CN113158514A - Automobile body material structure matching lightweight design method, system and storage medium - Google Patents

Automobile body material structure matching lightweight design method, system and storage medium Download PDF

Info

Publication number
CN113158514A
CN113158514A CN202110345933.4A CN202110345933A CN113158514A CN 113158514 A CN113158514 A CN 113158514A CN 202110345933 A CN202110345933 A CN 202110345933A CN 113158514 A CN113158514 A CN 113158514A
Authority
CN
China
Prior art keywords
design
parameters
parameter
design parameters
thickness
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110345933.4A
Other languages
Chinese (zh)
Other versions
CN113158514B (en
Inventor
王勇
刘角
孙光永
庞通
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hunan University
Central South University
Original Assignee
Hunan University
Central South University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hunan University, Central South University filed Critical Hunan University
Priority to CN202110345933.4A priority Critical patent/CN113158514B/en
Publication of CN113158514A publication Critical patent/CN113158514A/en
Application granted granted Critical
Publication of CN113158514B publication Critical patent/CN113158514B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/15—Vehicle, aircraft or watercraft design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/08—Probabilistic or stochastic CAD

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Automation & Control Theory (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明公开了一种汽车车身材料结构匹配轻量化设计方法、系统及存储介质，以汽车车身薄壁结构厚度和材料牌号作为设计参数，达到以B柱侵入量、B柱侵入速度、B柱比吸能或汽车车身结构重量为设计目标，达到在保证汽车耐撞性不下降的同时降低汽车重量的目的。

Figure 202110345933

The invention discloses a lightweight design method, system and storage medium for matching the material structure of an automobile body. The thickness and material grade of the thin-walled structure of the automobile body are used as design parameters to achieve the B-pillar intrusion amount, the B-pillar intrusion speed, and the B-pillar ratio. Energy absorption or the weight of the car body structure is the design goal to achieve the purpose of reducing the weight of the car while ensuring that the crashworthiness of the car does not decrease.

Figure 202110345933

Description

汽车车身材料结构匹配轻量化设计方法、系统及存储介质Automobile body material structure matching lightweight design method, system and storage medium

技术领域technical field

本发明涉及汽车结构设计领域，特别是一种汽车车身材料结构匹配轻量化设计方法、系统及存储介质。The invention relates to the field of automobile structure design, in particular to a lightweight design method, system and storage medium for matching the structure of an automobile body material.

背景技术Background technique

汽车工业是当代工业领域的重要支柱产业之一，随着汽车工业的快速发展，其所带来的环境污染、能源短缺和交通安全问题也日益凸显。环保、节能和安全已成为目前汽车工业发展所面临的主要问题，而这些问题都与汽车轻量化密切相关。世界铝业协会研究表明：当汽车自重每减10％，燃油消耗以及排放可分别降低6％～8％与5％～6％；油耗每减少1升，二氧化碳排放量就会相应减少2.45kg。从安全性的角度来说，对汽车进行减重能够有效减少刹车距离，从而降低事故发生率。无论是从节能减排还是从安全性来说，汽车的轻量化都将永无止境，与之相关的材料、成形以及设计方面的研究将一直是汽车行业中的前沿和热点。The automobile industry is one of the important pillar industries in the contemporary industrial field. With the rapid development of the automobile industry, the problems of environmental pollution, energy shortage and traffic safety brought about by it have become increasingly prominent. Environmental protection, energy saving and safety have become the main problems facing the development of the automobile industry, and these problems are closely related to the lightweight of automobiles. The research of the World Aluminum Association shows that when the weight of the car is reduced by 10%, the fuel consumption and emissions can be reduced by 6% to 8% and 5% to 6% respectively; for every 1 liter of fuel consumption, carbon dioxide emissions will be reduced by 2.45kg. From a safety point of view, reducing the weight of the car can effectively reduce the braking distance, thereby reducing the accident rate. Whether in terms of energy saving and emission reduction or safety, the lightweight of automobiles will never end, and the research on materials, forming and design related to it will always be the frontier and hotspot in the automotive industry.

汽车轻量化设计指的是在保证汽车耐撞性、安全性、稳定性、平顺性等性能指标不降低且造价不升高的情况下，合理减轻汽车重量。实现轻量化的一个途径是从汽车结构入手，通过多学科优化等先进的设计手段，优化汽车部件的拓扑结构、尺寸与形状参数，将零件复合化、薄壁化和中空化等。The lightweight design of automobiles refers to the reasonable reduction of the weight of automobiles under the condition that the performance indicators such as crashworthiness, safety, stability, and ride comfort of the automobile are not reduced and the cost is not increased. One way to achieve light weight is to start with the structure of the vehicle, optimize the topology, size and shape parameters of the auto parts through advanced design methods such as multidisciplinary optimization, and make the parts composite, thin-walled and hollow.

事实上，为了保证汽车车身结构的耐撞性并减轻其重量，不仅可以考虑对汽车车身结构中不同部件的尺寸参数进行优化，还可以通过为不同的部件合理分配不同的材料来达到这一目的。例如，在对汽车车侧身进行轻量化设计时，为了保证乘客的安全，通常希望在侧面碰撞发生时，汽车B柱的侵入量和侵入速度能够尽量小。从优化的角度来说，该问题具有两个特点：1)同时包含连续变量(即薄壁结构厚度)和离散变量(即部件材料)；2)汽车B柱的侵入量和侵入速度这两个性能指标没有具体的数学表达式，只能通过有限元分析或物理实验获得。因此，这种同时考虑部件尺寸和部件材料的设计问题通常被称为车身多材料结构匹配设计。In fact, in order to ensure the crashworthiness of the car body structure and reduce its weight, it is not only possible to consider optimizing the size parameters of different components in the car body structure, but also to reasonably allocate different materials to different components to achieve this goal. . For example, in the lightweight design of the side body of an automobile, in order to ensure the safety of passengers, it is usually hoped that the intrusion amount and intrusion speed of the B-pillar of the automobile can be as small as possible when a side collision occurs. From an optimization point of view, the problem has two characteristics: 1) it contains both continuous variables (ie, the thickness of the thin-walled structure) and discrete variables (ie, the component material); 2) the intrusion volume and intrusion speed of the automobile B-pillar are both The performance index has no specific mathematical expression and can only be obtained through finite element analysis or physical experiments. Therefore, this design problem that considers both component size and component material is often referred to as body multi-material structure matching design.

在对汽车进行轻量化设计时，通常将整个设计过程描述为一个优化问题，并通过求解该优化问题来实现具体的轻量化设计。在所描述的优化问题中，一些部件的物理性能，例如碰撞吸能、碰撞峰值力、部件质量等，将被作为性能指标来衡量汽车的耐撞性、安全性等性能，而部件尺寸、结构、材料等参数则被视为设计参数。通过调整设计参数，最大/最小化这些性能指标或让这些性能指标满足一定的设计需求，即可达到不降低汽车各项性能的同时实现汽车轻量化。而车身多材料结构匹配设计问题则主要包含以下三个特点：In the lightweight design of automobiles, the entire design process is usually described as an optimization problem, and the specific lightweight design is achieved by solving the optimization problem. In the described optimization problem, the physical properties of some components, such as collision energy absorption, collision peak force, component quality, etc., will be used as performance indicators to measure the car's crashworthiness, safety and other performance, while component size, structure , material and other parameters are regarded as design parameters. By adjusting the design parameters, maximizing/minimizing these performance indicators or allowing these performance indicators to meet certain design requirements, it is possible to achieve a lightweight car without reducing the performance of the car. The multi-material structure matching design problem of the car body mainly includes the following three characteristics:

·黑盒：汽车轻量化所描述的优化问题通常不具备一个显示的表达式。这意味着，在求解这类优化问题时，往往只能知道对应于一组参数的目标函数响应，而无法获得诸如梯度、二阶导数等数学特性。Black box: The optimization problem described by vehicle lightweighting usually does not have an explicit expression. This means that when solving such optimization problems, often only the response of the objective function corresponding to a set of parameters is known, but mathematical properties such as gradients and second derivatives cannot be obtained.

·昂贵：对于很多性能指标，只能通过有限元分析等仿真工具或实际物理实验才能获得其具体的数值。这一过程将会消耗大量的时间和财力。因此，大量多次地对一些性能指标进行评价是不现实的。Expensive: For many performance indicators, their specific values can only be obtained through simulation tools such as finite element analysis or actual physical experiments. This process will consume a lot of time and money. Therefore, it is unrealistic to evaluate some performance indicators many times.

·包含两种及以上的变量类型：汽车车身多材料结构匹配设计所描述的优化问题可能同时包含诸如部件尺寸这种连续变量和诸如材料选择这种离散变量。• Inclusion of two or more types of variables: Matching design of multi-material structures for automotive bodies The described optimization problem may contain both continuous variables such as component size and discrete variables such as material selection.

值得注意的是，车身多材料结构匹配设计问题具有黑盒和昂贵两个特性。这两个特性使得我们无法对汽车部件的一些性能指标进行大量评价，也无法使用传统的基于梯度的优化方法。为了有效处理带有黑盒和昂贵特性的汽车结构设计问题，在过去的十几年里，许多基于代理模型的优化算法被提出。然而，大部分方法都针对于仅包含连续设计参数的设计问题，针对于诸如车身多材料结构匹配设计这类同时包含多种变量类型的方法比较少见。现有的针对于包含多种变量类型的方法，仅采用单一种类的代理模型。这种方式往往难以有效的近似带有多种变量类型的目标函数及约束条件，进而难以有效的引导算法找到高质量的最优解。It is worth noting that the design problem of body multi-material structure matching has both black-box and expensive properties. These two properties make it impossible to perform extensive evaluations on some performance metrics of automotive components, nor to use traditional gradient-based optimization methods. In order to effectively deal with the problem of vehicle structure design with black-box and expensive properties, many optimization algorithms based on surrogate models have been proposed in the past decade. However, most of the methods are aimed at design problems involving only continuous design parameters, and it is relatively rare to address methods that include multiple types of variables at the same time, such as the matching design of multi-material structures of car bodies. Existing methods for involving multiple variable types employ only a single kind of surrogate model. This method is often difficult to effectively approximate the objective function and constraints with multiple variable types, and it is difficult to effectively guide the algorithm to find a high-quality optimal solution.

发明内容SUMMARY OF THE INVENTION

本发明所要解决的技术问题是，针对现有技术不足，提供一种汽车车身材料结构匹配轻量化设计方法、系统及存储介质，对汽车车身结构进行优化，达到降低汽车重量的目的。The technical problem to be solved by the present invention is to provide a lightweight design method, system and storage medium for matching the material structure of an automobile body to optimize the structure of the automobile body to reduce the weight of the automobile.

为解决上述技术问题，本发明所采用的技术方案是：一种汽车车身多材料结构匹配轻量化设计方法，包括以下步骤：In order to solve the above-mentioned technical problems, the technical solution adopted in the present invention is: a lightweight design method for multi-material structure matching of an automobile body, comprising the following steps:

S1、建立并初步求解以下优化问题：S1. Establish and initially solve the following optimization problem:

min:f(x)min:f(x)

s.t.c1(x)≤T1 stc 1 (x)≤T 1

c2(x)≤T2 c 2 (x)≤T 2

c3(x)≤T3 c 3 (x)≤T 3

x＝[xthick,xmat]x＝[x thick ,x mat ]

Figure BDA0003000812800000031
Figure BDA0003000812800000031

Figure BDA0003000812800000032
Figure BDA0003000812800000032

Figure BDA0003000812800000033
Figure BDA0003000812800000033

Figure BDA0003000812800000034
Figure BDA0003000812800000034

其中，x为设计参数，xthick为由厚度设计参数组成的向量，xmat为由材料设计参数组成的向量，f(x)为目标函数，c1(x)为第一个约束条件，c2(x)为第二个约束条件，c3(x)为第三个约束条件，T1，T2和T3分别为三个约束条件需要满足的指标，Li和Ui为第i个厚度设计参数的下限和上限，M1～Mp分别代表p种不同的材料，n1为待优化的厚度参数数目，n2为待优化的材料牌号参数数目；where x is the design parameter, x thick is the vector composed of the thickness design parameters, x mat is the vector composed of the material design parameters, f(x) is the objective function, c 1 (x) is the first constraint, c 2 (x) is the second constraint, c 3 (x) is the third constraint, T 1 , T 2 and T 3 are the indicators that the three constraints need to be satisfied, respectively, Li and U i are the i -th The lower and upper limits of the thickness design parameters, M 1 to M p represent p different materials respectively, n 1 is the number of thickness parameters to be optimized, and n 2 is the number of material grade parameters to be optimized;

S2、将上述优化问题转化为以下问题：S2. Transform the above optimization problem into the following problem:

Figure BDA0003000812800000035
Figure BDA0003000812800000035

Figure BDA0003000812800000036
Figure BDA0003000812800000036

Figure BDA0003000812800000037
Figure BDA0003000812800000037

Figure BDA0003000812800000038
Figure BDA0003000812800000038

Figure BDA0003000812800000039
Figure BDA0003000812800000039

Figure BDA00030008128000000310
Figure BDA00030008128000000310

S3、使用序列二次规划求解步骤S2的问题，获得解

Figure BDA00030008128000000311
该解代表由局部搜索给出的最优薄壁结构厚度设计参数，随后将
Figure BDA00030008128000000312
和
Figure BDA00030008128000000313
组合成为一个新的解
Figure BDA00030008128000000314
该解为由局部搜索给出的最优设计参数。S3. Use sequential quadratic programming to solve the problem of step S2, and obtain the solution
Figure BDA00030008128000000311
This solution represents the optimal thin-wall structure thickness design parameters given by a local search, which is then
Figure BDA00030008128000000312
and
Figure BDA00030008128000000313
combined into a new solution
Figure BDA00030008128000000314
The solution is the optimal design parameters given by the local search.

本发明的优势在于：The advantages of the present invention are:

S1着重于同时获取连续变量和离散变量的最优解。求解S1中所建立的优化问题，可以快速定位到最优解所在的区域，为后续优化提供质量较高质量的初始解。该初始解能够为后续优化过程提供一个较为优异的轻量化设计方案，在尽可能的满足设计指标的条件下获得较轻的车身轻量化设计方案。S1 focuses on obtaining optimal solutions for both continuous and discrete variables at the same time. Solving the optimization problem established in S1 can quickly locate the region where the optimal solution is located, and provide a high-quality initial solution for subsequent optimization. The initial solution can provide an excellent lightweight design solution for the subsequent optimization process, and obtain a lighter body lightweight design solution on the condition that the design indicators are met as much as possible.

S2着重于获得高质量的连续变量解。而在S3中，求解S2中所建立的优化问题，可以快速收敛到优异的最优解。在S1所获得的解，即，汽车轻量化设计方案的基础上，进一步快速的找到更好的、满足设计指标的，足够轻的车身轻量化设计方案。S2 focuses on obtaining high-quality continuous variable solutions. In S3, solving the optimization problem established in S2 can quickly converge to an excellent optimal solution. On the basis of the solution obtained in S1, that is, the vehicle lightweight design scheme, a better lightweight design scheme for the body that meets the design indicators and is light enough is further quickly found.

步骤S1的具体实现过程包括：The specific implementation process of step S1 includes:

A1、对所述目标函数f(x)，建立以下RBF代理模型：A1. For the objective function f(x), establish the following RBF surrogate model:

Figure BDA0003000812800000041
Figure BDA0003000812800000041

其中xl为归档集中的第l组设计参数，

Figure BDA0003000812800000042
为高斯核函数，dis(x,xl)表示x与xl之间的距离，
Figure BDA0003000812800000043
其中
Figure BDA0003000812800000044
表示x与xl厚度变量的向量差，
Figure BDA0003000812800000045
为x与xl材料牌号向量的异或操作，
Figure BDA00030008128000000414
为由
Figure BDA00030008128000000413
和
Figure BDA0003000812800000047
组成的向量，||·||表示2-范数，N为归档集中所存储的设计参数的数目，所述归档集A为A＝{xl,yl,(cl,1,cl,2,cl,3)|l＝1,...,N}，
Figure BDA0003000812800000048
(cl,1,cl,2,cl,3)为第l组设计参数的约束条件函数值，
Figure BDA0003000812800000049
表示第l组设计参数的第i个厚度参数，
Figure BDA00030008128000000410
表示第l组设计参数的第j个材料牌号变量，yl为第l组设计参数的目标函数值，wl为权重；where x l is the lth group of design parameters in the archive set,
Figure BDA0003000812800000042
is a Gaussian kernel function, dis(x, x l ) represents the distance between x and x l ,
Figure BDA0003000812800000043
in
Figure BDA0003000812800000044
represents the vector difference between x and x l thickness variables,
Figure BDA0003000812800000045
is the exclusive OR operation of x and x l material grade vectors,
Figure BDA00030008128000000414
reason
Figure BDA00030008128000000413
and
Figure BDA0003000812800000047
The composed vector, || · || represents the 2-norm, N is the number of design parameters stored in the archive set, and the archive set A is A={x l , y l , (c l,1 ,c l ,2 ,c l,3 )|l=1,...,N},
Figure BDA0003000812800000048
( cl,1 , cl,2 , cl,3 ) is the constraint function value of the lth group of design parameters,
Figure BDA0003000812800000049
represents the ith thickness parameter of the lth group of design parameters,
Figure BDA00030008128000000410
represents the jth material grade variable of the lth group of design parameters, y l is the objective function value of the lth group of design parameters, and w l is the weight;

对于约束条件c1(x)，c1(x)和c1(x)，建立三个RBF代理模型

Figure BDA00030008128000000411
及
Figure BDA00030008128000000412
For constraints c 1 (x), c 1 (x) and c 1 (x), three RBF surrogate models are established
Figure BDA00030008128000000411
and
Figure BDA00030008128000000412

对所述目标函数f(x)，在设计参数归档集的基础上建立以下梯度提升树代理模型：For the objective function f(x), the following gradient boosting tree surrogate model is established on the basis of the design parameter archive set:

Figure BDA0003000812800000051
Figure BDA0003000812800000051

T(·,·)为回归树，Θm为第m个回归树的参数，M为回归树的总数目，

Figure BDA0003000812800000052
T(·,·) is the regression tree, Θ m is the parameter of the mth regression tree, M is the total number of regression trees,
Figure BDA0003000812800000052

对于约束条件c1(x)，c1(x)和c1(x)，建立三个梯度提升树代理模型

Figure BDA0003000812800000053
Figure BDA0003000812800000054
及
Figure BDA0003000812800000055
For constraints c 1 (x), c 1 (x) and c 1 (x), build three gradient boosting tree surrogate models
Figure BDA0003000812800000053
Figure BDA0003000812800000054
and
Figure BDA0003000812800000055

根据目标函数的函数值对蚁群算法群体中的参数进行排序，并为每组参数分配序数rank(s)，其中s代表蚁群算法群体P＝{xs|s＝1,...,K}中的第s组参数，

Figure BDA0003000812800000056
Figure BDA0003000812800000057
表示第s组设计参数的第i个厚度参数，
Figure BDA0003000812800000058
表示第s组设计参数的第j个材料牌号变量；为蚁群算法群体中的每组参数分配权重
Figure BDA0003000812800000059
q为蚁群算法参数，K为蚁群算法种群规模；在生成第h个后代的第i个厚度设计变量
Figure BDA00030008128000000510
时，首先计算概率
Figure BDA00030008128000000511
并根据该概率随机从蚁群算法群体中选择一个薄壁结构部件的厚度参数，记为μj，随后根据高斯分布
Figure BDA00030008128000000512
生成后代所对应的薄壁结构厚度参数，其中
Figure BDA00030008128000000513
ξ为蚁群算法参数；在生成第h个后代的第j个后代的材料牌号变量
Figure BDA00030008128000000514
时，根据概率ps从蚁群算法群体中随机选择一个材料牌号，并给予0.1的概率使该材料牌号变异为任意一种其他材料牌号；不断重复以上过程，直到生成H个后代以后，所有这些后代都被存储到集合O＝{xh|h＝1,...,H}中，其中
Figure BDA00030008128000000515
表示第h个后代，该后代表达了一组设计参数。Sort the parameters in the ant colony algorithm colony according to the function value of the objective function, and assign an ordinal rank(s) to each group of parameters, where s represents the ant colony algorithm colony P={x s |s=1,..., The sth group of parameters in K},
Figure BDA0003000812800000056
Figure BDA0003000812800000057
represents the ith thickness parameter of the sth group of design parameters,
Figure BDA0003000812800000058
represents the jth material grade variable of the sth set of design parameters; assigns weights to each set of parameters in the ant colony algorithm population
Figure BDA0003000812800000059
q is the ant colony algorithm parameter, K is the ant colony algorithm population size; the i-th thickness design variable is used to generate the h-th descendant
Figure BDA00030008128000000510
, first calculate the probability
Figure BDA00030008128000000511
And according to this probability, randomly select a thickness parameter of thin-walled structural components from the ant colony algorithm group, denoted as μ j , and then according to the Gaussian distribution
Figure BDA00030008128000000512
Generate the thickness parameters of the thin-walled structure corresponding to the descendants, where
Figure BDA00030008128000000513
ξ is the ant colony algorithm parameter; the material grade variable of the j-th descendant of the h-th descendant is generated
Figure BDA00030008128000000514
When , randomly select a material grade from the ant colony algorithm group according to the probability p s , and give a probability of 0.1 to mutate the material grade into any other material grade; repeat the above process continuously until H offspring are generated, all these The descendants are stored in the set O={x h |h=1,...,H}, where
Figure BDA00030008128000000515
represents the h-th descendant, which expresses a set of design parameters.

A2、使用所有的代理模型评价集合O中的所有后代，即将集合中的后代逐一代入所建立的代理模型中，并得到所对应的

Figure BDA00030008128000000516
Figure BDA00030008128000000517
和
Figure BDA00030008128000000518
的值；随后，根据这些值从O中选出两组设计参数，具体为：若O中存在满足条件
Figure BDA0003000812800000061
的设计参数，则从O中选择出满足该条件的
Figure BDA0003000812800000062
值最小的设计参数，否则选择出
Figure BDA0003000812800000063
值最小的设计参数；若存在
Figure BDA0003000812800000064
的设计参数，则从O中选择出满足该条件的
Figure BDA0003000812800000065
值最小的设计参数，否则选择出
Figure BDA0003000812800000066
值最小的设计参数；选择完毕后，再从O中随机选择出一组设计参数；使用目标函数f(x)和约束条件c1(x)，c2(x)，c3(x)评价选择出来的三组设计参数，并将这三组设计参数及其目标函数、约束条件值存入到设计参数归档集A中。A2. Use all surrogate models to evaluate all descendants in the set O, that is, insert the descendants in the set into the established surrogate model one by one, and get the corresponding
Figure BDA00030008128000000516
Figure BDA00030008128000000517
and
Figure BDA00030008128000000518
Then, two sets of design parameters are selected from O according to these values, specifically: if there is a satisfying condition in O
Figure BDA0003000812800000061
design parameters, then select the one that satisfies this condition from O
Figure BDA0003000812800000062
The design parameter with the smallest value, otherwise select the
Figure BDA0003000812800000063
The design parameter with the smallest value; if any
Figure BDA0003000812800000064
design parameters, then select the one that satisfies this condition from O
Figure BDA0003000812800000065
The design parameter with the smallest value, otherwise select the
Figure BDA0003000812800000066
The design parameter with the smallest value; after the selection is completed, a set of design parameters is randomly selected from O; the objective function f(x) and the constraints c 1 (x), c 2 (x), c 3 (x) are used to evaluate Three groups of design parameters are selected, and the three groups of design parameters, their objective functions, and constraints are stored in the design parameter archive set A.

步骤S2的具体实现过程为The specific implementation process of step S2 is as follows

A3、确定归档集A中最好的解，记为

Figure BDA0003000812800000067
其中
Figure BDA0003000812800000068
为厚度设计参数，
Figure BDA0003000812800000069
为材料牌号；A3. Determine the best solution in the archive set A, denoted as
Figure BDA0003000812800000067
in
Figure BDA0003000812800000068
is the thickness design parameter,
Figure BDA0003000812800000069
is the material grade;

A4、从归档集A中找到所有材料牌号变量和

Figure BDA00030008128000000610
相同的设计参数，并将这些设计参数的厚度变量存入到归档集Alocal中；根据Alocal，对目标函数和约束条件建立若干个连续RBF代理模型，得到步骤S2所述的问题。A4. Find all material grade variables and
Figure BDA00030008128000000610
The same design parameters are stored, and the thickness variables of these design parameters are stored in the archive set A local ; according to A local , several continuous RBF surrogate models are established for the objective function and constraints, and the problem described in step S2 is obtained.

与现有技术相比，本发明具备以下技术效果：Compared with the prior art, the present invention has the following technical effects:

1.在步骤A1中，本发明采用了混合变量同时采用两种代理模型：RBF代理模型和梯度提升树代理模型处理带有多种变量类型的目标函数和约束条件。这样做的优点是，RBF代理模型和梯度提升树代理模型能够适应于不同类型的变量，即RBF代理模型适合于处理带有连续变量的目标函数和约束条件，梯度提升树模型适合于处理带有离散变量的目标函数和约束条件。由于所要求解的汽车车身多材料结构匹配问题，即，同时获得的汽车车身结构中的尺寸参数和材料参数的最优值，同时包含连续变量和离散变量。因此，同时使用这两种代理模型更有助于处理这类问题。进而，在进行汽车轻量化的时候，这种方法更容易帮助设计人员找到较为优秀的汽车轻量化设计方案，即，在满足耐撞性要求的情况下获得极轻的车身。1. In step A1, the present invention adopts mixed variables and simultaneously adopts two surrogate models: RBF surrogate model and gradient boosting tree surrogate model to deal with objective functions and constraints with multiple variable types. The advantage of this is that the RBF surrogate model and the gradient boosting tree surrogate model can adapt to different types of variables, that is, the RBF surrogate model is suitable for dealing with objective functions and constraints with continuous variables, and the gradient boosting tree model is suitable for dealing with variables with Objective function and constraints for discrete variables. Due to the multi-material structure matching problem of the automobile body to be solved, that is, the optimal values of dimensional parameters and material parameters in the structure of the automobile body obtained at the same time, both continuous variables and discrete variables are included. Therefore, using both surrogate models at the same time is more helpful to deal with this kind of problem. Furthermore, when reducing the weight of the vehicle, this method is easier to help designers find a better lightweight design solution for the vehicle, that is, to obtain an extremely light body while meeting the crashworthiness requirements.

2.在步骤A2中，我们使用了两个指标，即

Figure BDA00030008128000000611
和
Figure BDA00030008128000000612
对不同解的优劣进行判断。其中，
Figure BDA0003000812800000071
注重于满足约束条件，即
Figure BDA0003000812800000072
而
Figure BDA0003000812800000073
则侧重于目标函数。以这种方式对A1中所生成的后代，有助于引导算法快速进入可行域，并在此基础上进一步提升目标函数的值。进而，这种方式有助于帮助设计人员快速找到符合设计需求的汽车轻量化设计方案，有利于提升设计效率，缩短设计周期。2. In step A2, we used two metrics, namely
Figure BDA00030008128000000611
and
Figure BDA00030008128000000612
Judge the pros and cons of different solutions. in,
Figure BDA0003000812800000071
Focus on satisfying constraints, i.e.
Figure BDA0003000812800000072
and
Figure BDA0003000812800000073
focus on the objective function. In this way, the offspring generated in A1 can help guide the algorithm to quickly enter the feasible region, and further improve the value of the objective function on this basis. Furthermore, this method helps designers to quickly find a lightweight automotive design solution that meets the design requirements, which is conducive to improving design efficiency and shortening the design cycle.

3.在步骤A3和A4中，我们仅使用归档集中的一部分解建立S2中所提到的优化问题，这种方式可以有效的专注于在一个局部范围内获得高精度的问题模型，进而提升获得高质量的解的效率。进而，该方法可以进一步提升轻量化设计的质量，获得更为优异的汽车轻量化设计方案。3. In steps A3 and A4, we only use a part of the solutions in the archive to establish the optimization problem mentioned in S2. This method can effectively focus on obtaining a high-precision problem model in a local range, thereby improving the Efficiency of high-quality solutions. Furthermore, this method can further improve the quality of lightweight design, and obtain a more excellent automotive lightweight design solution.

步骤A3之前，将所述三组设计参数并入到种群中，随后根据这三组设计参数对应的目标函数值和约束条件值进行排序，排序方法为：首先对于

Figure BDA0003000812800000074
的设计参数从大到小排序，随后对于
Figure BDA0003000812800000075
的设计参数，则根据目标函数值从大到小排序，将
Figure BDA0003000812800000076
的设计参数排到
Figure BDA0003000812800000077
的设计参数之前，排序完毕后将排序前的三组设计参数从种群中删除，即完成精英选择，将排序后的三组设计参数及其目标函数、约束条件值存入到设计参数归档集A中。Before step A3, the three groups of design parameters are incorporated into the population, and then the three groups of design parameters are sorted according to the corresponding objective function values and constraint values. The sorting method is as follows: first, for
Figure BDA0003000812800000074
The design parameters are sorted from largest to smallest, and then for
Figure BDA0003000812800000075
The design parameters are sorted according to the value of the objective function from large to small, and the
Figure BDA0003000812800000076
The design parameters of
Figure BDA0003000812800000077
Before sorting the design parameters, delete the three sets of design parameters before sorting from the population after sorting, that is, complete the elite selection, and store the three sets of design parameters after sorting, their objective functions, and constraint values in the design parameter archive set A. middle.

本发明还提供了一种汽车车身多材料结构匹配轻量化设计系统，包括计算机设备；所述计算机设备被配置或编程为用于执行上述方法的步骤。The present invention also provides an automotive body multi-material structure matching lightweight design system, comprising computer equipment; the computer equipment is configured or programmed to perform the steps of the above method.

作为一个发明构思，本发明还提供了一种计算机可读存储介质，其存储有程序；所述程序被配置为用于执行上述方法的步骤。As an inventive concept, the present invention also provides a computer-readable storage medium storing a program; the program is configured to execute the steps of the above method.

与现有技术相比，本发明的技术效果为：Compared with the prior art, the technical effect of the present invention is:

1、本发明能够处理同时带有连续变量(如汽车某部件结构尺寸)和离散变量(如汽车某部件的材料选择)的汽车轻量化设计问题；1. The present invention can deal with the lightweight design of automobiles with both continuous variables (such as the structural size of a certain part of the automobile) and discrete variables (such as the material selection of a certain automobile part);

1、本发明合理利用多个代理模型，能够在较短的设计周期内，获得优质的设计方案。1. The present invention reasonably utilizes multiple proxy models, and can obtain high-quality design solutions within a short design cycle.

附图说明Description of drawings

图1为本发明方法流程图；Fig. 1 is the flow chart of the method of the present invention;

图2为设计参数示意图。Figure 2 is a schematic diagram of the design parameters.

具体实施方式Detailed ways

现以图2所示的汽车车身结构轻量化设计为例对本发明所提出的方法进行说明。在该示例中，五个薄壁结构部件的厚度和材料牌号被设置为设计参数。本实施例的目的在于通过调整这五个薄壁结构部件的厚度和材料牌号，尽可能减小汽车B柱侵入量，同时保证B柱侵入速度、侧身结构整体重量以及结构吸能分别不超过T1，T2和T3。如图1，本实施例实施步骤如下：The method proposed by the present invention will now be described by taking the lightweight design of the vehicle body structure shown in FIG. 2 as an example. In this example, the thickness and material grade of five thin-walled structural components are set as design parameters. The purpose of this embodiment is to reduce the intrusion amount of the B-pillar of the automobile as much as possible by adjusting the thickness and material grade of the five thin-walled structural components, and at the same time ensure that the B-pillar intrusion speed, the overall weight of the side body structure and the structural energy absorption do not exceed T respectively. 1 , T 2 and T 3 . As shown in Figure 1, the implementation steps of this embodiment are as follows:

步骤1：建立优化问题；Step 1: Establish an optimization problem;

步骤2：算法初始化；Step 2: Algorithm initialization;

步骤3：建立代理模型；Step 3: Establish a proxy model;

步骤4：后代生成；Step 4: Generation of offspring;

步骤5：多代理模型辅助选择；Step 5: Multi-agent model assisted selection;

步骤6：更新群体；Step 6: Update the group;

步骤7：代理模型辅助局部搜索；Step 7: The proxy model assists the local search;

步骤8：不断执行2、3、4、5、6、7六个步骤，直至满足终止条件。Step 8: Repeat the six steps 2, 3, 4, 5, 6, and 7 until the termination condition is met.

在步骤1中，设计参数确定方式为，从汽车车身选择若干个薄壁结构部件，将这些部件的厚度作为连续设计参数，材料牌号作为离散设计参数，具体如图2所示。设计目标的确定方法为，将B柱侵入量、B柱侵入速度、B柱吸能或汽车车身结构重量其中一个指标作为目标函数，将其他三个指标作为约束条件，另这些指标小于一定的值，达到在不提升低汽车重量的情况下最大化汽车耐撞性，或在不降低汽车耐撞性的情况下最小化汽车重量。这些指标可通过有限元分析软件获得，如LS-DYNA。最终，建立以下优化问题：In step 1, the design parameters are determined by selecting several thin-walled structural components from the automobile body, using the thickness of these components as a continuous design parameter, and the material grade as a discrete design parameter, as shown in Figure 2. The method for determining the design objective is to take one of the B-pillar intrusion amount, B-pillar intrusion speed, B-pillar energy absorption or the weight of the vehicle body structure as the objective function, and the other three indicators as the constraint conditions, and the other indicators are less than a certain value. , to maximize car crashworthiness without lifting low car weight, or minimize car weight without reducing car crashworthiness. These indicators can be obtained by finite element analysis software, such as LS-DYNA. Finally, the following optimization problem is established:

min:f(x)min:f(x)

s.t.c1(x)≤T1 stc 1 (x)≤T 1

c2(x)≤T2 c 2 (x)≤T 2

c3(x)≤T3 c 3 (x)≤T 3

x＝[xthick,xmat]x＝[x thick ,x mat ]

Figure BDA0003000812800000091
Figure BDA0003000812800000091

Figure BDA0003000812800000092
Figure BDA0003000812800000092

Figure BDA0003000812800000093
Figure BDA0003000812800000093

Figure BDA0003000812800000094
Figure BDA0003000812800000094

其中x为一组设计参数，xthick为由厚度设计参数组成的向量，xmat为由材料设计参数组成的向量，f(x)为目标函数，即B柱侵入量，c1(x)为第一个约束条件，即B柱侵入速度，c2(x)为第二个约束条件，即侧身结构重量，c3(x)为第三个约束条件，即B柱吸能，目标函数和所有约束条件的函数值可由有限元分析获得，T1，T2和T3分别为三个约束条件需要满足的指标，即，要求汽车的B柱侵入速度不超过T1，B柱吸能不超过T2，侧身重量不超过T3，Li和Ui为第i个厚度设计参数的下限和上限，M1～M5分别代表五种高强钢，牌号分别为DP440，DP500，DP600，DP780，DP980。where x is a set of design parameters, x thick is a vector composed of thickness design parameters, x mat is a vector composed of material design parameters, f(x) is the objective function, namely the B-pillar intrusion, and c 1 (x) is The first constraint is the B-pillar intrusion velocity, c 2 (x) is the second constraint, the sideways structure weight, c 3 (x) is the third constraint, the B-pillar energy absorption, the objective function and The function values of all constraints can be obtained by finite element analysis. T 1 , T 2 and T 3 are the indicators that the three constraints need to be satisfied respectively, that is, the B-pillar intrusion speed of the car is required not to exceed T 1 , and the energy absorption of the B-pillar should not exceed T 1 . Over T 2 , the side weight does not exceed T 3 , Li and U i are the lower and upper limits of the i -th thickness design parameter, M 1 to M 5 represent five kinds of high-strength steels, and the grades are DP440, DP500, DP600, DP780 , DP980.

在步骤2中，以下参数将被初始化：In step 2, the following parameters will be initialized:

蚁群算法参数：包括蚁群算法群体规模K，用于生成后代设计参数的两个常量q和ξ。Ant colony algorithm parameters: including the ant colony algorithm colony size K, two constants q and ξ used to generate the design parameters of the offspring.

蚁群算法群体：包含一系列汽车车身结构参数及其设计函数响应值，记为P＝{xs|s＝1,...,K}，其中

Figure BDA0003000812800000095
Figure BDA0003000812800000096
表示第s组设计参数的第i个厚度参数，
Figure BDA0003000812800000097
表示第s组设计参数的第j个材料牌号变量。汽车车身结构参数具体如图2所示，为汽车车身薄壁结构部件的厚度和每个薄壁结构部件的材料牌号。设计函数响应值具体可以为B柱腰线侵入量、B柱腰线侵入速度、B柱吸能以及B柱比吸能，可由有限元分析获得。在后续步骤中，该群体将存储经有限元分析评价过的、具有最好性能指标的K组设计参数。Ant colony algorithm colony: contains a series of car body structure parameters and their design function response values, denoted as P={x s |s=1,...,K}, where
Figure BDA0003000812800000095
Figure BDA0003000812800000096
represents the ith thickness parameter of the sth group of design parameters,
Figure BDA0003000812800000097
represents the jth material grade variable of the sth group of design parameters. The structural parameters of the automobile body are specifically shown in Figure 2, which are the thickness of the thin-walled structural components of the automobile body and the material grade of each thin-walled structural component. The response value of the design function can specifically be the B-pillar waistline intrusion amount, the B-pillar waistline intrusion speed, the B-pillar energy absorption, and the B-pillar specific energy absorption, which can be obtained by finite element analysis. In a subsequent step, the population will store the K sets of design parameters that have been evaluated by finite element analysis and have the best performance indicators.

设计参数归档集：包含一系列汽车车身结构参数及其设计函数响应值，记为A＝{xl,yl,(cl,1,cl,2,cl,3)|l＝1,...,N}，其中

Figure BDA0003000812800000098
yl为第l组设计参数的目标函数值，(cl,1,cl,2,cl,3)为第l组设计参数的约束条件函数值，
Figure BDA0003000812800000101
表示第l组设计参数的第j个厚度参数，
Figure BDA0003000812800000102
表示第l组设计参数的第j个材料牌号变量。Design parameter archive set: contains a series of auto body structural parameters and their design function response values, denoted as A={x l ,y l ,( cl,1 , cl,2 , cl,3 )|l=1 ,...,N}, where
Figure BDA0003000812800000098
y l is the objective function value of the lth group of design parameters, ( cl,1 , cl,2 , cl,3 ) is the constraint function value of the lth group of design parameters,
Figure BDA0003000812800000101
represents the jth thickness parameter of the lth group of design parameters,
Figure BDA0003000812800000102
represents the jth material grade variable of the lth group of design parameters.

在步骤3中，分别对目标函数和每个约束条件建立两种代理模型，即RBF模型和梯度提升树模型。具体如下：In step 3, two surrogate models are established for the objective function and each constraint, namely, the RBF model and the gradient boosting tree model. details as follows:

对于目标函数，在设计参数归档集的基础上建立以下RBF代理模型For the objective function, the following RBF surrogate model is established on the basis of the design parameter archive set

Figure BDA0003000812800000103
Figure BDA0003000812800000103

其中xl为归档集中的第i组设计参数，dis(x,xl)表示x与xl之间的距离，具体为

Figure BDA0003000812800000104
其中
Figure BDA0003000812800000105
表示x与xl厚度变量的向量差，
Figure BDA0003000812800000106
为x与xl材料牌号向量的异或操作，
Figure BDA0003000812800000107
为由
Figure BDA0003000812800000108
和
Figure BDA0003000812800000109
组成的向量，||·||表示2-范数。wl为权重，具体计算方式为：where x l is the ith group of design parameters in the archive set, dis(x, x l ) represents the distance between x and x l , specifically
Figure BDA0003000812800000104
in
Figure BDA0003000812800000105
represents the vector difference between x and x l thickness variables,
Figure BDA0003000812800000106
is the exclusive OR operation of x and x l material grade vectors,
Figure BDA0003000812800000107
reason
Figure BDA0003000812800000108
and
Figure BDA0003000812800000109
A vector consisting of ||·|| represents the 2-norm. w l is the weight, the specific calculation method is:

w＝(ΦTΦ)-1(ΦTy)w=(Φ T Φ) -1 (Φ T y)

其中w＝(w1,...,wN)为权重向量，y＝(y1,...,yN)为设计参数归档集中所存储的目标函数响应值，Φ为以下矩阵：where w=(w 1 ,...,w N ) is the weight vector, y=(y 1 ,...,y N ) is the response value of the objective function stored in the design parameter archive set, and Φ is the following matrix:

Figure BDA00030008128000001010
Figure BDA00030008128000001010

类似的，对于约束条件c1(x)，c2(x)和c3(x)，以类似的方法建立三个RBF代理模型

Figure BDA00030008128000001011
及
Figure BDA00030008128000001012
具体为Similarly, for the constraints c 1 (x), c 2 (x) and c 3 (x), three RBF surrogate models are established in a similar way
Figure BDA00030008128000001011
and
Figure BDA00030008128000001012
Specifically

Figure BDA00030008128000001013
Figure BDA00030008128000001013

Figure BDA00030008128000001014
Figure BDA00030008128000001014

Figure BDA00030008128000001015
Figure BDA00030008128000001015

其中权重

Figure BDA00030008128000001016
和
Figure BDA00030008128000001017
分别按以下方式计算：where the weight
Figure BDA00030008128000001016
and
Figure BDA00030008128000001017
Calculated as follows:

wc1＝(ΦTΦ)-1(ΦTc1)w c1 = (Φ T Φ) -1 (Φ T c 1 )

wc2＝(ΦTΦ)-1(ΦTc2)w c2 = (Φ T Φ) -1 (Φ T c 2 )

wc3＝(ΦTΦ)-1(ΦTc3)w c3 = (Φ T Φ) -1 (Φ T c 3 )

其中c1＝(c1,1,...,cN,1)，c2＝(c1,2,...,cN,2)和c3＝(c1,3,...,cN,3)代替y＝(y1,...,yN)为数据归档集中所存储的约束条件响应值所构成的向量。where c 1 =(c 1,1 ,...,c N,1 ), c 2 =(c 1,2 ,...,c N,2 ) and c 3 =(c 1,3 ,.. .,c N,3 ) instead of y=(y 1 , . . . , y N ) is a vector formed by the constraint response values stored in the data archive set.

对于目标函数，在设计参数归档集的基础上建立以下梯度提升树代理模型For the objective function, the following gradient boosting tree surrogate model is established on the basis of the design parameter archive set

Figure BDA0003000812800000111
Figure BDA0003000812800000111

其中T(·,·)为回归树，Θm为第m个回归树的参数，其计算方式如下where T(·,·) is the regression tree, and Θ m is the parameter of the mth regression tree, which is calculated as follows

Figure BDA0003000812800000112
Figure BDA0003000812800000112

其中

Figure BDA0003000812800000113
in
Figure BDA0003000812800000113

类似的，对于约束条件c1(x)，c2(x)和c3(x)，以类似的方法建立三个梯度提升树模型

Figure BDA0003000812800000114
及
Figure BDA0003000812800000115
具体如下：Similarly, for the constraints c 1 (x), c 2 (x) and c 3 (x), three gradient boosted tree models are built in a similar way
Figure BDA0003000812800000114
and
Figure BDA0003000812800000115
details as follows:

Figure BDA0003000812800000116
Figure BDA0003000812800000116

Figure BDA0003000812800000117
Figure BDA0003000812800000117

Figure BDA0003000812800000118
Figure BDA0003000812800000118

梯度提升树模型参数

Figure BDA0003000812800000119
和
Figure BDA00030008128000001110
时按以下方式获得：Gradient boosted tree model parameters
Figure BDA0003000812800000119
and
Figure BDA00030008128000001110
is obtained as follows:

Figure BDA00030008128000001111
Figure BDA00030008128000001111

Figure BDA00030008128000001112
Figure BDA00030008128000001112

Figure BDA00030008128000001113
Figure BDA00030008128000001113

其中

Figure BDA00030008128000001114
Figure BDA00030008128000001115
in
Figure BDA00030008128000001114
Figure BDA00030008128000001115

在步骤4中，采用混合变量蚁群算法(ACOMV)生成后代，每个后代代表一组设计参数，并被表示为

Figure BDA00030008128000001116
其中
Figure BDA00030008128000001117
代表第h个后代的第i个厚度设计参数，
Figure BDA00030008128000001118
代表第h个后代的第j个材料设计参数，具体生成方式如下所述。首先，根据设计函数响应值对蚁群算法群体中的参数进行排序，排序方式具体为，首先对于对于
Figure BDA00030008128000001119
的设计参数则根据目标函数值从大到小排序，随后对于
Figure BDA0003000812800000121
的设计参数，根据
Figure BDA0003000812800000122
的值从小到大排序，最后将
Figure BDA0003000812800000123
的设计参数排到
Figure BDA0003000812800000124
后面。排序后，为每组参数分配序数rank(s)，其中s代表蚁群算法群体中的第s组参数；随后为蚁群算法群体中的每组参数分配权重
Figure BDA0003000812800000125
其中q为蚁群算法参数，K为蚁群算法种群规模；首先，根据设计函数响应值对蚁群算法群体中的参数进行排序，并为每组参数分配序数rank(s)，其中s代表蚁群算法群体P＝{xs|s＝1,...,K}中的第s组参数，
Figure BDA0003000812800000126
Figure BDA0003000812800000127
表示第s组设计参数的第i个厚度参数，
Figure BDA0003000812800000128
表示第s组设计参数的第j个材料牌号变量。随后为蚁群算法群体中的每组参数分配权重
Figure BDA0003000812800000129
其中q为蚁群算法参数，K为蚁群算法种群规模；接下来，在生成
Figure BDA00030008128000001210
时，首先计算概率
Figure BDA00030008128000001211
并根据该概率随机从蚁群算法群体中选择一个薄壁结构部件的厚度参数，记为μj，随后根据高斯分布
Figure BDA00030008128000001212
生成后代所对应的薄壁结构厚度参数，其中
Figure BDA00030008128000001213
ξ为蚁群算法参数；在生成后代的材料牌号变量
Figure BDA00030008128000001214
时，根据概率ps从蚁群算法群体中随机选择一个材料牌号，并给与0.1的概率使其可以变异为任意一种其他材料牌号。不断重复以上过程，直到生成H个后代以后，所有这些后代都被存储到集合O＝{xh|h＝1,...,H}中，其中
Figure BDA00030008128000001215
In step 4, the ant colony algorithm with mixed variables (ACO MV ) is used to generate offspring, each offspring represents a set of design parameters and is denoted as
Figure BDA00030008128000001116
in
Figure BDA00030008128000001117
represents the ith thickness design parameter of the hth descendant,
Figure BDA00030008128000001118
represents the jth material design parameter of the hth descendant, and is generated as follows. First, the parameters in the ant colony algorithm group are sorted according to the response value of the design function. The sorting method is as follows:
Figure BDA00030008128000001119
The design parameters are sorted according to the objective function value from large to small, and then for
Figure BDA0003000812800000121
design parameters, according to
Figure BDA0003000812800000122
The values are sorted from small to large, and finally the
Figure BDA0003000812800000123
The design parameters of
Figure BDA0003000812800000124
Behind. After sorting, assign an ordinal rank(s) to each group of parameters, where s represents the s-th group of parameters in the ant colony algorithm colony; then assign weights to each group of parameters in the ant colony algorithm colony
Figure BDA0003000812800000125
Among them, q is the ant colony algorithm parameter, and K is the ant colony algorithm population size; first, the parameters in the ant colony algorithm colony are sorted according to the response value of the design function, and the ordinal rank(s) is assigned to each group of parameters, where s represents the ant colony. Swarm algorithm population P={x s |s=1,...,K} the sth group of parameters,
Figure BDA0003000812800000126
Figure BDA0003000812800000127
represents the ith thickness parameter of the sth group of design parameters,
Figure BDA0003000812800000128
represents the jth material grade variable of the sth group of design parameters. Then assign weights to each set of parameters in the ant colony algorithm population
Figure BDA0003000812800000129
Where q is the ant colony algorithm parameter, K is the ant colony algorithm population size; next, in the generation
Figure BDA00030008128000001210
, first calculate the probability
Figure BDA00030008128000001211
And according to this probability, randomly select a thickness parameter of thin-walled structural components from the ant colony algorithm group, denoted as μ j , and then according to the Gaussian distribution
Figure BDA00030008128000001212
Generate the thickness parameters of the thin-walled structure corresponding to the descendants, where
Figure BDA00030008128000001213
ξ is the ant colony algorithm parameter; the material grade variable in the generation of offspring
Figure BDA00030008128000001214
When , randomly select a material grade from the ant colony algorithm group according to the probability p s , and give it a probability of 0.1 so that it can mutate into any other material grade. Repeat the above process until H offspring are generated, and all these offspring are stored in the set O={x h |h=1,...,H}, where
Figure BDA00030008128000001215

在步骤5中，首先使用步骤3所建立的代理模型评价集合O中的所有后代，具体为将集合中的后代逐一带入步骤3中所建立的代理模型中，并得到所对应的

Figure BDA00030008128000001216
和
Figure BDA00030008128000001217
的值。随后，根据这些值从O中选出两组设计参数，具体为：若O中存在满足条件
Figure BDA0003000812800000131
的设计参数，则从O中选择出满足该条件的
Figure BDA0003000812800000132
值最小的设计参数，否则则选择出
Figure BDA0003000812800000133
值最小的设计参数；类似的，若存在
Figure BDA0003000812800000134
的设计参数，则从O中选择出满足该条件的
Figure BDA0003000812800000135
值最小的设计参数，否则则选择出
Figure BDA0003000812800000136
值最小的设计参数。选择完毕后，再从O中随机选择出一组设计参数。接下来使用目标函数f(x)和约束条件c1(x)，c2(x)，c3(x)评价选择出来的三组设计参数，并将这三组设计参数及其目标函数、约束条件值存入到设计参数归档集中。In step 5, first use the surrogate model established in step 3 to evaluate all descendants in the set O, specifically bringing the descendants in the set into the surrogate model established in step 3 one by one, and obtain the corresponding
Figure BDA00030008128000001216
and
Figure BDA00030008128000001217
value of . Then, two sets of design parameters are selected from O according to these values, specifically: if there is a satisfying condition in O
Figure BDA0003000812800000131
design parameters, then select the one that satisfies this condition from O
Figure BDA0003000812800000132
The design parameter with the smallest value, otherwise select the
Figure BDA0003000812800000133
the design parameter with the smallest value; similarly, if present
Figure BDA0003000812800000134
design parameters, then select the one that satisfies this condition from O
Figure BDA0003000812800000135
The design parameter with the smallest value, otherwise select the
Figure BDA0003000812800000136
The design parameter with the smallest value. After the selection is complete, a set of design parameters is randomly selected from O. Next, use the objective function f(x) and the constraints c 1 (x), c 2 (x), c 3 (x) to evaluate the selected three sets of design parameters, and compare the three sets of design parameters and their objective functions, Constraint values are stored in the design parameter archive.

在步骤6中，采用精英选择对群体进行更新。具体为，首先将步骤5中选择得到的三组设计参数并入到种群中，随后根据他们的目标函数值和约束条件值进行排序，排序方法为，首先对于

Figure BDA0003000812800000137
的设计参数从大到小排序，随后对于
Figure BDA0003000812800000138
的设计参数则根据目标函数值从大到小排序，将
Figure BDA0003000812800000139
的设计参数排到
Figure BDA00030008128000001310
的设计参数之前。排序完毕后将最前面的三组设计参数从种群中删除，最终完成精英选择。In step 6, the population is updated with elite selection. Specifically, firstly, the three sets of design parameters selected in step 5 are incorporated into the population, and then they are sorted according to their objective function values and constraint values. The sorting method is as follows: first for
Figure BDA0003000812800000137
The design parameters are sorted from largest to smallest, and then for
Figure BDA0003000812800000138
The design parameters are sorted according to the value of the objective function from large to small, and the
Figure BDA0003000812800000139
The design parameters of
Figure BDA00030008128000001310
before the design parameters. After sorting, the first three sets of design parameters are deleted from the population, and the elite selection is finally completed.

在步骤7中，首先确定归档集中最好的解，记为

Figure BDA00030008128000001311
其中
Figure BDA00030008128000001312
为厚度设计参数，
Figure BDA00030008128000001313
为材料牌号。随后，从归档集A中找到所有材料牌号变量和
Figure BDA00030008128000001314
相同的设计参数，并将这些设计参数的厚度变量存入到归档集Alocal中。接下来，根据Alocal，对目标函数和约束条件建立若干个连续RBF代理模型构造方法与步骤5相同，最终得到以下优化问题In step 7, first determine the best solution in the archive set, denoted as
Figure BDA00030008128000001311
in
Figure BDA00030008128000001312
is the thickness design parameter,
Figure BDA00030008128000001313
is the material grade. Then, find all material grade variables from archive set A and
Figure BDA00030008128000001314
The same design parameters and the thickness variables of these design parameters are stored in the archive set A local . Next, according to A local , establish several continuous RBF surrogate models for the objective function and constraints. The construction method is the same as step 5, and finally the following optimization problem is obtained

Figure BDA00030008128000001315
Figure BDA00030008128000001315

Figure BDA00030008128000001316
Figure BDA00030008128000001316

Figure BDA00030008128000001317
Figure BDA00030008128000001317

Figure BDA00030008128000001318
Figure BDA00030008128000001318

Figure BDA00030008128000001319
Figure BDA00030008128000001319

Figure BDA00030008128000001320
Figure BDA00030008128000001320

使用序列二次规划求解该问题，得到解

Figure BDA00030008128000001321
该解代表由局部搜索给出的最优薄壁结构厚度设计参数，随后将
Figure BDA00030008128000001322
和
Figure BDA00030008128000001323
组合成为一个新的解
Figure BDA00030008128000001324
该解为由局部搜索给出的最优设计参数。最终，使用有限元分析评价解xlocal，即，根据该xlocal所提供的参数设置，将该参数设置带入到有限元分析软件LS-DYNA中，利用LS-DYNA获得的仿真结果并得到目标函数和约束条件的值。随后将该解存入到设计参数归档集中，并同步骤6一样，使用精英选择更新群体。值得注意的是，我们将所有评价过的解都放入了归档集。这是因为归档集中的数据越多，模型建立的就越准确，进而更有助于优化求解。Solve the problem using sequential quadratic programming, and get the solution
Figure BDA00030008128000001321
This solution represents the optimal thin-wall structure thickness design parameters given by a local search, which is then
Figure BDA00030008128000001322
and
Figure BDA00030008128000001323
combined into a new solution
Figure BDA00030008128000001324
The solution is the optimal design parameters given by the local search. Finally, the finite element analysis is used to evaluate the solution x local , that is, according to the parameter settings provided by the x local , the parameter settings are brought into the finite element analysis software LS-DYNA, and the simulation results obtained by LS-DYNA are used to obtain the target. Values for functions and constraints. This solution is then stored in the design parameter archive, and as in step 6, the population is updated using elite selection. Notably, we put all evaluated solutions into the archive set. This is because the more data in the archive set, the more accurate the model can be built, which in turn can help optimize the solution.

Claims (6)

1. A matching lightweight design method for an automobile body material structure is characterized by comprising the following steps:
s1, establishing the following optimization problems:
Figure FDA0003000812790000011
wherein x is a design parameter，xthickFor vectors composed of thickness design parameters, xmatIs a vector composed of material design parameters, f (x) is an objective function, c1(x) As a first constraint, c2(x) As a second constraint, c3(x) As a third constraint, T1，T2And T3Are indexes that three constraint conditions need to satisfy, LiAnd UiLower and upper limits of the design parameter for the ith thickness, M1～MpEach representing p different materials, n1For the number of thickness parameters to be optimized, n2Number of parameters for the grade of the material to be optimized;
s2, converting the optimization problem into the following problems:
Figure FDA0003000812790000012
s3, solving the problem of the step S2 by using sequential quadratic programming to obtain a solution
Figure FDA0003000812790000013
This solution represents the optimal thin-wall structure thickness design parameter given by the local search, which will then follow
Figure FDA0003000812790000014
And
Figure FDA0003000812790000015
combined into a new solution
Figure FDA0003000812790000021
The solution is the optimal design parameter given by the local search.
2. The method for matching and designing the lightweight automobile body material structure according to claim 1, wherein the step S1 is realized by the following steps:
a1, establishing the following RBF proxy model for the objective function f (x):
Figure FDA0003000812790000022
wherein xlParameters are designed for the l-th group in the archive set,
Figure FDA0003000812790000023
is a Gaussian kernel function, dis (x, x)l) Denotes x and xlThe distance between the two or more of the two or more,
Figure FDA0003000812790000024
Figure FDA0003000812790000025
wherein
Figure FDA0003000812790000026
Denotes x and xlThe vector difference in the thickness variation,
Figure FDA0003000812790000027
is x and xlThe xor operation of the vector of the material brand,
Figure FDA0003000812790000028
is composed of
Figure FDA0003000812790000029
And
Figure FDA00030008127900000210
the vector is formed, wherein | · | | represents a 2-norm, N is the number of design parameters stored in an archive set, and A is { x ═ x |, in the archive setl,yl,(cl,1,cl,2,cl,3)|l＝1,...,N}，
Figure FDA00030008127900000211
(cl,1,cl,2,cl,3) The constraint function values for the l-th set of design parameters,
Figure FDA00030008127900000212
an ith thickness parameter representing the ith set of design parameters,
Figure FDA00030008127900000213
j material grade variable, y, representing the l set of design parameterslDesigning the objective function value of the parameter for the l-th group, wlIs a weight;
for constraint c1(x)，c1(x) And c1(x) Establishing three RBF proxy models
Figure FDA00030008127900000214
And
Figure FDA00030008127900000215
for the target function f (x), establishing the following gradient lifting tree proxy model on the basis of a design parameter filing set:
Figure FDA00030008127900000216
t (,) is a regression tree, ΘmIs a parameter of the mth regression tree, M is the total number of regression trees,
Figure FDA0003000812790000031
for constraint c1(x)，c1(x) And c1(x) Establishing three gradient lifting tree agent models
Figure FDA0003000812790000032
Figure FDA0003000812790000033
And
Figure FDA0003000812790000034
sorting the parameters in the ant colony algorithm colony according to the function value of the objective function, and distributing ordinal rank(s) to each group of parameters, wherein s represents the ant colony algorithm colony P ═ { x ═sThe s-th set of parameters in 1., K },
Figure FDA0003000812790000035
Figure FDA0003000812790000036
an ith thickness parameter representing an s-th set of design parameters,
Figure FDA0003000812790000037
a jth material grade variable representing a set s of design parameters; assigning weights to each set of parameters in an ant colony algorithm population
Figure FDA0003000812790000038
q is an ant colony algorithm parameter, and K is an ant colony algorithm population scale; design variable of ith thickness in generation of h-th offspring
Figure FDA0003000812790000039
First, the probability is calculated
Figure FDA00030008127900000310
And randomly selecting a thickness parameter of a thin-wall structural part from the ant colony algorithm colony according to the probability, and recording the thickness parameter as mujThen according to a Gaussian distribution
Figure FDA00030008127900000311
Generating a thin-wall structure thickness parameter corresponding to the offspring, wherein
Figure FDA00030008127900000312
Xi is an ant colony algorithm parameter; material grade variable at the jth offspring from which the h offspring was generated
Figure FDA00030008127900000313
According to the probability psRandomly selecting a material grade from an ant colony algorithm colony, and giving a probability of 0.1 to change the material grade into any other material grade; the above process is repeated until after H offspring are generated, all of which are stored in the set O ═ xh1,. H, H ], wherein
Figure FDA00030008127900000314
Represents the h-th offspring expressing a set of design parameters.
A2, evaluating all descendants in the set O by using all the agent models, namely substituting the descendants in the set into the established agent models one by one to obtain the corresponding agent models
Figure FDA00030008127900000315
Figure FDA00030008127900000316
And
Figure FDA00030008127900000317
a value of (d); then, two sets of design parameters are selected from O according to these values, specifically: if there is a sufficient condition in O
Figure FDA0003000812790000041
The design parameter of (1), then selecting the one satisfying the condition from O
Figure FDA0003000812790000042
The design parameter with the smallest value is selected otherwise
Figure FDA0003000812790000043
The design parameter with the smallest value; if present, is
Figure FDA0003000812790000044
The design parameter of (1), then selecting the one satisfying the condition from O
Figure FDA0003000812790000045
The design parameter with the smallest value is selected otherwise
Figure FDA0003000812790000046
The design parameter with the smallest value; after the selection is finished, randomly selecting a group of design parameters from O; using an objective function f (x) and a constraint c1(x)，c2(x)，c3(x) And evaluating the three selected groups of design parameters, and storing the three groups of design parameters, the target functions and the constraint condition values into a design parameter filing set A to obtain the optimization problem.
3. The method for designing an automobile body material with a matched structure and reduced weight according to claim 2,
the specific implementation process of step S2 includes:
a3, determining the best solution in archive set A, and recording as
Figure FDA0003000812790000047
Wherein
Figure FDA0003000812790000048
The parameters are designed for the thickness of the film,
Figure FDA0003000812790000049
is a material brand;
a4, finding all material trade mark variables and
Figure FDA00030008127900000410
the same design parameters, and setting themStoring the thickness variable of the metering parameter into an archive set AlocalPerforming the following steps; according to AlocalAnd establishing a plurality of continuous RBF proxy models for the objective function and the constraint condition to obtain the problem of the step S2.
4. The method for matching and designing the light weight of the automobile body material structure according to claim 3, wherein before step A3, the three sets of design parameters are incorporated into a population, and then the three sets of design parameters are sorted according to objective function values and constraint condition values corresponding to the three sets of design parameters, wherein the sorting method comprises the following steps: first, for
Figure FDA00030008127900000411
Is ordered from large to small, then for
Figure FDA00030008127900000412
The design parameters of (2) are sorted from large to small according to the objective function value
Figure FDA00030008127900000413
Is arranged to
Figure FDA00030008127900000414
Before the design parameters are sorted, deleting the three groups of design parameters before sorting from the population after sorting is finished, namely finishing elite selection, and storing the three groups of design parameters after sorting, target functions of the three groups of design parameters and constraint condition values into a design parameter filing set A.
5. A matching lightweight design system for an automobile body material structure is characterized by comprising computer equipment; the computer device is configured or programmed for carrying out the steps of the method according to one of claims 1 to 4.
6. A computer-readable storage medium characterized by storing a program; the program is configured for carrying out the steps of the method according to one of claims 1 to 4.
CN202110345933.4A 2021-03-31 2021-03-31 Automobile body material structure matching lightweight design method, system and storage medium Active CN113158514B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110345933.4A CN113158514B (en) 2021-03-31 2021-03-31 Automobile body material structure matching lightweight design method, system and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110345933.4A CN113158514B (en) 2021-03-31 2021-03-31 Automobile body material structure matching lightweight design method, system and storage medium

Publications (2)

Publication Number Publication Date
CN113158514A true CN113158514A (en) 2021-07-23
CN113158514B CN113158514B (en) 2022-05-24

Family

ID=76885675

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110345933.4A Active CN113158514B (en) 2021-03-31 2021-03-31 Automobile body material structure matching lightweight design method, system and storage medium

Country Status (1)

Country Link
CN (1) CN113158514B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115730382A (en) * 2022-11-03 2023-03-03 中南大学 A Rapid Design Method for Crashworthiness and Lightweight of Automobile Body
CN115935522A (en) * 2023-01-10 2023-04-07 河南科技大学 An optimization method for the B-pillar structure of a lightweight hybrid material vehicle
WO2023233931A1 (en) * 2022-06-03 2023-12-07 株式会社日立製作所 Design assistance device and design assistance method

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5729463A (en) * 1995-09-01 1998-03-17 Ulsab Trust Designing and producing lightweight automobile bodies
CN102799704A (en) * 2012-05-21 2012-11-28 湖南大学 Optimal design method for spliced and welded plate automobile door
CN106919767A (en) * 2017-03-09 2017-07-04 江铃汽车股份有限公司 Automobile body-in-white lightweight analysis method
CN109063389A (en) * 2018-09-28 2018-12-21 重庆长安汽车股份有限公司 A kind of vehicle structure lightweight forward design method and system based on more performance constraints
CN109190189A (en) * 2018-08-10 2019-01-11 武汉理工大学 A kind of body side wall safety component hybrid variable design method for optimization of matching
CN111310281A (en) * 2020-03-24 2020-06-19 郑州大学 Multi-disciplinary design method for light weight of automobile frame
CN111881931A (en) * 2020-06-10 2020-11-03 中南大学 Implementation method and device for solving response discontinuous expensive optimization problem in automobile body lightweight structure design

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5729463A (en) * 1995-09-01 1998-03-17 Ulsab Trust Designing and producing lightweight automobile bodies
CN102799704A (en) * 2012-05-21 2012-11-28 湖南大学 Optimal design method for spliced and welded plate automobile door
CN106919767A (en) * 2017-03-09 2017-07-04 江铃汽车股份有限公司 Automobile body-in-white lightweight analysis method
CN109190189A (en) * 2018-08-10 2019-01-11 武汉理工大学 A kind of body side wall safety component hybrid variable design method for optimization of matching
CN109063389A (en) * 2018-09-28 2018-12-21 重庆长安汽车股份有限公司 A kind of vehicle structure lightweight forward design method and system based on more performance constraints
CN111310281A (en) * 2020-03-24 2020-06-19 郑州大学 Multi-disciplinary design method for light weight of automobile frame
CN111881931A (en) * 2020-06-10 2020-11-03 中南大学 Implementation method and device for solving response discontinuous expensive optimization problem in automobile body lightweight structure design

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
李金武: "《不确定性时变参数下结构的响应边界计算与可靠性分析》", 《中国优秀硕士学位论文全文数据库》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023233931A1 (en) * 2022-06-03 2023-12-07 株式会社日立製作所 Design assistance device and design assistance method
CN115730382A (en) * 2022-11-03 2023-03-03 中南大学 A Rapid Design Method for Crashworthiness and Lightweight of Automobile Body
CN115730382B (en) * 2022-11-03 2023-08-11 中南大学 A Rapid Design Method for Crashworthiness and Lightweight of Automobile Body
CN115935522A (en) * 2023-01-10 2023-04-07 河南科技大学 An optimization method for the B-pillar structure of a lightweight hybrid material vehicle

Also Published As

Publication number Publication date
CN113158514B (en) 2022-05-24

Similar Documents

Publication Publication Date Title
CN113158514B (en) Automobile body material structure matching lightweight design method, system and storage medium
Song et al. Sensitivity analysis and reliability based design optimization for high-strength steel tailor welded thin-walled structures under crashworthiness
Duan et al. Parametric modeling and multiobjective crashworthiness design optimization of a new front longitudinal beam
Duan et al. Multi-objective reliability-based design optimization for the VRB-VCS FLB under front-impact collision
CN110020466A (en) Negative poisson's ratio structure energy-absorption box optimization design method based on agent model
CN105608286B (en) Bumper of vehicles optimization method based on particle swarm optimization algorithm
CN103034766A (en) Method for determining ply angle direction and thickness of laminated board composite material
CN106184085A (en) A kind of closed loop negative poisson's ratio construction car security system and method for designing
Du et al. The vehicle’s velocity prediction methods based on rnn and lstm neural network
CN113408046B (en) Multidisciplinary optimization method for riveting point variable-pitch arrangement based on crashworthiness
CN106294972A (en) A kind of vehicle bridge multidisciplinary reliability design optimization method
Hosseini et al. Optimal design of the S-rail family for an automotive platform with novel modifications on the product-family optimization process
CN108984901A (en) A kind of automobile body crash-worthiness optimization method
CN117350154A (en) Parameter matching optimization method for electric drive system of pure electric vehicle
CN114912195A (en) Aerodynamic sequence optimization method for commercial vehicle
CN104392031B (en) A kind of design method of white body non-uniform beam
Xiao et al. Optimization design of crashworthiness of polyurethane foam filled Origami thin wall square tube based on an surrogate model
CN110110480A (en) A kind of carbon fiber bumper anti-collision girder construction optimum design method considering laying compatibility
CN114896688B (en) A design method for energy absorption box of vehicle body structure
CN115510561B (en) Multitask-based automobile energy absorption box structure optimization design method and system
CN117313248A (en) Vehicle body structure optimization method and system considering vehicle body performance and collision injury
Zhou et al. An enhanced hybrid and adaptive meta-model based global optimization algorithm for engineering optimization problems
CN108133068B (en) Truss type unmanned vehicle body lightweight design method
CN115795678A (en) Parameter optimization method and storage medium for conceptual design of vehicle body structure
Ao et al. Artificial intelligence-aided design (AIAD) for structures and engineering: a state-of-the-art review and future perspectives

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant