An intelligent design method for stainless steel structural components based on generative adversarial networks

Abstract

The present invention relates to a steel structure member, specifically an intelligent design method for stainless steel structure members based on a generative adversarial network, belonging to the field of steel structure applications. A feature mapping relationship is established between the design parameters and design schemes of stainless steel structure members, making full use of the strain hardening ability, plastic deformation ability of stainless steel materials, and the post-buckling bearing capacity of members, and finally providing a relatively accurate and economical design scheme for stainless steel structure members, such as the member cross-section type, cross-section size, etc. The present invention can more accurately consider the material properties of stainless steel, and on the premise of ensuring safety and reliability, achieve the refined design of the cross-section of stainless steel members, reducing material waste caused by low design accuracy.

Description

Technical Field

[0001] The present invention relates to a steel structure member, specifically an intelligent design method for stainless steel structure members based on a generative adversarial network, belonging to the field of steel structure applications. Background Art

[0002] Due to its strong corrosion resistance, easy maintenance, environmental friendliness and other advantages, stainless steel structures have become a new trend in engineering structure anti-corrosion. However, their high cost limits their large-scale application in the engineering field. As a typical non-linear material, stainless steel often exhibits obvious buckling mode coupling phenomena when the member fails. In addition, the stress-strain curve of stainless steel has no obvious yield point and yield platform. In structural design, the stress f corresponding to a plastic strain of 0.2% is usually taken as its nominal yield strength, which is much lower than the ultimate tensile strength of the material. Therefore, conventional structural design is difficult to fully consider the considerable strain hardening ability, plastic deformation ability of stainless steel and the buckling mode coupling phenomenon of members, often resulting in material waste and increased cost. 0.2 As its nominal yield strength, this strength is much lower than the ultimate tensile strength of the material. Therefore, conventional structural design is difficult to fully consider the considerable strain hardening ability, plastic deformation ability of stainless steel and the buckling mode coupling phenomenon of members, often resulting in material waste and increased cost. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention provides a method for establishing a characteristic mapping relationship between the design parameters and design schemes of stainless steel structure members, making full use of the strain hardening ability, plastic deformation ability of stainless steel materials and the post-buckling bearing capacity of members, and finally providing a relatively accurate and economical design scheme for stainless steel structure members, such as member section types, section sizes, etc. The present invention can more accurately consider the material properties of stainless steel, and on the premise of ensuring safety and reliability, realize the refined design of stainless steel member sections, and reduce the material waste caused by low design accuracy. An intelligent design method for stainless steel structure members based on a generative adversarial network.

[0004] To achieve the above objectives, the present invention is realized through the following technical solutions:

[0005] An intelligent design method for stainless steel structure members based on a generative adversarial network is carried out according to the following steps:

[0006] (1) Collect the experimental data of existing stainless steel structure members, and extract the key parameters related to load types, member lengths and restraint conditions, stainless steel grades, section types and sizes, member bearing capacities, and buckling failure modes;

[0007] (2) Based on the experimental data, use the finite element software Abaqus to establish a finite element analysis model of stainless steel structure members;

[0008] (3) Using the finite element analysis model of stainless steel structural members, parametric analysis is carried out on various factors affecting their bearing capacity, such as common stainless steel grades of stainless steel structural members: S304 and S316, section types: I-shaped section, square and rectangular tubes, and circular tubes, slenderness ratio, and section dimensions.

[0009] (4) Summarize and organize the results of the parametric analysis, and establish a generative adversarial network dataset of design requirements related to different load types, load levels, member lengths, required material grades, and deflection limits, and the corresponding optimal sections.

[0010] (5) Build a generative adversarial network model suitable for the section optimization design of stainless steel structural members.

[0011] (6) Train the generative adversarial network model, and realize the intelligent design of stainless steel structural members based on the trained generative adversarial network model; the design scheme of stainless steel structural members mainly includes the section type and section dimensions of the designed section.

[0012] Preferably, in step (6), the expression in the training process of the generative adversarial network model is:

[0013] Y = G(X)

[0014] minZ(Y)

[0015] N Ed or M Ed ≤N Rd or M Rd

[0016] ω Ed ≥ω Rd

[0017] Among them, X is the design requirements of stainless steel structural members, mainly including load type, load level, member length, required material grade, deflection limit, etc.; Y is the optimal section type and section dimensions of stainless steel structural members that meet the design requirements output by the generative network model; Z is the loss function of the intelligent design generative adversarial network model defined for stainless steel structural members; N Ed and ω Ed are the bearing capacity and allowable deflection required for the design of stainless steel structural members respectively; N Rd and ω Rd are the ultimate load and maximum deflection of the optimal solution given by the generative model respectively.

[0018] This method can fully consider the strain hardening ability of stainless steel materials, the plastic deformation ability, and the post-buckling bearing capacity of components, and provide a reasonable design scheme. The generative adversarial network in this method mainly includes two parts: a generative network and a discriminative network. Among them, the generative network is used to establish the mapping relationship between the design requirements of stainless steel structural components in the input layer and the optimal cross-section design scheme in the output layer. The discriminative network is used to identify the cross-section design scheme of stainless steel structural components output by the generative network, optimize the model parameters of the generative network, and improve the accuracy and reliability of the generative network in generating the optimal cross-section of stainless steel structural components.

[0019] Stainless steel materials have strong non-linearity and obvious coupling phenomena of buckling modes when components fail. Therefore, it is difficult for traditional design methods to reasonably and fully utilize the bearing performance of stainless steel structural components. The present invention introduces a generative adversarial network model, which is constructed based on a dataset of a large amount of experimental data and numerical simulation analysis results, can more accurately consider the material properties of stainless steel, and intelligently output an optimal cross-section design scheme that meets the requirements according to the design requirements of the input stainless steel structural components.

[0020] Fully consider the strain hardening ability of stainless steel materials, the plastic deformation ability, and the post-buckling bearing capacity of components, and provide the optimal cross-section design scheme. Compared with the existing traditional design methods, it can improve the design accuracy, make full use of the mechanical properties of stainless steel materials, and reduce the waste of stainless steel materials caused by low design accuracy. Description of the Drawings

[0021] Figure 1 It is a training scheme diagram of the generative adversarial network of the present invention. Detailed Embodiments

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] Embodiment: As shown in the figure, an intelligent design method for stainless steel structural components based on a generative adversarial network is carried out according to the following steps:

[0024] (1). Collect the experimental data of existing stainless steel structural components, and extract the key parameters related to load types, component lengths and restraint conditions, stainless steel grades, cross-section types and sizes, component bearing capacities, and buckling failure modes;

[0025] (2). Based on the experimental data, use the finite element software Abaqus to establish a finite element analysis model of stainless steel structural components;

[0026] (3) Using the finite element analysis model of stainless steel structural members, parametric analysis is carried out on various factors affecting their bearing capacity, such as common stainless steel grades of stainless steel structural members: S304 and S316, cross-section types: I-shaped cross-section, square and rectangular tubes, and circular tubes, slenderness ratio, and cross-section dimensions.

[0027] (4) Summarize and organize the results of the parametric analysis, and establish a generative adversarial network dataset of design requirements related to different load types, load levels, member lengths, required material grades, and deflection limits and the corresponding optimal cross-sections.

[0028] (5) Build a generative adversarial network model suitable for the cross-section optimization design of stainless steel structural members.

[0029] (6) Train the generative adversarial network model, and realize the intelligent design of stainless steel structural members based on the trained generative adversarial network model; the design scheme of stainless steel structural members mainly includes the cross-section type and cross-section dimensions of the designed cross-section.

[0030] In step (6), the expression of the training process of the generative adversarial network model is:

[0031] Y = G(X)

[0032] min Z(Y)

[0033] N Ed or M Ed ≤ N Rd or M Rd

[0034] ω Ed ≥ ω Rd

[0035] Among them, X is the design requirements of stainless steel structural members, mainly including load type, load level, member length, required material grade, deflection limit, etc.; Y is the optimal cross-section type and cross-section dimensions of stainless steel structural members that meet the design requirements output by the generative network model; Z is the loss function of the intelligent design generative adversarial network model defined for stainless steel structural members; N Ed and ω Ed are the bearing capacity and allowable deflection required for the design of stainless steel structural members, respectively; N Rd and ω Rd are the ultimate load and maximum deflection of the optimal scheme given by the generative model, respectively.

Claims

1. An intelligent design method for stainless steel structural members based on a generative adversarial network, characterized in that The steps are as follows: (1) Collect the experimental data of existing stainless steel structural members, and extract the load type, member length and constraint conditions, stainless steel grade, section type and size, member bearing capacity, and buckling failure mode parameters; (2) Based on the experimental data, establish a finite element analysis model of stainless steel structural members using the finite element software Abaqus; (3) Use the finite element analysis model of stainless steel structural members to perform parametric analysis on the factors that affect the bearing capacity of common stainless steel grades: S304 and S316, section types: I-shaped section, square and rectangular tubes, and circular tubes, slenderness ratio, and section size of stainless steel structural members; (4) Summarize and organize the results of the parametric analysis, and establish an adversarial generation network dataset of relevant design requirements and corresponding optimal sections for different load types, load levels, member lengths, stainless steel grades, and deflection limits; (5) Build an adversarial generation network model suitable for the cross-section optimization design of stainless steel structural members; (6) Train the generative adversarial network model, and realize the intelligent design of stainless steel structural members based on the trained generative adversarial network model; The design scheme of the stainless steel structural member includes the section type and section size of the designed section; In step (6), the expression of the training process of the generative adversarial network model is: N Ed ≤ N Rd M Ed ≤ M Rd Among them, X is the design requirement of the stainless - steel structural member, including load type, load level, member length, stainless - steel grade, and deflection limit; Y is the optimal cross - section type and cross - section size of the stainless - steel structural member that meet the design requirements output by the generation network model; Z is the loss function of the intelligent design adversarial generation network model defined for the stainless - steel structural member. N Ed and ω Ed are the bearing capacity and allowable deflection required for the design of the stainless - steel structural member, respectively; N Rd and ω Rd are the ultimate load and maximum deflection of the optimal solution given by the generation model, respectively.

Images