CN114218686A - Multi-precision data smooth scale approximate modeling method for aircraft - Google Patents
Multi-precision data smooth scale approximate modeling method for aircraft Download PDFInfo
- Publication number
- CN114218686A CN114218686A CN202210154400.2A CN202210154400A CN114218686A CN 114218686 A CN114218686 A CN 114218686A CN 202210154400 A CN202210154400 A CN 202210154400A CN 114218686 A CN114218686 A CN 114218686A
- Authority
- CN
- China
- Prior art keywords
- precision
- low
- function
- sample point
- scale
- 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
Links
Images
Classifications
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Geometry (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (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
The invention discloses a smooth scale approximate modeling method for multi-precision data of an aircraft. In the process of establishing the scale function, the relevance of the high-precision model and the low-precision model is considered, the scale function is enabled to be as smooth as possible, namely the linear term of the low-precision model, the shape parameter of the scale function and the smooth factor of the scale function are subjected to parameter training by taking the minimum average curvature of the scale function as a target, and finally the prediction precision of the multi-precision proxy model is obviously improved, so that the high-efficiency and accurate construction of the multi-precision model of the aircraft is realized, the number of high-precision sample points required by modeling is reduced, the modeling efficiency is obviously improved, and the subsequent optimization is guided.
Description
Technical Field
The invention relates to the technical field of optimization design of aircrafts, in particular to a smooth scale approximate modeling method for multi-precision data of an aircraft.
Background
With the rapid development of computer technology, an optimization method based on a proxy model becomes one of the methods widely applied in the design process of an aircraft. However, in the process of optimization design, multiple times of calling of a high-precision simulation model cannot be avoided, and the calculation time is consumed, so that the requirements of high speed and high efficiency of aircraft design cannot be met. The multi-precision model fusion order-reduction characterization method is used for establishing the multi-precision agent model by introducing low-precision numerical simulation analysis, so that the calculation consumption can be obviously reduced while the accuracy of the model is ensured, and the subsequent optimization design is guided.
The commonly used multi-precision agent model establishing method at present comprises the following steps:
1. the multi-precision proxy model modeling method based on the scale function comprises the following steps: establishing a low-precision proxy model by using a large number of low-precision sample points, establishing a scale function between the high-precision model and the low-precision model by taking the error between the high-precision real output of the sample points and the prediction output of the low-precision model as output, and fusing the scale function and the low-precision proxy model by adopting an additive scale or multiplication scale method to establish a multi-precision proxy model;
2. the multi-precision agent model modeling method based on the space mapping comprises the following steps: and searching a proper transfer function, mapping the design variable space of the high-precision analysis model to the design variable space of the low-precision analysis model, or mapping the low-precision output space to the output space of the high-precision analysis model to construct a multi-precision proxy model. Converting the high-precision model optimization problem into a low-precision model optimization problem through a transfer function;
3. the Co-Kriging multi-precision agent model modeling method comprises the following steps: based on the Bayesian theory, a trend is provided by a low-precision analysis model, and a multi-precision agent model is constructed by interpolating high-precision sample points.
The conventional multi-precision agent model modeling method has the following defects:
1. the relevance research of high-precision and low-precision analysis models is insufficient based on a scale function and a Co-Kriging multi-precision agent model modeling method, high-precision numerical simulation and low-precision numerical simulation are still regarded as two independent modeling problems, the strong relevance of the high-precision numerical simulation and the low-precision numerical simulation in engineering problems is not searched, and the further improvement of the model precision is limited;
2. the multi-precision agent model modeling method based on space mapping mainly enables the optimal solution of a low-precision simulation function to approach the optimal solution of a high-precision simulation function through the design space of a low-precision simulation function, the core of the process lies in finding a proper mapping relation to convert the high-precision simulation function and the low-precision simulation function, however, the form of a conversion function is complicated and difficult to judge the accuracy of the mapping relation, continuous trial is needed, and meanwhile, the capability of quantifying errors is lacked, so that the method is poor in applicability.
Disclosure of Invention
Aiming at the problems that in the prior art, when a multi-precision agent model is constructed in the process of aircraft optimization design, the model precision is poor, and the follow-up optimization design cannot be effectively guided, the invention provides a multi-precision data smooth scale approximate modeling method for an aircraft, which can effectively improve the performance and realize the accurate construction of the multi-precision agent model in the aircraft optimization design.
To achieve the above object, the present invention provides a smooth scale approximate modeling method for multi-precision data of an aircraft, comprising the steps of:
step 1, obtaining design variables and a design domain, and initially sampling to obtain high-precision sample points and low-precision sample points;
step 4, constructing a scale function based on the scale function sample point set to obtain the average curvature of the scale function in the design domain, and performing smoothness training on the scale function by taking the minimum average curvature of the scale function in the design domain as a target;
and 5, obtaining a multi-precision agent model.
In another embodiment, in step 1, the initial sampling obtains a high-precision sample point and a low-precision sample point, specifically:
respectively selecting by adopting a Latin hypercube sampling methodA low precision sampling point andeach high-precision sampling point is obtained by running a simulation model with corresponding precision at each sampling point respectivelyA low precision sample point andthe high-precision sample points of each sample are respectively as follows:
in the formula (I), the compound is shown in the specification,X iLis shown asiThe input value of one of the low-precision sample points,Y iLis shown asiThe response value of the sample point of low precision,X iHis shown asiThe input value of each sample point with high precision,Y iHis shown asiAnd (5) high-precision sample point response values.
In another embodiment, in step 2, the low-precision proxy model is:
in the formula (I), the compound is shown in the specification,f L(x) In order to be a low-precision proxy model,for any point in the design domainAnd a firstA low precision sample pointIs a distance therebetween, i.e.;Is as followsThe basis function coefficients of the low precision sample points,is as followsThe basis functions of the low precision sample points.
In another embodiment, a Gauss function is selected as a basis function of the low-precision sample points, which is:
in the formula (I), the compound is shown in the specification,is as followsThe shape parameters of the basis functions are:
in the formula (I), the compound is shown in the specification,is as followsThe distance between the one low precision sample point to the farthest sample point,to design spatial dimensions;
will be provided withSubstituting the input value and the response value of each low-precision sample point into the low-precision proxy model to obtain the basis function coefficient of the low-precision sample pointAnd solving the equation set to obtain a basis function coefficient to obtain the low-precision proxy model.
In another embodiment, step 3 specifically includes:
calculating the difference between the response value of the high-precision sample point and the predicted value of the low-precision proxy model as a scaling function sample point, wherein the difference is as follows:
in the formula (I), the compound is shown in the specification,is as followsThe difference between the individual high-precision sample point response values and the low-precision proxy model prediction values,、respectively linear correction constants for the low-precision proxy model,is as followsThe predicted value of the low-precision agent model at each high-precision sample point;
obtaining a set of scale function sample points:
in the formula (I), the compound is shown in the specification,X iDis shown asiThe sample point input values of the individual scaling functions,Y iLis shown asiThe point response values of the sample points of the individual scaling functions,n Dthe number of sample points is a scaling function.
In another embodiment, in step 4, the constructing a scale function based on the scale function sample point set specifically includes:
in the formula (I), the compound is shown in the specification,for any point in the design domainAnd a firstSample points of a scaling functionThe distance between the two or more of the two or more,;is as followsThe basis function coefficients of the individual scale function sample points;is as followsThe basis functions of the sample points of the individual scaling functions.
In another embodiment, the Gauss function is chosen as the basis function for the scaling function sample points as:
in the formula (I), the compound is shown in the specification,a shape parameter correction factor;is as followsThe shape parameters of the basis functions are:
in the formula (I), the compound is shown in the specification,is as followsThe distance between the sample point of the individual scaling functions to the farthest sample point;
will be provided withSubstituting the input value and the response value of the sample point of the scaling function into the scaling function to obtain the coefficient of the sample point of the scaling functionThe system of linear equations of (1) is:
in the formula (I), the compound is shown in the specification,and solving the equation set to obtain a basic function coefficient for a scale function smoothing factor to obtain a scale function.
In another embodiment, the average curvature of the scaling function in the design domain is:
in the formula:is the average curvature of the scaling function in the design domain,as a function of scale inLocal curvature of the surface.
In another embodiment, in step 4, the smoothness training of the scale function is performed with the goal of minimizing the average curvature of the scale function in the design domain, specifically:
in the formula (I), the compound is shown in the specification,for the maximum error between the multi-precision proxy model prediction value and the high-precision sample point response value,constraining an upper bound for errors
In another embodiment, in step 5, the multi-precision proxy model is:
in the formula (I), the compound is shown in the specification,f(x) Representing a multi-precision proxy model.
Compared with the prior art, the smooth scale approximate modeling method for the multi-precision data of the aircraft has the following beneficial technical effects:
1. aiming at the problems that the multi-precision agent model is inaccurate and the relevance between high-precision data and low-precision data is neglected in the optimization design process of the aircraft, the high-efficiency and accurate multi-precision agent model modeling method is provided. The hyper-parameters in the multi-precision proxy model are trained, the average curvature of the scaling function in the proxy model is optimized, the prediction precision of the multi-precision proxy model is effectively improved, the number of high-precision sample points required by modeling is reduced, the modeling efficiency is remarkably improved, and the subsequent optimization is guided;
2. according to the method, a low-precision proxy model is constructed by using low-precision simulation data, linear term parameters are added to correct the low-precision proxy model, and a scale function is constructed through a difference value between a high-precision sample point and a low-precision predicted value of a corresponding position of the high-precision sample point. In the process of establishing the scale function, the relevance of the high-precision model and the low-precision model is considered, the scale function is made to be as smooth as possible, namely, the linear term of the low-precision model, the shape parameter of the scale function and the smooth factor of the scale function are subjected to parameter training by taking the minimum average curvature of the scale function as a target, and finally, the prediction precision of the multi-precision proxy model is obviously improved, so that the high-efficiency and accurate construction of the multi-precision model of the aircraft is realized.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the structures shown in the drawings without creative efforts.
FIG. 1 is a flow chart of a smooth scale approximation modeling method for multi-precision data in an embodiment of the invention;
FIG. 2 is a schematic structural representation of an exemplary heavy launch vehicle concentration force dispersion cabin segment in an embodiment of the invention;
FIG. 3 is a schematic diagram illustrating an exemplary multi-zone skin partitioning method according to an embodiment of the present disclosure;
FIG. 4 is a schematic view of an exemplary variable cross-section main beam in an embodiment of the invention;
FIG. 5 is a schematic illustration of an exemplary non-uniform secondary beam/stringer layout in an embodiment of the present invention;
fig. 6 is a schematic diagram of an exemplary layout format of middle boxes and end boxes in the embodiment of the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that all the directional indicators (such as up, down, left, right, front, and rear … …) in the embodiment of the present invention are only used to explain the relative position relationship between the components, the movement situation, etc. in a specific posture (as shown in the drawing), and if the specific posture is changed, the directional indicator is changed accordingly.
In addition, the descriptions related to "first", "second", etc. in the present invention are only for descriptive purposes and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
In the present invention, unless otherwise expressly stated or limited, the terms "connected," "secured," and the like are to be construed broadly, and for example, "secured" may be a fixed connection, a removable connection, or an integral part; the connection can be mechanical connection, electrical connection, physical connection or wireless communication connection; they may be directly connected or indirectly connected through intervening media, or they may be connected internally or in any other suitable relationship, unless expressly stated otherwise. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
In addition, the technical solutions in the embodiments of the present invention may be combined with each other, but it must be based on the realization of those skilled in the art, and when the technical solutions are contradictory or cannot be realized, such a combination of technical solutions should not be considered to exist, and is not within the protection scope of the present invention.
The embodiment provides an efficient and accurate multi-precision agent model modeling method, and particularly provides a smooth scale approximate modeling method for multi-precision data of an aircraft, aiming at the problems that a multi-precision agent model is inaccurate and ignores the correlation between high-precision data and low-precision data in the optimization design process of the aircraft. The method comprises the steps of firstly constructing a low-precision proxy model by using low-precision simulation data, adding linear term parameters to correct the low-precision proxy model, constructing a scale function through a difference value between a high-precision sample point and a low-precision predicted value of a corresponding position of the high-precision sample point, and in the process of establishing the scale function, considering the relevance of the high-precision model and the low-precision model, enabling the scale function to be as smooth as possible, namely performing parameter training on the linear term of the low-precision model, the shape parameters of the scale function and the smooth factors of the scale function by taking the minimum average curvature of the scale function as a target, and finally obviously improving the prediction precision of the multi-precision proxy model, so that the high-efficiency and accurate construction of the multi-precision model of the aircraft is realized.
Referring to FIG. 1, in an implementation, a method for smooth scale approximate modeling of multi-precision data for an aircraft includes the following steps 1-5.
Step 1, obtaining design variables and a design domain of a given aircraft optimization design, and initially sampling to obtain high-precision sample points and low-precision sample points. The initial application process of the high-precision sample points and the low-precision sample points is specifically as follows:
respectively selecting by adopting a Latin hypercube sampling methodA low precision sampling point andeach high-precision sampling point is obtained by running a simulation model with corresponding precision at each sampling point respectivelyA low precision sample point andthe high-precision sample points of each sample are respectively as follows:
in the formula (I), the compound is shown in the specification,X iLis shown asiThe input value of one of the low-precision sample points,Y iLis shown asiThe response value of the sample point of low precision,X iHis shown asiThe input value of each sample point with high precision,Y iHis shown asiAnd (5) high-precision sample point response values.
in the formula (I), the compound is shown in the specification,f L(x) In order to be a low-precision proxy model,for any point in the design domainAnd a firstA low precision sample pointIs a distance therebetween, i.e.;Is as followsThe basis function coefficients of the low precision sample points,is as followsThe basis functions of the low precision sample points. Selecting a Gauss function as a basic function of the low-precision sample point, wherein the basic function comprises the following steps:
in the formula (I), the compound is shown in the specification,is as followsThe shape parameter of each basis function adopts a direct determination method, and comprises the following steps:
in the formula (I), the compound is shown in the specification,is as followsThe distance between the one low precision sample point to the farthest sample point,to design spatial dimensions;
will be provided withSubstituting the input value and the response value of each low-precision sample point into the low-precision proxy model to obtain the basis function coefficient of the low-precision sample pointThe system of linear equations of (1) is:
and solving the equation set to obtain a basis function coefficient to obtain the low-precision proxy model.
And 3, taking the difference between the high-precision sample point and the low-precision proxy model predicted value as a scale function sample point to obtain a scale function sample point set. In the specific implementation process:
calculating the difference between the response value of the high-precision sample point and the predicted value of the low-precision proxy model as a scaling function sample point, wherein the difference is as follows:
in the formula (I), the compound is shown in the specification,is as followsThe difference between the individual high-precision sample point response values and the low-precision proxy model prediction values,、respectively linear correction constants for the low-precision proxy model,is as followsThe predicted value of the low-precision agent model at each high-precision sample point;
obtaining a set of scale function sample points:
in the formula (I), the compound is shown in the specification,X iDis shown asiThe sample point input values of the individual scaling functions,Y iLis shown asiThe point response values of the sample points of the individual scaling functions,n Dthe number of sample points is a scaling function.
And 4, constructing a scale function based on the scale function sample point set to obtain the average curvature of the scale function in the design domain, and performing smoothness training on the scale function by taking the minimum average curvature of the scale function in the design domain as a target. Wherein the scaling function is specifically:
in the formula (I), the compound is shown in the specification,for any point in the design domainAnd a firstSample points of a scaling functionThe distance between the two or more of the two or more,;is as followsThe basis function coefficients of the individual scale function sample points;is as followsThe basis functions of the sample points of the individual scaling functions. Selecting a Gauss function as a basic function of a scale function sample point, wherein the shape parameter adopts a direct determination method and comprises the following steps:
in the formula (I), the compound is shown in the specification,a shape parameter correction factor;is as followsThe shape parameters of the basis functions are:
in the formula (I), the compound is shown in the specification,is as followsThe distance between the sample point of the individual scaling functions to the farthest sample point;
will be provided withSubstituting the input value and the response value of the sample point of the scaling function into the scaling function to obtain the coefficient of the sample point of the scaling functionThe system of linear equations of (1) is:
in the formula (I), the compound is shown in the specification,and solving the equation set to obtain a basic function coefficient for a scale function smoothing factor to obtain a scale function.
The average curvature of the scaling function in the design domain is found from its local curvature at each low precision sample point location, i.e.:
in the formula:is the average curvature of the scaling function in the design domain,as a function of scale inLocal curvature of the surface.
Since the mean curvature of the scale function in the design domain is a multi-dimensional problem, the hessian matrix is usedSolving (Hessian Matrix), wherein the specific implementation process comprises the following steps:
the hessian matrix is of the form:
the form of the scaling function can be written as:
in the formula (I), the compound is shown in the specification,as a gaussian function, can be written as:
is composed ofAnd a firstThe distance between sample points of the individual scaling functions can be written as:
then scaling functionIn thatIs to be treated asThe first order partial derivatives of the dimensions are:
according to the above process, the product is obtainedAt any low precision sample pointHessian matrix ofAnd then solving the local curvature of the calibration function at the position to finally obtain the average curvature of the calibration function in the design domain.
In the above steps 1 to 4, there areFour hyper-parameters, by training four parameters, such that the scaling functionWith minimal mean curvature and due to the introduction of a smoothing factorIf the multi-precision agent model cannot accurately pass through each high-precision sample point, the maximum error between the predicted value of the multi-precision agent model and the response value of the high-precision sample point is taken as a constraint, and then the mathematical description of parameter training is as follows:
in the formula (I), the compound is shown in the specification,the maximum error between the multi-precision agent model predicted value and the high-precision sample point response value is obtained;an upper bound is constrained for error.
And 5, finally obtaining a multi-precision proxy model, which is as follows:
in the formula (I), the compound is shown in the specification,f(x) Representing a multi-precision proxy model.
The modeling method in this embodiment is further explained below by taking the structural performance prediction of the launch vehicle cabin segment as an example.
The concentrated force diffusion cabin section is used as a main connecting cabin section of the main binding device and plays a role in transferring and diffusing the thrust of the booster to a core stage, and the axial pressure bearing capacity is a main performance index for designing the structure. For the problem of structural optimization of the concentrated force diffusion cabin, static analysis is used as a low-precision analysis model, implicit kinetic analysis is used as a high-precision analysis model, and the problem is used for calculating the bearing capacity of the axle load. Fig. 2 is a schematic diagram, in which a portion in fig. 2 is a schematic diagram of a structure of a concentrated force diffusion cabin of a conventional single bundled booster, and b portion is a schematic diagram of a structure of a concentrated force diffusion cabin of a double-layer bundled booster.
The implementation process of the modeling method in the embodiment is as follows:
1. an optimization target and design variables are given, a proxy model is established by taking the bearing load bearing capacity as a response value, and 50 parameters such as the skin thickness of different regions of the concentrated force diffusion cabin section and the design parameters of a main beam, an auxiliary beam, a stringer and a middle frame are taken as design variables, wherein the design variables are as follows:
skin multi-zone variable thickness design:
according to the bearing characteristics, skins with different thicknesses are designed at different force bearing parts, the region division form is shown in fig. 3, and the value range is shown in table 1.
TABLE 1 initial design and value range of different areas of skin thickness
The variable cross section design of the main beam:
as a main bearing part of the concentrated force diffusion cabin section, a main beam adopts a variable cross-section design, and the structural parameters are shown in fig. 4, wherein part a in fig. 4 is the top dimension, part b is the bottom dimension, and the value range is shown in table 2.
TABLE 2 initial design and value range of relevant parameters of variable section girder
And (3) proportional layout design of secondary beams and stringers:
the distances between the auxiliary beams and the stringers are distributed according to an equal ratio number series to realize the non-uniform layout, and three equal ratio coefficients are introducedThe spatial positions of the secondary beam and the stringer are respectively characterized, the mathematical description is as follows, the layout diagram is shown in fig. 5, and the design range is shown in table 3.
Wherein the content of the first and second substances,the central angle representing the corresponding position of the adjacent ribs (secondary beams or stringers).Representing the number of secondary beams arranged in the main diffusion B region;representing the number of secondary beams arranged in the main diffusion C region;representing the number of stringers placed in the non-main diffusion D region.Represents rounding up;indicating a rounding down.
TABLE 3 initial design and value range of parameters related to secondary beams and stringers
Layout design of a middle frame and an end frame:
the layout of the middle frame and the end frame is shown in FIG. 6, and the design range is shown in Table 4.
TABLE 4 initial design and value range of related parameters of middle frame and end frame
2. Design of experimentsRespectively selecting by adopting a Latin hypercube methodOne low precision sample point and 50 high precision sample points.
3. Performing calculation by using the static analysis model as a low-precision simulation modelA low precision sample point response value. 4. And calculating 50 high-precision sample point response values by using an implicit kinetic analysis model as a high-precision simulation model.
6. The difference between the high-precision response value and the low-precision proxy model prediction value is calculated at the 50 high-precision sample point positions.
7. And establishing a proxy model by using the difference values of the 50 high-precision sample points and the corresponding positions.
8. And reducing the mean curvature of the scale function model by using a parameter training method.
9. And combining the low-precision proxy model with the scaling function to obtain the final multi-precision proxy model.
And (3) reselecting 50 high-precision sample points by using a Latin hypercube method, calculating the response value by using an implicit dynamics simulation model, comparing the result with the constructed multi-precision proxy model, and calculating to obtain the relative Root Mean Square Error (RMSE). Number of low precision sample pointsThe results are shown in the following table 5 when taken together:
TABLE 5 comparison of modeling accuracy between the method and the conventional method
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention, and all modifications and equivalents of the present invention, which are made by the contents of the present specification and the accompanying drawings, or directly/indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A smooth scale approximate modeling method for multi-precision data of an aircraft is characterized by comprising the following steps:
step 1, obtaining design variables and a design domain, and initially sampling to obtain high-precision sample points and low-precision sample points;
step 2, establishing a low-precision proxy model based on the low-precision sample points;
step 3, taking the difference between the high-precision sample point and the low-precision proxy model predicted value as a scale function sample point to obtain a scale function sample point set;
step 4, constructing a scale function based on the scale function sample point set to obtain the average curvature of the scale function in the design domain, and performing smoothness training on the scale function by taking the minimum average curvature of the scale function in the design domain as a target;
and 5, obtaining a multi-precision agent model.
2. The method for modeling an aircraft according to claim 1, wherein in step 1, the initial sampling is performed to obtain high-precision sample points and low-precision sample points, specifically:
respectively selecting by adopting a Latin hypercube sampling methodA low precision sampling point andeach high-precision sampling point is provided with a simulation model with corresponding precisionForm (a) to obtainA low precision sample point andthe high-precision sample points of each sample are respectively as follows:
in the formula (I), the compound is shown in the specification,X iLis shown asiThe input value of one of the low-precision sample points,Y iLis shown asiThe response value of the sample point of low precision,X iHis shown asiThe input value of each sample point with high precision,Y iHis shown asiAnd (5) high-precision sample point response values.
3. The method according to claim 2, wherein in step 2, the low-precision proxy model is:
in the formula (I), the compound is shown in the specification,f L(x) In order to be a low-precision proxy model,for any point in the design domainAnd a firstA low precision sample pointIs a distance therebetween, i.e.;Is as followsThe basis function coefficients of the low precision sample points,is as followsThe basis functions of the low precision sample points.
4. The method according to claim 3, wherein the Gauss function is selected as the basis function of the low-precision sample points as follows:
in the formula (I), the compound is shown in the specification,is as followsThe shape parameters of the basis functions are:
in the formula (I), the compound is shown in the specification,is as followsThe distance between the one low precision sample point to the farthest sample point,to design spatial dimensions;
will be provided withSubstituting the input value and the response value of each low-precision sample point into the low-precision proxy model to obtain the basis function coefficient of the low-precision sample pointAnd solving the equation set to obtain a basis function coefficient to obtain the low-precision proxy model.
5. The method for smooth-scale approximate modeling of multi-precision data for an aircraft according to claim 3 or 4, wherein step 3 specifically comprises:
calculating the difference between the response value of the high-precision sample point and the predicted value of the low-precision proxy model as a scaling function sample point, wherein the difference is as follows:
in the formula (I), the compound is shown in the specification,is as followsThe difference between the individual high-precision sample point response values and the low-precision proxy model prediction values,、respectively linear correction constants for the low-precision proxy model,is as followsThe predicted value of the low-precision agent model at each high-precision sample point;
obtaining a set of scale function sample points:
in the formula (I), the compound is shown in the specification,X iDis shown asiThe sample point input values of the individual scaling functions,Y iLis shown asiThe point response values of the sample points of the individual scaling functions,n Dthe number of sample points is a scaling function.
6. The method according to claim 5, wherein in step 4, the step of constructing the scale function based on the scale function sample point set comprises:
in the formula (I), the compound is shown in the specification,for any point in the design domainAnd a firstSample points of a scaling functionThe distance between the two or more of the two or more,;is as followsThe basis function coefficients of the individual scale function sample points;is as followsThe basis functions of the sample points of the individual scaling functions.
7. The method according to claim 6, wherein the Gauss function is selected as the basis function of the scale function sample points as follows:
in the formula (I), the compound is shown in the specification,a shape parameter correction factor;is as followsThe shape parameters of the basis functions are:
in the formula (I), the compound is shown in the specification,is as followsThe distance between the sample point of the individual scaling functions to the farthest sample point;
will be provided withSubstituting the input value and the response value of the sample point of the scaling function into the scaling function to obtain the coefficient of the sample point of the scaling functionThe system of linear equations of (1) is:
9. The method according to claim 8, wherein in step 4, the smooth scale approximation modeling of the multi-precision data for the aircraft is performed by performing smoothness training on the scale function with the goal of minimizing the average curvature of the scale function in the design domain, specifically:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210154400.2A CN114218686B (en) | 2022-02-21 | 2022-02-21 | Multi-precision data smooth scale approximate modeling method for aircraft |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210154400.2A CN114218686B (en) | 2022-02-21 | 2022-02-21 | Multi-precision data smooth scale approximate modeling method for aircraft |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114218686A true CN114218686A (en) | 2022-03-22 |
CN114218686B CN114218686B (en) | 2022-05-10 |
Family
ID=80708985
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210154400.2A Active CN114218686B (en) | 2022-02-21 | 2022-02-21 | Multi-precision data smooth scale approximate modeling method for aircraft |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114218686B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106295043A (en) * | 2016-08-18 | 2017-01-04 | 中国人民解放军国防科学技术大学 | A kind of parameter determination method of gaussian radial basis function agent model |
WO2021008628A1 (en) * | 2019-07-12 | 2021-01-21 | 深圳市道通智能航空技术有限公司 | Aircraft control method and aircraft |
CN112434448A (en) * | 2021-01-27 | 2021-03-02 | 中国人民解放军国防科技大学 | Proxy model constraint optimization method and device based on multipoint adding |
US20210141981A1 (en) * | 2019-11-08 | 2021-05-13 | Dalian University Of Technology | Structural non-gradient topology optimization method based on sequential kriging surrogate model |
CN113657029A (en) * | 2021-08-10 | 2021-11-16 | 北京理工大学 | Efficient approximate optimization method for aircraft driven by heterogeneous data |
CN113673027A (en) * | 2021-07-28 | 2021-11-19 | 北京航空航天大学 | Hypersonic aircraft aerodynamic load optimization design method based on proxy model |
-
2022
- 2022-02-21 CN CN202210154400.2A patent/CN114218686B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106295043A (en) * | 2016-08-18 | 2017-01-04 | 中国人民解放军国防科学技术大学 | A kind of parameter determination method of gaussian radial basis function agent model |
WO2021008628A1 (en) * | 2019-07-12 | 2021-01-21 | 深圳市道通智能航空技术有限公司 | Aircraft control method and aircraft |
US20210141981A1 (en) * | 2019-11-08 | 2021-05-13 | Dalian University Of Technology | Structural non-gradient topology optimization method based on sequential kriging surrogate model |
CN112434448A (en) * | 2021-01-27 | 2021-03-02 | 中国人民解放军国防科技大学 | Proxy model constraint optimization method and device based on multipoint adding |
CN113673027A (en) * | 2021-07-28 | 2021-11-19 | 北京航空航天大学 | Hypersonic aircraft aerodynamic load optimization design method based on proxy model |
CN113657029A (en) * | 2021-08-10 | 2021-11-16 | 北京理工大学 | Efficient approximate optimization method for aircraft driven by heterogeneous data |
Non-Patent Citations (2)
Title |
---|
王志祥: "基于序列径向基函数的运载火箭蒙皮桁条结构轻质优化", 《国防科技大学学报》 * |
魏锋涛等: "基于径向基函数代理模型的改进协同优化方法", 《西安理工大学学报》 * |
Also Published As
Publication number | Publication date |
---|---|
CN114218686B (en) | 2022-05-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP7058902B2 (en) | High-speed coordinated optimization method for plate-wound shell structure of hybrid fiber composite material | |
Sobieski et al. | Collaborative optimization using response surface estimation | |
Martins et al. | A coupled-adjoint sensitivity analysis method for high-fidelity aero-structural design | |
Ghoreyshi et al. | Accelerating the numerical generation of aerodynamic models for flight simulation | |
CN112016167B (en) | Aircraft aerodynamic shape design method and system based on simulation and optimization coupling | |
CN104866692A (en) | Aircraft multi-objective optimization method based on self-adaptive agent model | |
Chittick et al. | An asymmetric suboptimization approach to aerostructural optimization | |
CN115496006A (en) | High-precision numerical simulation method suitable for hypersonic aircraft | |
CN109255141B (en) | Optimization method for cross section shape of forward conceptual design of automobile body | |
Kanazaki et al. | Multi-objective aerodynamic exploration of elements' setting for high-lift airfoil using Kriging model | |
CN114218686B (en) | Multi-precision data smooth scale approximate modeling method for aircraft | |
Bhardwaj et al. | Computational fluid dynamics/computational structural dynamics interaction methodology for aircraft wings | |
Dalle et al. | Adjustments and Uncertainty Quantification for SLS Aerodynamic Sectional Loads | |
US8473267B2 (en) | Computer-aided method for a cost-optimized calculation of variable distributions over an aerodynamic surface | |
CN114564787A (en) | Bayesian optimization method, device and storage medium for target-related airfoil design | |
Kim et al. | Enhancement of class of adjoint design methods via optimization of parameters | |
CN111310328B (en) | Point adding update optimization method of gradient enhancement collaborative proxy model | |
CN111611646A (en) | Rapid design method for aerodynamic configuration of aircraft forebody | |
CN109635346B (en) | Reliability analysis method of mechanical connection structure | |
CN114676639A (en) | Aircraft aerodynamic shape optimization method, device and medium based on neural network | |
CN108984864A (en) | Higher-dimension complex aircraft model characterizing method based on Gaussian process variance evaluation | |
CN113657029A (en) | Efficient approximate optimization method for aircraft driven by heterogeneous data | |
Bekemeyer et al. | Nonlinear unsteady reduced order models based on computational fluid dynamics for gust loads predictions | |
CN112182739A (en) | Aircraft structure non-probability credible reliability topological optimization design method | |
CN102385651B (en) | Method for rapidly obtaining speed of transonic flutter of composite airfoil surface |
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 |