Design optimization method based on combined function proxy model

Abstract

The invention relates to a design optimization method based on a combined function proxy model, which comprises the following steps: constructing an initial simulation model according to a to-be-optimized product, and determining a design variable of the initial simulation model and a definition domain corresponding to the design variable; sampling the design variables based on a preset sampling method to generate initial sampling points; inputting the initial sampling points into an initial simulation model, calculating initial sampling values, and generating an initial sample set according to the relationship between the initial sampling points and the initial sampling values; constructing a combined function agent model according to the initial sample set and a pre-selected primary function; solving the combined function agent model through a preset iterative algorithm to obtain a primary function coefficient of the combined function agent model; and carrying out optimization calculation on the combination function agent model to obtain an optimal design point. According to the method provided by the invention, the calling times of the black box function can be greatly reduced, and the product design optimization efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of model optimization, and in particular to a design optimization method based on a combined function proxy model. Background Art

[0002] As the functions of modern electromechanical products (such as aircraft, automobiles, high-end CNC machine tools, etc.) become more complex and intelligent, the system composition and control become more and more complex. Therefore, modeling and simulation technology has been widely used in the design process to simulate and analyze the design scheme and optimize the decision-making to improve the comprehensive performance of the product. However, in order to accurately simulate the real product structure and system, the simulation model often presents multidisciplinary and nonlinear characteristics, which makes the simulation model solution time very long. Taking the car collision time as an example, the simulation time may reach 30-160 hours. The design optimization of the product requires a large number of simulation models, which leads to a longer time required for modern product design optimization.

[0003] The existing design method can regard the simulation model of the product as a black box function, and use the optimization algorithm to perform a global search in the design space to obtain the optimal solution, so as to minimize the number of calls to the black box function. Since the running time of the black box function is very long, a proxy model can be constructed. Commonly used proxy models include: polynomial regression, Kriging model, radial basis function, support vector regression, moving least squares, etc. The premise that the proxy model can be used in engineering product design optimization is that its approximation accuracy of the black box function is high enough. In order to ensure the approximation accuracy, the proxy model often requires a large number of sampling points to solve the model parameters. Sampling requires calling the black box function or the original simulation model, which is very time-consuming and computationally expensive. How to use a small number of sampling points to build a high-precision proxy model has become a difficult problem in the field of engineering design optimization. Summary of the invention

[0004] In order to solve the problems existing in the above-mentioned background technology, the present invention provides a design optimization method based on a combinatorial function proxy model, which can solve the basis function coefficients of the proxy model through a small number of sampling points, complete the construction of the proxy model, reduce the number of calls to the black box function, and improve the efficiency of product design optimization.

[0005] In a first aspect, the present application provides a design optimization method based on a combinatorial function proxy model, comprising:

[0006] Constructing an initial simulation model of the product to be optimized according to the product to be optimized, and determining design variables of the initial simulation model and their corresponding definition domains; wherein the number of the design variables is greater than 1;

[0007] Sampling the design variables based on a preset sampling method to generate initial sampling points;

[0008] Inputting the initial sampling points into an initial simulation model, calculating initial sampling values, and generating an initial sample set according to the relationship between the initial sampling points and the initial sampling values;

[0009] Constructing a combined function proxy model according to the initial sample set and the pre-selected basis functions; wherein the basis functions are of more than one type;

[0010] Solving the combined function proxy model through a preset iterative algorithm to obtain basis function coefficients of the combined function proxy model, thereby completing the construction of the combined function proxy model;

[0011] The combined function proxy model is optimized and calculated to obtain the optimal design point.

[0012] In a second aspect, the present application provides a design optimization system based on a combinatorial function agent model, comprising:

[0013] An initial simulation model acquisition module is used to construct an initial simulation model of the product to be optimized according to the product to be optimized, and determine the design variables of the initial simulation model and their corresponding definition domains; wherein the number of the design variables is greater than 1;

[0014] A sampling module, used to sample the design variables based on a preset sampling method to generate initial sampling points;

[0015] A sample set generation module, used to input the initial sampling point into the initial simulation model, calculate the initial sampling value, and generate an initial sample set according to the relationship between the initial sampling point and the initial sampling value;

[0016] A model building module, used to build a combined function proxy model according to the initial sample set and the pre-selected basis function; wherein the type of the basis function is more than one;

[0017] An iterative calculation module, used to solve the combined function proxy model through a preset iterative algorithm, obtain the basis function coefficients of the combined function proxy model, and complete the construction of the combined function proxy model;

[0018] The optimization calculation module is used to optimize the combination function proxy model to obtain the optimal design point.

[0019] In a third aspect, the present application provides a computer device, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method as described above are implemented.

[0021] The design optimization method based on the combinatorial function proxy model described in the present application constructs an initial simulation model according to product requirements, determines the design variables of the product and the corresponding definition domains, samples the sampling points of the design variables, inputs the sampling points into the initial simulation model output to obtain the sampling values ​​corresponding to the sampling points of the model, constructs a combinatorial function proxy model according to a plurality of pre-selected basis functions and a sample set generated by combining the sampling points and the sampling values, uses an iterative algorithm to iteratively calculate the coefficients of the proxy model, completes the construction of the combinatorial function proxy model, uses an optimization algorithm to optimize the proxy model and finally obtains the optimal design point of the product. The constructed proxy model has high approximation accuracy and stability, and can effectively improve the efficiency of product design optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 A flow chart of a design optimization method based on a combinatorial function proxy model provided in an embodiment of the present application;

[0024] Figure 2 A structural diagram of a design optimization system based on a combinatorial function proxy model provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present invention are clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention. The flowchart used in this application shows the operations implemented in some embodiments of the embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical contextual relationships can be reversed in order or implemented simultaneously. In addition, under the guidance of the content of this application, those skilled in the art can add at least one other operation to the flowchart, or remove at least one operation from the flowchart.

[0026] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0027] First, see Figure 1 , Figure 1 A flowchart of a design optimization method based on a combinatorial function proxy model provided in an embodiment of the present application, the method comprising:

[0028] S101, constructing an initial simulation model of the product to be optimized according to the product to be optimized, and determining design variables of the initial simulation model and their corresponding definition domains; wherein the number of the design variables is greater than 1;

[0029] S102, sampling the design variables based on a preset sampling method to generate initial sampling points;

[0030] S103, inputting the initial sampling point into the initial simulation model, calculating an initial sampling value, and generating an initial sample set according to the relationship between the initial sampling point and the initial sampling value;

[0031] S104, constructing a combined function proxy model according to the initial sample set and the pre-selected basis functions; wherein the basis functions may be of more than one type;

[0032] S105, solving the combined function proxy model by a preset iterative algorithm to obtain basis function coefficients of the combined function proxy model, thereby completing the construction of the combined function proxy model;

[0033] S106, performing optimization calculation on the combined function proxy model to obtain the optimal design point.

[0034] The design optimization method based on the combinatorial function proxy model provided in the present application constructs an initial simulation model according to product requirements, determines the design variables of the product and the corresponding definition domains, samples the sampling points of the design variables, inputs the sampling points into the initial simulation model output to obtain the sampling values ​​corresponding to the sampling points of the model, constructs a combinatorial function proxy model according to a sample set generated by a combination of a plurality of pre-selected basis functions and the sampling points and the sampling values, uses an iterative algorithm to iteratively calculate the coefficients of the proxy model, completes the construction of the combinatorial function proxy model, uses an optimization algorithm to optimize the proxy model, and finally obtains the optimal design point of the product. The constructed proxy model has high approximation accuracy and stability, and can effectively improve the efficiency of product design optimization.

[0035] In step S101, an initial simulation model of the product to be optimized is constructed according to the product to be optimized, and design variables of the initial simulation model and their corresponding definition domains are determined.

[0036] Design variables are parameters that require decision making or optimization during product design. Design variables can be continuous (such as length, weight, temperature, etc.) or discrete (such as material type, model, etc.). The choice of design variables depends on the specific design problem and design goals. The domain refers to the range of values ​​allowed for the design variable. For continuous variables, the domain is usually a closed interval, such as [a, b], where a and b are the minimum and maximum values ​​of the variable, respectively; for discrete variables, the domain is usually a finite set.

[0037] In this application, an initial simulation model of the product to be optimized is constructed, and the design variables of the initial simulation model and the definition domain corresponding to each design variable are determined according to the actual needs of the product.

[0038] For step S102, the design variables are sampled based on a preset sampling method to generate initial sampling points. Specifically, the preset sampling method described in the present application is a Latin hypercube sampling method.

[0039] The number of initial sampling points is determined by the running time of the simulation model. If the running time of the simulation model is short, more sampling can be performed, and if the running time is long, fewer sampling can be performed. Generally, the number of initial sampling points is between 10 times and 20 times the design variables.

[0040] Among them, the initial sampling points of the present application are generated by the Latin Hypercube sampling method. In complex systems involving multiple design variables, Latin Hypercube Sampling (LHS) is a method of approximate random sampling from a multi-source parameter distribution. It belongs to the stratified sampling technology. By dividing the value range of each independent variable into several equally spaced intervals, and randomly selecting a point in each interval as a sampling point, it ensures that the sample points are evenly distributed throughout the value range, effectively avoiding the situation where some areas are not covered or repeatedly covered due to randomness, so as to better reflect the characteristics of the design space under fewer sample conditions. This method is particularly suitable for situations where a large number of simulations or experiments are required to evaluate the design performance. The present application adopts the Latin Hypercube sampling method to sample the design variables in the initial simulation model, and uses the sampled sampling points and sampling values ​​to construct a combinatorial function proxy model, which can effectively reduce the call of the simulation model and reduce the design optimization time.

[0041] For step S103, the initial sampling points are input into the initial simulation model, initial sampling values ​​are calculated, and an initial sample set is generated according to the relationship between the initial sampling points and the initial sampling values.

[0042] The initial sampling points are sampling points obtained by sampling the design variables of the product. In the present application, the initial simulation model is a black box function model. The initial sampling points are input into the model as input values, and the output values ​​are calculated as the initial sampling values. The initial sample set is generated based on the input-output relationship between the initial sampling points and the initial sampling values.

[0043] A black box function is a type of function whose input-output correspondence is known but not its internal structure, and whose specific expression cannot be written. Each time a set of inputs is given, a set of output values ​​can be obtained through the calculation of the black box function, but the precise expression of the black box function cannot be written. The black box function in product design refers to a template or system with specific inputs and outputs, but its internal implementation details are encapsulated and not disclosed to the outside world. Users or external systems only need to care about its inputs and outputs without understanding its internal working principles.

[0044] For step S104, a combined function proxy model is constructed according to the initial sample set and the pre-selected basis functions; wherein the basis functions are of more than one type.

[0045] Specifically, the step S104 includes:

[0046] S1041, combining basis functions according to the types and quantities of pre-selected basis functions to form a basis function combination; wherein the basis functions are polynomial functions;

[0047] S1042, constructing a combined function proxy model based on the basis function combination and the initial sample set; wherein the initial sampling point is an input variable of the combined function proxy model, and the initial sampling value is an output variable of the combined function proxy model.

[0048] The basis functions include Legendre polynomials, Chebyshev polynomials, sine functions, cosine functions, and user-defined functions. In the embodiment of the present application, the combined function proxy model is composed of multiple types of basis functions, and the types of basis functions and the number of each type of basis functions are set by the user.

[0049] In some embodiments, 2-3 types of basis functions are selected, and each basis function is composed of a low-order polynomial. For a single design variable, the number of basis functions of each type is at least 5; for multiple design variables, the number of basis functions of each type is not less than 5 times the number of initial sampling points, and the total number does not exceed 1000. The total order of each basis function starts from a low order, and the total order is the sum of the polynomial functions corresponding to each design variable.

[0050] A surrogate model (or meta model) is an approximate mathematical model built through a data-driven approach to simulate and predict the behavior of a complex system. It aims to be as close to the behavior of the real system as possible while reducing computational cost and time. Surrogate models are usually modeled based on a limited number of intelligently selected data points, which can be experimental data, simulation data, or other types of observational data.

[0051] This application combines the initial sample set with the pre-selected basis function to construct a combined function proxy model. The combined function proxy model is:

[0052]

[0053] Among them, y is the output variable, x is the input variable, d1 is the number of basis functions of the first type, and d2 is the number of basis functions of the second type. is the i-th basis function in the first basis function, θ 1,i is the corresponding coefficient, dp is the number of the p-th basis function, is the i-th basis function in the p-th basis function, θ p,iis the coefficient of the i-th basis function in the p-th basis function. There are p types of basis functions in the combined function proxy model, and the total number of functions D = (d1+d2+…dp).

[0054] The x and y in the combined function proxy model are respectively the initial sampling point and the initial sampling value in the initial sample set obtained by the above steps. The present application constructs a proxy model by combining the basis functions pre-selected by the user, obtains the corresponding coefficients by solving, and then constructs a proxy model that can approximately represent the relationship between the input variable x and the output variable y, thereby solving the optimization problem of design variables in product design.

[0055] Further, the number of design variables m is defined. In the present application, the number of design variables is greater than 1 (i.e., m>1). For the combination of basis functions in the combination function proxy model, for the ith basis function of the dth class, it can be expressed as:

[0056]

[0057] Where m is the number of design variables, for The exponential vector, F(x,η (i) ) is the form of basis function, is a single variable x j No. Order function.

[0058] For step S105, the combined function proxy model is solved by a preset iterative algorithm to obtain basis function coefficients of the combined function proxy model, thereby completing the construction of the combined function proxy model.

[0059] The core idea of ​​iterative weighted algorithms is to update the weight of each data point or parameter according to the result of the previous iteration or some criteria during the iteration process. This weight update is usually aimed at improving the accuracy, stability or convergence speed of the algorithm. Through continuous iteration, the algorithm can gradually approach the optimal solution or meet specific optimization goals.

[0060] The present application proposes a method of solving the coefficients of the basis functions in the combined function proxy model by iterative calculation. Specifically, step S105 includes:

[0061] S1051, based on the combined function proxy model, obtaining a matrix expression of the combined function proxy model:

[0062] y=φΘ

[0063] Wherein, Φ is the set of basis functions of the combined function proxy model, Θ is the set of basis function coefficients of the combined function proxy model, φ=[Φ1, Φ2,…, Φp , Θ = [Θ1, Θ2, …, Θ p T , where p is the number of types of basis functions.

[0064] In the construction of the combined function surrogate model, it is assumed that after n samplings, y is a vector of [n, 1], x is a vector of [n, m], and solving for the coefficients of the basis functions is equivalent to solving a linear equation system with n rows and D columns (n << D). This application needs to solve the sparsest solution of this linear equation system, that is, the solution with the fewest non-zero numbers in Θ.

[0065] S1052, construct a preset basis function coefficient solving formula according to the matrix expression:

[0066]

[0067] where λ d is the weight of the d-th type of basis function, W d = diag(w d,1 , w d,2 , …, w d,dp ) is the weight matrix of the d-th type of basis function, w d,i is the weight of the i-th basis function in the d-th type of basis function, θ d is the coefficient of the d-th type of basis function, and p is the number of types of basis functions.

[0068] S1053, initialize λ d and W d to make

[0069] S1054, use a preset iterative algorithm to solve the basis function coefficient solving formula:

[0070]

[0071] where y is the output variable of the combined function surrogate model.

[0072] The iterative process includes the steps of:

[0073] Compare and judge whether θ (t) meets the preset threshold, and iteratively update the weight λ d and the weight matrix W d :

[0074]

[0075]

[0076] where nd is the number of the d-th type of basis function, δ is the preset parameter value of the iterative algorithm, θ d,i ​is the coefficient of the ith basis function in the dth basis function. In this application, δ=0.0001 is set, and the preset parameter value is the preset parameter value in the iterative algorithm model. The selection of δ will affect the convergence speed and calculation results of the algorithm calculation. This application sets the δ parameter value according to experimental data.

[0077] If θ (t) Meet the preset threshold, output θ (t) are the basis function coefficients of the combined function proxy model.

[0078] Wherein, t=1, 2, 3 ... T, T is the number of iterations in the iterative algorithm, in some embodiments, the number of iterations T may be 10 times, and in other embodiments, the number of iterations T may also be adjusted according to actual conditions. The calculation process of the iterative algorithm may be implemented by programming with mathematical software.

[0079] For step S106, the combined function proxy model is optimized and calculated to obtain the optimal design point. Based on the combined function proxy model generated in the above steps, the combined function proxy model is optimized and calculated using an optimization algorithm to obtain the optimal design point. Specifically, step S106 includes:

[0080] S1061, obtaining an optimization problem of the combined function proxy model, wherein the optimization problem is expressed as:

[0081]

[0082] Wherein, the optimization problem is an optimization problem with equality constraints, and h(x) is an equality constraint;

[0083] S1062, converting the optimization problem into a Lagrangian function:

[0084]

[0085] in, is a combined function proxy model, h(x) is an equality constraint, λ is the multiplier vector of h(x), and λ T is the transpose of λ;

[0086] where h(x)=[h1(x),…,h l (x)] T Indicates that there are l equality constraints, λ T =[λ1,…,λ l ] T ;

[0087] S1063, constructing an unconstrained sub-problem of the optimization problem:

[0088]

[0089] in, is the penalty term for the unconstrained subproblem, σ (k) is the penalty factor of the penalty term;

[0090] S1064, initialize the unconstrained subproblem parameters so that x (0) ∈R m ,λ (1) ∈R l ,σ (1) >0,0≤ε≤1,υ∈(0,1),η>1,k=1; where ε is the preset threshold;

[0091] S1065, x (k-1) Iteratively calculate the minimum point x of the unconstrained problem for the initial point (k) ; In which, the k value increases by 1 in each iteration;

[0092] During the iteration, if ‖h(x)‖≤ε, output x (k) as the optimal design point.

[0093] Furthermore, for each iterative calculation, the penalty term and multiplier vector in the unconstrained subproblem are checked and updated, including:

[0094] If ||h(x (k) )||≥υ||h(x (k-1) )||, so that σ (k+1) =ησ (k) ;

[0095] According to x (k) and h(x (k) ) value to update the multiplier vector so that λ (k+1) =λ (k) -σ (k) h(x (k) ).

[0096] This application uses an optimization algorithm to solve the minimum point of the unconstrained subproblem. By setting a threshold ε, h(x) is compared with the threshold. When h(x) meets the preset threshold, it is judged that the equation satisfies the equality constraint, and then the iterative calculation is stopped, and x is output. (k) As the optimal solution of the optimization problem, the optimal solution is used as the optimal design point.

[0097] In this application, in order to further verify whether the obtained optimal design point is suitable for the product, after obtaining the optimal design point, the following steps are also included:

[0098] S107, inputting the optimal design point into the initial simulation model to verify the optimization result.

[0099] The optimal design point is input into the initial simulation model corresponding to the product and run to verify whether the point meets the product design and improve the product optimization effect.

[0100] The design optimization method based on the combinatorial function proxy model provided by the present application adopts a preset sampling method to sample the design variables of the product to obtain sampling points, and inputs the sampling points into the simulation model of the product to calculate and obtain the corresponding sampling values; constructs a combinatorial function proxy model based on multiple types of basis functions and sampling sample sets, calculates the coefficients of each basis function in the model through a preset iterative algorithm, completes the construction of the proxy model, uses an optimization algorithm to optimize and calculate to obtain the optimal design point of the model, and finally inputs the final design point into the simulation model of the product to verify the result. The method uses a small number of sampling points to construct a proxy model of the combinatorial function, which can save time and computing costs, and uses an iterative algorithm and optimization method to solve the model coefficients to finally obtain the optimal solution of the model, which has high approximation accuracy and stability, can greatly reduce the number of calls to the black box function, and improve the design optimization efficiency of the product.

[0101] Second, see Figure 2 , Figure 2 A structural diagram of a design optimization system based on a combinatorial function proxy model provided in an embodiment of the present application includes:

[0102] An initial simulation model acquisition module 11 is used to construct an initial simulation model of the product to be optimized according to the product to be optimized, and determine the design variables of the initial simulation model and their corresponding definition domains;

[0103] A sampling module 12, used to sample the design variables based on a preset sampling method to generate initial sampling points;

[0104] A sample set generation module 13, used to input the initial sampling point into the initial simulation model, calculate the initial sampling value, and generate an initial sample set according to the relationship between the initial sampling point and the initial sampling value;

[0105] A model building module 14 is used to build a combined function proxy model based on the initial sample set and the pre-selected basis function; wherein the type of the basis function is more than one;

[0106] An iterative calculation module 15 is used to solve the combined function proxy model through a preset iterative algorithm to obtain basis function coefficients of the combined function proxy model;

[0107] The optimization calculation module 16 is used to perform optimization calculation on the combined function proxy model to obtain the optimal design point.

[0108] In a third aspect, the present application provides a computer device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned design optimization method based on the combinatorial function proxy model when executing the computer program.

[0109] The processor may be a general-purpose processor, such as a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and may implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of the present application. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in conjunction with the embodiments of the present application may be directly embodied as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0110] As a non-volatile computer-readable storage medium, the memory can be used to store non-volatile software programs, non-volatile computer executable programs and modules. The memory can include at least one type of storage medium, for example, it can include flash memory, hard disk, multimedia card, card-type memory, random access memory (Random Access Memory, RAM), static random access memory (Static Random Access Memory, SRAM), programmable read-only memory (Programmable Read Only Memory, PROM), read-only memory (Read Only Memory, ROM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), magnetic memory, disk, optical disk, etc. The memory is any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 21 in the embodiment of the present application can also be a circuit or any other device that can realize a storage function, for storing program instructions and / or data.

[0111] In a fourth aspect, the present application provides a readable storage medium storing a computer program, which, when executed by a processor, executes the steps of the above-mentioned design optimization solution based on the combinatorial function proxy model.

[0112] The above embodiments are only used to introduce the technical solutions of the present application in detail, but the description of the above embodiments is only used to help understand the methods of the embodiments of the present application and should not be understood as limiting the present application. Any changes or substitutions that can be easily thought of by those skilled in the art should be included in the protection scope of the embodiments of the present application.

Claims

1. A design optimization method based on a combinatorial function agent model, characterized in that: include: Constructing an initial simulation model of the product to be optimized according to the product to be optimized, and determining design variables of the initial simulation model and their corresponding definition domains; wherein the number of the design variables is greater than 1; Sampling the design variables based on a preset sampling method to generate initial sampling points; Inputting the initial sampling point into the initial simulation model, calculating an initial sampling value, and generating an initial sample set according to the relationship between the initial sampling point and the initial sampling value; Constructing a combined function proxy model according to the initial sample set and the pre-selected basis functions; wherein the basis functions are of more than one type; Solving the combined function proxy model through a preset iterative algorithm to obtain basis function coefficients of the combined function proxy model, thereby completing the construction of the combined function proxy model; The combined function proxy model is optimized and calculated to obtain the optimal design point.

2. The design optimization method based on the combinatorial function proxy model according to claim 1 is characterized in that: The step of constructing a combined function proxy model according to the initial sample set and the pre-selected basis functions includes: Combining basis functions according to the types and quantities of pre-selected basis functions to form a basis function combination; wherein the basis functions are polynomial functions; Based on the basis function combination and the initial sample set, a combined function proxy model is constructed; wherein the initial sampling point is an input variable of the combined function proxy model, and the initial sampling value is an output variable of the combined function proxy model; The combined function proxy model is: Among them, y is the output variable, x is the input variable, dp is the number of the p-th type basis function, is the i-th basis function in the p-th basis function, θ p,i is the coefficient of the i-th basis function in the p-th type basis function.

3. The design optimization method based on the combinatorial function proxy model according to claim 2 is characterized in that: The basis functions are: in, is the ith basis function in the dth type basis function, m is the number of design variables, for The exponential vector, F(x,η (i) ) is the form of basis function, is a single variable x j No. Order function.

4. The design optimization method based on the combinatorial function proxy model according to claim 1 is characterized in that: The step of solving the combined function proxy model by a preset iterative algorithm to obtain basis function coefficients of the combined function proxy model includes: Based on the combined function proxy model, a matrix expression of the combined function proxy model is obtained: y=ΦΘ Wherein, Φ is the set of basis functions of the combined function proxy model, Θ is the set of basis function coefficients of the combined function proxy model, Φ=[Φ1, Φ2,…, Φ p ],Θ=[Θ1,Θ2,…,Θ p ] T , p is the number of basis functions; According to the matrix expression, a preset basis function coefficient solution formula is constructed: Among them, λ d is the weight of the d-th type basis function, W d =diag(w d,1 ,w d,2 ,…,w d,dp ) is the weight matrix of the d-th type basis function, w d,i is the weight of the ith basis function in the dth type of basis function, θ d is the coefficient of the d-th type basis function, and p is the number of types of basis functions; Initialize λ d and W d ,make The preset iterative algorithm is used to solve the basis function coefficient solution formula: Among them, y is the output variable of the combined function proxy model; The iterative process includes the steps: Comparative judgment Θ (t) Whether the preset threshold is met and the weight λ is updated iteratively d and the weight matrix W d : Among them, nd is the number of the d-th type of basis function, δ is the preset parameter value of the iterative algorithm, and θ d,i is the coefficient of the i-th basis function in the d-th type basis function; If θ (t) Meet the preset threshold, output θ (t) are the basis function coefficients of the combined function proxy model.

5. The design optimization method based on the combinatorial function proxy model according to claim 1 is characterized in that: The step of optimizing the combined function proxy model to obtain the optimal design point includes: The optimization problem of the combined function proxy model is obtained, and the optimization problem is expressed as: Wherein, the optimization problem is an optimization problem with equality constraints, and h(x) is an equality constraint; Convert the optimization problem into a Lagrangian function: in, is a combined function proxy model, h(x) is an equality constraint, λ is the multiplier vector of h(x), and λ T is the transpose of λ; where h(x)=[h1(x),…,h l (x)] T Indicates that there are l equality constraints, λ T =[λ1,…,λ l ] T ; Construct an unconstrained subproblem of the optimization problem: in, is the penalty term for the unconstrained subproblem, σ (k) is the penalty factor of the penalty term; Initialize the unconstrained subproblem parameters so that x (0) ∈R m ,λ (1) ∈R l ,σ (1) >0,0≤ε≤1,υ∈(0,1),η>1,k=1; where ε is the preset threshold; x (k-1) Iteratively calculate the minimum point x of the unconstrained problem for the initial point (k) ; In which, the k value increases by 1 in each iteration; During the iteration, if ‖h(x)‖≤ε, output x (k) as the optimal design point.

6. The design optimization method based on the combinatorial function proxy model according to claim 5 is characterized in that: Each iteration calculation also includes: If ||h(x (k) )|| ≥ υ||h(x (k-1) )‖, then make σ (k+1) = ησ (k) ; According to x (k) and h(x (k) ) value to update the multiplier vector so that λ (k+1) =λ (k) -σ (k) h(x (k) ).

7. The design optimization method based on the combinatorial function proxy model according to claim 1 is characterized in that: After obtaining the optimal design point, the method further includes: The optimal design point is input into the initial simulation model to verify the optimization result.

8. A design optimization system based on a combinatorial function agent model, characterized in that: include: An initial simulation model acquisition module is used to construct an initial simulation model of the product to be optimized according to the product to be optimized, and determine the design variables of the initial simulation model and their corresponding definition domains; wherein the number of the design variables is greater than 1; A sampling module, used to sample the design variables based on a preset sampling method to generate initial sampling points; A sample set generation module, used to input the initial sampling point into the initial simulation model, calculate the initial sampling value, and generate an initial sample set according to the relationship between the initial sampling point and the initial sampling value; A model building module, used to build a combined function proxy model according to the initial sample set and the pre-selected basis function; wherein the type of the basis function is more than one; An iterative calculation module, used to solve the combined function proxy model through a preset iterative algorithm, obtain the basis function coefficients of the combined function proxy model, and complete the construction of the combined function proxy model; The optimization calculation module is used to optimize the combination function proxy model to obtain the optimal design point.

9. A computer device comprising: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Images