent variable is predicted by using a neural network, and the text feature data is obtained; the text feature data is normalized, and the dimension of the text feature data is matched with the dimension of the image feature data; the image feature data and the text feature data are spliced together and passed to the residual block network "Residual blocks", and the final merged feature merge is obtained; (7.4) The output feature of the encoder and the merged feature merge are added to obtain a temporary feature representation; the temporary feature and the text feature data are spliced together to obtain a spliced feature tensor; the spliced feature tensor is used as input to decode by the decoder to obtain the final output result; (7.5) The text feature data is converted in dimension to match the image and text data input in the generator in dimension; the image and text data input in the generator and the text feature data are spliced together to form a comprehensive input feature, and the final output result is obtained through the discriminator model.

2. The method of claim 1, wherein the method is based on a generative adversarial network (GAN) that fuses image-text features for the design of the steel plate shear wall. In step S1, the S.Mises diagram and the ESEDEN diagram are made as follows: (2.1) According to the steel plate shear wall structure design drawing, a corresponding steel plate shear wall model is established in the finite element analysis software ABAQUS; (2.2) Create a field output in the analysis step module, the scope is the middle steel plate part of the steel plate shear wall, and the output variable is set to MISES, Mises equivalent stress and ELEDEN, all energy density components; (2.3) Create and submit a Job file, run it and open the Job.odb file, draw a cloud chart on the undeformed diagram in the visualization module to obtain the S.Mises diagram and the ESEDEN diagram of the middle steel plate.

3. The method of claim 2, wherein the method is based on a generative adversarial network (GAN) that fuses image-text features for the design of the steel plate shear wall. The modeling method in step (2.1) is as follows: (3.1) In terms of finite element modeling unit selection, the middle embedded steel plate of the steel plate shear wall belongs to a plate shell structure, so a linear reduced thin shell element is selected; (3.2) The initial defect distribution adopts the deformation distribution pattern of the first-order elastic modal of the component, and the amplitude of the defect is determined according to the actual measurement.

4. The method of claim 1, wherein the method is based on a generative adversarial network (GAN) that fuses image-text features for the design of the steel plate shear wall. The evaluation method and index for the BraceGAN-TXT intelligent generation design in step S2 are as follows: (8.1) The evaluation method and index suitable for BraceGAN-TXT are proposed, namely Score IoU , and Score IoU is composed of η SRratio , Wiou, SIoU three indexes, η SRratio as a reduction coefficient to control the number of entity units, and the control coefficient representing the proportion of stiffening ribs; Wiou measures the weighted generation of four elements in the image, and evaluates the pixel-level accuracy of the generated image; SIoU is used to measure the consistency of the generated design and the target design in the design of the stiffened steel plate shear wall structure, and Score IoU The evaluation index evaluates the design from two aspects of the number and reasonable arrangement position of the stiffening rib design, and Score IoU greater than 0.5 is a good design performance; (8.2) Score is calculated using formula (1) IoU η is calculated using formula (2) SRratio ; Score IoU = η SRratio × (η SIoU × SIoU + η WIoU × WIoU) (1) wherein, WloU, SIoU are the weighted intersection over union, the stiffening rib contour intersection over union respectively; η SIoU and η WIoU represent the weight coefficients of SIoU and WIoU respectively, both of which are 0.5; represents the total amount of stiffening ribs in the BraceGAN design drawing of the steel plate shear wall, represents the total amount of stiffening ribs in the target design of the steel plate shear wall. (8.3) Calculate WIoU using formula (3); where k represents the category, i.e. 0 class is background, 1 class is solid unit, 2 class is virtual unit, and 3 class is black grid line; i represents the true value, and j represents the predicted value; P ij represents predicting i as j, which is false negative, i.e. FN; P ji represents predicting j as i, which is false positive, i.e. FP; P ii represents predicting i as i, which is true positive, i.e. TP; w i is a weight factor, i.e. w0=0, w1=0.5, w2=0.4, and w3=0.1; (8.4) Calculate SIoU using formula (4); where A inter denotes the intersection area of the entity units in the BraceGAN design and the target design, A union denotes the union area of the entity units in the BraceGAN design and the target design, A union = A target + A GAN - A inter , A target denotes the area of the entity units in the target design, A GAN denotes the area of the entity units in the BraceGAN design. 5.The method of claim 1, wherein the method is characterized by, The specific implementation process of the optimization scheme deepening design in step S3 is as follows: (9.1) Steel plate shear wall stiffening rib layout design based on BraceGAN-TXT: Obtain the S. Mises and ESEDEN figures of the steel plate filled in the steel plate shear wall required for design, call the PIL library (Python Imaging Library) in python to convert the resolution of the picture to 2048*1024, and judge the category according to the classification standard. If it belongs to N category, add the prefix "N_" before the picture file name, if it belongs to W category, add the prefix "W_" before the picture file name, input the picture with modified file name into BraceGAN-TXT, and output the stiffening rib layout figure; (9.2) Deepening design based on stiffener layout: the equivalent bending stiffness method is used for deepening design of the stiffener, the post-buckling strength design of the stiffened steel plate shear wall is considered, and when the stiffener is a steel strip, the width-thickness ratio of the stiffener meets the requirements of formula (5), wherein λ s is the width-thickness ratio of the stiffener, and the width and thickness of the stiffened plate member are used. The size of the stiffener is adjusted by the equivalent bending stiffness method so that the sectional moment of inertia of the stiffener before and after design remains unchanged, and the sectional moment of inertia I s is calculated by formula (6). 6 < λ s ≤ 12 (5) wherein h s represents the overhanging width of the stiffened panel member, t s represents the thickness of the stiffened panel member; (9.3) Construction of corresponding stiffened steel plate shear wall analysis model: model the structure by using ABAQUS finite element analysis software, and carry out performance evaluation on buckling mode, bearing capacity, stiffness, out-of-plane deformation and energy dissipation capacity of the model.

Images