CN110334449A - A kind of aerofoil profile Fast design method based on online agent model algorithm - Google Patents
A kind of aerofoil profile Fast design method based on online agent model algorithm Download PDFInfo
- Publication number
- CN110334449A CN110334449A CN201910612597.8A CN201910612597A CN110334449A CN 110334449 A CN110334449 A CN 110334449A CN 201910612597 A CN201910612597 A CN 201910612597A CN 110334449 A CN110334449 A CN 110334449A
- Authority
- CN
- China
- Prior art keywords
- aerofoil profile
- agent model
- sampled point
- design
- algorithm
- 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.)
- Pending
Links
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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Aviation & Aerospace Engineering (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Automation & Control Theory (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of aerofoil profile Fast design methods based on online agent model algorithm, comprising the following steps: 1) determines wing range of variables；2) sampled point is initialized, the CFD simulation result of corresponding sampled point is calculated；3) online agent model is constructed using decision tree；4) based on the agent model and multi-objective optimization algorithm after fitting, the Parameters Optimal Design of aircraft wing is carried out.The design of agent model technology, multiple-objection optimization technology and aircraft wing is fused together by the present invention, can obtain preferable aircraft wing design scheme with faster speed, lower cost.
Description
Technical field
The invention belongs to airplane design technical fields, and in particular to a kind of aircraft wing design method is suitable for general fly
Airfoil type design.
Background technique
Wing design is the important link of airplane design, and the design of aerofoil profile is the primary basis of wing design.According to aircraft
Different purposes and flying condition, need for the suitable aerofoil profile of airplane design.Such as the aerofoil profile of transporter wing is thicker, to provide
The aerofoil profile of biggish lift and stable performance, supersonic plane is generally relatively thin, to reduce resistance.Airfoil Design directly influences
The development of follow-up link, thus engineering is very high to the design efficiency requirement of aerofoil profile in practice, that is, requires within the time short as far as possible
Design aerofoil profile of good performance.
Wind tunnel experiment is essential link in Airfoil Design.Early stage Airfoil Design is mainly carried out by wind tunnel experiment,
I.e. Airfoil Design personnel rule of thumb design aerofoil profile, make model, experiment are blowed, in wind-tunnel then to assess aerofoil profile
Aeroperformance.Wind tunnel experiment passes through the airflow state that artificial environment simulated flight device encounters in flight course, to obtain
The performance data of aircraft is the detection most basic and most reliable means of Airfoil Aerodynamic Performance.However, in order to carry out wind-tunnel reality
It tests, needs to put into a large amount of time and funds, thus the Airfoil Design period of early stage is long, at high cost.
Modern Airfoil Design more carries out on computers.With the development of computing technique and computer technology, pass through meter
Calculation machine is possibly realized to simulate and analyze the flowing of fluid, and here it is Fluid Mechanics Computation (Computational Fluid
Dynamics, brief note make CFD).Airfoil Design person designs aerofoil profile on computers, passes through the flow fields environment pair of computer simulation
The aeroperformance of aerofoil profile is detected, and the use of CFD approach reduces dependence of the Airfoil Design process to wind tunnel experiment.In recent years
Coming, CFD is calculated as the important means of assessment Airfoil Aerodynamic Performance, instead of wind tunnel experiment a large amount of during Airfoil Design,
To improve the speed and efficiency of Airfoil Design.
Optimal Design Method is to resolve establishing equation function according to flow field, using the geometric parameter of aerofoil profile as the defeated of function
Enter, using aeroperformance as the output of function, geometric parameter is solved by optimization tool.It can carry out initial aerofoil profile
It quickly and efficiently modifies, it is made to meet design requirement.During optimized design frequently with a kind of method be of overall importance
Intelligent search algorithm.Airfoil Design is carried out by intelligent search algorithm, its essence is using the powerful computing capability of computer,
It is scanned in wide aerofoil profile space, to find the aerofoil profile met the requirements.Intelligent search algorithm is often found in aerofoil profile space
As soon as new aerofoil calculates aeroperformance using CFD approach.Due to new aerofoil enormous amount calculative in search process,
It is very time-consuming for thus carrying out Airfoil Optimization using intelligent search algorithm, it is difficult to requirement needed for reaching engineering.
Summary of the invention
[goal of the invention]
In order to overcome the above-mentioned deficiencies of the prior art, it is fast to provide a kind of aerofoil profile based on online agent model algorithm by the present invention
Fast design method is calculated Airfoil Aerodynamic Performance using agent model, relatively time-consuming CFD is replaced to calculate, improved and searched based on intelligence
The speed of the optimized design process of rope algorithm, designs more preferably aerofoil profile within the shorter time.
[technical solution]
The present invention is implemented with the following technical solutions.
A kind of aerofoil profile Fast design method based on online agent model algorithm, flow chart is as shown in Figure 1, include following step
It is rapid:
(1) target and design variable for determining aircraft wing, establish the Model for Multi-Objective Optimization of aircraft wing parameter；
(2) initial samples point set is generated；
(3) simulation calculation is carried out to all sampled points generated in step (2) one by one and obtains aeroperformance as target value,
Constitute the initial population of multi-objective optimization algorithm；
(4) simulation result of analytical procedure (3) constructs agent model；
(5) real simulation in step (1) is replaced to calculate using the agent model constructed in step (4), using multiple target
Optimization algorithm generates new sampled point and assesses；
(6) the agent model assessed value obtained according to step (5) is selected from new sampled point using agent model management strategy
It selects a sampled point and carries out simulation calculation；
(7) environmental selection is carried out to all sampled points using multi-objective optimization algorithm, obtains preferable solution and forms new population；
(8) the population recruitment agent model obtained using step (7)；
(9) it if simulation calculation number reaches preset times, executes step (10), otherwise executes step (5) again；
(10) the non-dominant disaggregation in the population for obtaining step (7) is as final optimization pass as a result, parameter optimization terminates.
Further, step (1) the following steps are included:
(a) Airfoil Design variable is determined: using maximum opposite camber, the relative position of maximum camber and maximum relative thickness
Three variables as Airfoil Design.
It is an aerofoil profile as shown in Figure 3, the vertical segment in aerofoil profile between lower aerofoil is referred to as the thickness of aerofoil profile, wherein longest
Line segment be referred to as maximum gauge, the length is t for note；The line at the midpoint of thickness is referred to as middle camber line, between middle camber line and wing chord
Line segment is referred to as camber, wherein longest line segment is known as maximum camber, the length is f for note；The position of maximum camber and it is up-front away from
From being denoted as xf。
Use following values as the variable of Airfoil Design:
Maximum opposite camber:
The relative position of maximum camber:
Maximum relative thickness:
(b) it determines Airfoil Design target: selecting to maximize the lift coefficient of aerofoil profile and minimize resistance coefficient as aerofoil profile
Two targets of design.
When aircraft is toward when the flight of front, air-flow (incoming flow) is separated in aerofoil profile leading edge, a part of upper surface for flowing through aerofoil profile, and one
Part flows through the lower surface of aerofoil profile.Since aerofoil profile upper surface is thicker, the air-flow for flowing through aerofoil profile upper surface is squeezed, air-flow velocity
Accelerate, internal pressure reduces；Aerofoil profile lower surface is more flat, then flows through extruding degree that the air-flow of lower surface is subject to relatively
Small, air-flow velocity is relatively slow, internal pressure is relatively large.Since air flow pressure is smaller near aerofoil profile upper surface, lower surface is attached
Nearly air flow pressure is larger, and the pressure difference of air-flow produces the lift that aerofoil profile is held up upwards near upper lower aerofoil.
Use lift coefficient CLThe lift of aerofoil profile is described:
L is the lift of aerofoil profile in formula, and c is chord length, and ρ is to carry out current density, V∞For speed of incoming flow.Lift coefficient is to indicate
One dimensionless number of profile lift size, it is related with the shape of aerofoil profile, unrelated with the physics size of aerofoil profile.
Incoming flow can also generate resistance to aerofoil profile, using resistance coefficient C other than generating lift to aerofoil profileDIndicate aerofoil profile
Drag characteristic:
D is the resistance of aerofoil profile in formula, and c is chord length, and ρ is to carry out current density, V∞For speed of incoming flow.Resistance coefficient is to indicate
One dimensionless number of profile drag size, it is related with the shape of aerofoil profile, unrelated with the physics size of aerofoil profile.
(c) restriction range of design parameter is determined:
Maximum opposite camber:
The relative position of maximum camber:
Maximum relative thickness:
The generation sampled point of step (2) is according to method of samplings such as Latin Hypercube Sampling, Grid Samplings, in step (1)
A certain number of parameter combinations are generated in the parameter section of middle setting.
Simulation calculation is carried out one by one to all sampled points in step (3), comprising the following steps:
(a) grid dividing is carried out to aerofoil profile using grid dividing software；
(b) variable of each sampled point for using CFD software to be obtained according to step (2) solves Airfoil Aerodynamic Performance as adopting
Sampling point target value；
(c) variable of the sampled point obtained using step (2) and the target value being calculated constitute multi-objective optimization algorithm
Initial population.
The building agent model of step (4), comprising the following steps:
(a) Pareto sequence is carried out to the target value that step (3) obtains, the one kind that is divided into for being one by ranking results, label
For " 1 " class, ranking results are not one to be divided into one kind, are labeled as " 0 " class；
(b) as feature, corresponding category is denoted as target, building decision tree the sampling point variable for obtaining step (2)
Disaggregated model is as agent model.
Wherein, decision tree is binary tree, and the value of internal node feature is "Yes" and "No", and left branch is that value is "Yes"
Branch, right branch be value be "No" branch.Decision tree is equivalent to recursively two points of each features, and feature space is divided
For limited unit, and determine on these units the probability distribution of prediction, i.e., the condition exported under conditions of input is given
Probability distribution.Criterion is minimized to carry out feature selecting using Gini coefficient in the present invention.Assuming that there is K class, sample point belongs to
The probability of kth class is pk, then the Gini index of probability distribution is defined as: If sample set D is divided into D according to some feature A1, D2Two parts, then under conditions of feature A,
The Gini index of set D is defined as:Gini index Gini (D, A) table
Show the uncertainty of the data set A of feature A different grouping.Gini index value is bigger, and the uncertainty of sample set is also bigger.
Therefore, the present invention determines the optimal cut-off of some feature in decision tree by Gini index.
Step (5) the following steps are included:
(a) cross and variation is carried out using the individual in current population, generates new sampled point；
Wherein, each individual includes one group of parameter combination, the corresponding simulation result of parameter, and multiple individuals constitute kind
Group；
(b) simulation calculation is replaced using currently available agent model, obtained new sampled point parameter is input to agency
The class label of corresponding parameter point is calculated in model.
It is the class label obtained according to step (5) that one sampled point of selection in step (6), which carries out simulation calculation, to mark
Label are that the sampled point of " 1 " class is randomly ordered, select the parameter of first sampled point, solve Airfoil Aerodynamic Performance using CFD software
Target value as this sampled point.
Environmental selection in step (7) is to apply the environmental selection strategy in NSGAII algorithm in current population and step
(6) in the solution selected, all solutions that are calculated of non-dominated ranking and crowding distance are carried out according to the target value to all solutions
The solution of partial order, the forward Population Size of selected and sorted forms new population.
The renewal agency model of step (8), comprising the following steps:
(a) target value of sampled point carries out Pareto sequence in the population obtained to step (7), is one by ranking results
It is divided into one kind, is labeled as " 1 " class, ranking results is not one to be divided into one kind, are labeled as " 0 " class；
(b) as feature, obtained correspondence category is denoted as mesh the variable of sampled point in the population for obtaining step (7)
Mark, on-line training renewal agency model.
Proceed to step (9) if when simulation calculation number reach preset number, carry out step (10), otherwise repeat to hold
Row step (5).
Individual target value carries out non-dominated ranking in the population that step (10) obtains step (8), in ranking results
Non-domination solution is final as a result, the variate-value of non-domination solution is final parameter optimization as a result, the target value of non-domination solution is
Corresponding optimal aeroperformance.
[beneficial effect]
Compared with prior art, the aerofoil profile Fast design method provided by the invention based on online agent model algorithm, passes through
Improvement to the solution procedure of the Optimal Design Method for the intelligent optimization algorithm for using global search, is replaced using agent model
Originally extremely time-consuming simulation calculation, greatly accelerates the solution efficiency of aerofoil profile parameter designing, saves a large amount of time and gold
Money cost, the aerofoil profile for being quite suitable for aircraft quickly design the process of especially initial designs, while being existed by agent model
Line updates, and parameter space reduces the error of agent model near optimized parameter, excellent so as to obtain high-precision parameter
Change result.
Detailed description of the invention
Fig. 1 is a kind of aerofoil profile Fast design method flow chart based on online agent model algorithm of the present invention.
Fig. 2 is typical aerofoil profile figure of the present invention.
Fig. 3 is airfoil geometry Parameter Map of the present invention.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples, it should be understood that these examples are only used for
It is bright the present invention rather than limit the scope of the invention, after the present invention has been read, those skilled in the art are to of the invention
The modification of various equivalent forms falls within the application range as defined in the appended claims.
The present invention proposes a kind of aerofoil profile Fast design method based on online agent model algorithm.Specific example step is such as
Under:
(1) target and design variable for determining aircraft wing establish aircraft wing parameter Model for Multi-Objective Optimization.Step is such as
Under:
(a) Airfoil Design variable is determined: using maximum opposite camber, the relative position of maximum camber and maximum relative thickness
Three variables as Airfoil Design.
It is an aerofoil profile as shown in Figure 3, the vertical segment in aerofoil profile between lower aerofoil is referred to as the thickness of aerofoil profile, wherein longest
Line segment be referred to as maximum gauge, the length is t for note；The line at the midpoint of thickness is referred to as middle camber line, between middle camber line and wing chord
Line segment is referred to as camber, wherein longest line segment is known as maximum camber, the length is f for note；The position of maximum camber and it is up-front away from
From being denoted as xf。
Use following values as the variable of Airfoil Design:
Maximum opposite camber:
The relative position of maximum camber:
Maximum relative thickness:
(b) it determines Airfoil Design target: selecting to maximize the lift coefficient of aerofoil profile and minimize resistance coefficient as aerofoil profile
Two targets of design.
Use lift coefficient CLThe lift of aerofoil profile is described:
Using resistance coefficient CDIndicate the drag characteristic of aerofoil profile:
L is the lift of aerofoil profile in formula, and D is the resistance of aerofoil profile, and c is chord length, and ρ is to carry out current density, V∞For speed of incoming flow.
(c) restriction range of design parameter is determined:
Maximum opposite camber:
The relative position of maximum camber:
Maximum relative thickness:
(2) a certain number of ginsengs are generated in the parameter section for adding random perturbation to be arranged in step (1) using Grid Sampling
Array is closed, and initial samples point as shown in Table 1 is obtained.
1 initial samples point variable of table
| Maximum opposite camber (%) | Maximum camber relative position (%) | Maximum relative thickness (%) |
| 1 | 3 | 11 |
| 1 | 4 | 14 |
| 1 | 5 | 13 |
| 1.12 | 4.79 | 13.56 |
| 1.34 | 3.77 | 11.53 |
| 1.66 | 3.83 | 13.67 |
| 2 | 3 | 12 |
| 2 | 4 | 15 |
| 2 | 5 | 11 |
| 2.04 | 4.2 | 17.05 |
| 2.29 | 5.15 | 15.1 |
| 2.63 | 3.02 | 16.72 |
1 initial samples point variable (Continued) of table
| Maximum opposite camber (%) | Maximum camber relative position (%) | Maximum relative thickness (%) |
| 2.67 | 2.96 | 12.43 |
| 2.94 | 4.92 | 11.28 |
| 2.99 | 3.09 | 14.17 |
| 3 | 3 | 16 |
| 3 | 4 | 11 |
| 3 | 5 | 13 |
| 3.51 | 4.12 | 13.33 |
| 3.88 | 3.75 | 11.61 |
| 3.93 | 3.2 | 17.07 |
| 4 | 3 | 11 |
| 4 | 4 | 13 |
| 4 | 5 | 16 |
| 4.08 | 5.03 | 11.01 |
| 4.26 | 3.4 | 12.04 |
| 4.53 | 4.75 | 12.43 |
| 4.59 | 2.76 | 13.21 |
| 4.77 | 4.69 | 14.34 |
| 5 | 3 | 11 |
| 5 | 4 | 12 |
| 5 | 5 | 17 |
| 5.76 | 4.92 | 13.11 |
| 6 | 3 | 14 |
| 6 | 4 | 15 |
| 6 | 5 | 12 |
(3) simulation calculation is carried out to all sampled points generated in step (2) one by one and obtains aeroperformance as target value,
Constitute the initial population of multi-objective optimization algorithm.Steps are as follows:
(a) grid dividing is carried out to aerofoil profile using grid dividing software；
(b) variable of each sampled point for using CFD software to be obtained according to step (2) solves Airfoil Aerodynamic Performance as adopting
Sampling point target value is set as 0 ° of incoming flow generator rotor angle in CFD software, free stream Mach number 0.76, and obtained target value is as shown in table 2；
2 initial samples point attribute of table
| Maximum opposite camber (%) | Maximum camber relative position (%) | Maximum relative thickness (%) | Lift coefficient | Resistance coefficient |
| 1 | 3 | 11 | 0.215062 | 0.019171 |
| 1 | 4 | 14 | 0.207816 | 0.030013 |
| 1 | 5 | 13 | 0.219388 | 0.025379 |
2 initial samples point attribute (Continued) of table
| Maximum opposite camber (%) | Maximum camber relative position (%) | Maximum relative thickness (%) | Lift coefficient | Resistance coefficient |
| 1.12 | 4.79 | 13.56 | 0.244236 | 0.028508 |
| 1.34 | 3.77 | 11.53 | 0.252698 | 0.022809 |
| 1.66 | 3.83 | 13.67 | 0.335567 | 0.034829 |
| 2 | 3 | 12 | 0.377917 | 0.031254 |
| 2 | 4 | 15 | 0.401921 | 0.047074 |
| 2 | 5 | 11 | 0.421032 | 0.025063 |
| 2.04 | 4.2 | 17.05 | 0.393182 | 0.061071 |
| 2.29 | 5.15 | 15.1 | 0.498069 | 0.051526 |
| 2.63 | 3.02 | 16.72 | 0.45774 | 0.069519 |
| 2.67 | 2.96 | 12.43 | 0.501469 | 0.043817 |
| 2.94 | 4.92 | 11.28 | 0.616105 | 0.040014 |
| 2.99 | 3.09 | 14.17 | 0.551507 | 0.060685 |
| 3 | 3 | 16 | 0.52453 | 0.072357 |
| 3 | 4 | 11 | 0.586277 | 0.040289 |
| 3 | 5 | 13 | 0.63151 | 0.051762 |
| 3.51 | 4.12 | 13.33 | 0.66955 | 0.065608 |
| 3.88 | 3.75 | 11.61 | 0.72581 | 0.063425 |
| 3.93 | 3.2 | 17.07 | 0.613867 | 0.098513 |
| 4 | 3 | 11 | 0.726039 | 0.063392 |
| 4 | 4 | 13 | 0.736595 | 0.074996 |
| 4 | 5 | 16 | 0.7106 | 0.091773 |
| 4.08 | 5.03 | 11.01 | 0.813003 | 0.064062 |
| 4.26 | 3.4 | 12.04 | 0.766116 | 0.076099 |
| 4.53 | 4.75 | 12.43 | 0.835683 | 0.083228 |
| 4.59 | 2.76 | 13.21 | 0.782953 | 0.093123 |
| 4.77 | 4.69 | 14.34 | 0.812293 | 0.099781 |
| 5 | 3 | 11 | 0.863718 | 0.089827 |
| 5 | 4 | 12 | 0.86639 | 0.093418 |
| 5 | 5 | 17 | 0.751281 | 0.117117 |
| 5.76 | 4.92 | 13.11 | 0.918576 | 0.114913 |
| 6 | 3 | 14 | 0.898093 | 0.134271 |
| 6 | 4 | 15 | 0.859578 | 0.133198 |
| 6 | 5 | 12 | 0.96122 | 0.114541 |
(c) variable of the sampled point obtained using step (2) and the target value being calculated constitute multi-objective optimization algorithm
Initial population.
(4) simulation result of analytical procedure (3) constructs agent model.Steps are as follows:
(a) Pareto sequence is carried out to the target value that step (3) obtains, the one kind that is divided into for being one by ranking results, label
For " 1 " class, ranking results are not one to be divided into one kind, are labeled as " 0 " class, the class label of initial samples is as shown in table 3；
3 initial samples point class label of table
| Maximum opposite camber (%) | Maximum camber relative position (%) | Maximum relative thickness (%) | Class label |
| 1 | 3 | 11 | 1 |
| 1 | 4 | 14 | 0 |
| 1 | 5 | 13 | 0 |
| 1.12 | 4.79 | 13.56 | 0 |
| 1.34 | 3.77 | 11.53 | 1 |
| 1.66 | 3.83 | 13.67 | 0 |
| 2 | 3 | 12 | 0 |
| 2 | 4 | 15 | 0 |
| 2 | 5 | 11 | 1 |
| 2.04 | 4.2 | 17.05 | 0 |
| 2.29 | 5.15 | 15.1 | 0 |
| 2.63 | 3.02 | 16.72 | 0 |
| 2.67 | 2.96 | 12.43 | 0 |
| 2.94 | 4.92 | 11.28 | 1 |
| 2.99 | 3.09 | 14.17 | 0 |
| 3 | 3 | 16 | 0 |
| 3 | 4 | 11 | 0 |
| 3 | 5 | 13 | 1 |
| 3.51 | 4.12 | 13.33 | 0 |
| 3.88 | 3.75 | 11.61 | 0 |
| 3.93 | 3.2 | 17.07 | 0 |
| 4 | 3 | 11 | 1 |
| 4 | 4 | 13 | 0 |
| 4 | 5 | 16 | 0 |
| 4.08 | 5.03 | 11.01 | 1 |
| 4.26 | 3.4 | 12.04 | 0 |
| 4.53 | 4.75 | 12.43 | 1 |
| 4.59 | 2.76 | 13.21 | 0 |
| 4.77 | 4.69 | 14.34 | 0 |
| 5 | 3 | 11 | 1 |
3 initial samples point class label (Continued) of table
| Maximum opposite camber (%) | Maximum camber relative position (%) | Maximum relative thickness (%) | Class label |
| 5 | 4 | 12 | 1 |
| 5 | 5 | 17 | 0 |
| 5.76 | 4.92 | 13.11 | 0 |
| 6 | 3 | 14 | 0 |
| 6 | 4 | 15 | 0 |
| 6 | 5 | 12 | 1 |
(b) the sampling point variable for obtaining step (2) is as feature, and obtained correspondence category is denoted as constructing for target
Decision-Tree Classifier Model is as agent model.
(5) cross and variation is carried out using the individual in current population, generates new sampled point.Use building in step (4)
Agent model replace step (1) in real simulation calculate, the new sampled point parameter of generation is input to agent model, calculate
Obtain the class label of corresponding parameter point.
(6) the agent model assessed value obtained according to step (5) is selected from new sampled point using agent model management strategy
It selects a sampled point and carries out simulation calculation.It is randomly ordered to the sampled point that label is " 1 " class, select the ginseng of first sampled point
Number uses CFD software to solve Airfoil Aerodynamic Performance as the target value of this sampled point.
(7) environmental selection is carried out to all sampled points using multi-objective optimization algorithm, the environment in NSGA II algorithm is selected
Strategy is selected to apply in the solution that current population and step (6) are selected, according to all solutions target value carry out non-dominated ranking and
The solution of the partial ordering relation that all solutions are calculated of crowding distance, the forward Population Size number of selected and sorted forms new kind
Group.
(8) the population recruitment agent model obtained using step (7).Steps are as follows:
(a) target value of sampled point carries out Pareto sequence in the population obtained to step (7), is one by ranking results
It is divided into one kind, is labeled as " 1 " class, ranking results is not one to be divided into one kind, are labeled as " 0 " class；
(b) as feature, obtained correspondence category is denoted as mesh the variable of sampled point in the population for obtaining step (7)
Mark, on-line training renewal agency model.
(9) it if simulation calculation number reaches 300 times, executes step (10), otherwise circulation executes step (5) and arrives step
(8) process.
(10) target value individual in the population obtained to step (7) carries out non-dominated ranking, the non-branch in ranking results
With solution to be final as a result, the variate-value of non-domination solution is final parameter optimization as a result, the target value of non-domination solution is corresponding
Optimal aeroperformance.Parameter optimization terminates.
The present invention is changed by the solution procedure of the Optimal Design Method to the intelligent optimization algorithm for using global search
Into, originally extremely time-consuming simulation calculation is replaced using agent model, greatly accelerates the solution efficiency of aerofoil profile parameter designing, section
A large amount of time and monetary cost are saved, the aerofoil profile for being quite suitable for aircraft quickly designs the process of especially initial designs, together
When by the online updating of agent model, parameter space reduces the error of agent model near optimized parameter, so as to
Obtain high-precision parameter optimization result.
Claims (7)
1. a kind of aerofoil profile Fast design method based on online agent model algorithm, it is characterised in that the following steps are included:
(1) target and design variable for determining aircraft wing, establish the Model for Multi-Objective Optimization of aircraft wing parameter；
(2) initial samples point set is generated；
(3) simulation calculation is carried out to all sampled points generated in step (2) one by one and obtains aeroperformance as target value, is constituted
The initial population of multi-objective optimization algorithm；
(4) simulation result of analytical procedure (3) constructs agent model；
(5) real simulation in step (1) is replaced to calculate using the agent model constructed in step (4), using multiple-objection optimization
Algorithm generates new sampled point and assesses；
(6) the agent model assessed value obtained according to step (5) selects one from new sampled point using agent model management strategy
A sampled point carries out simulation calculation；
(7) environmental selection is carried out to all sampled points using multi-objective optimization algorithm, obtains preferable solution and forms new population；
(8) the population recruitment agent model obtained using step (7)；
(9) it if simulation calculation number reaches preset times, executes step (10), otherwise executes step (5) again；
(10) non-domination solution in the population for obtaining step (7) is as final optimization pass as a result, parameter optimization terminates；
By above step, the present invention provides a kind of aerofoil profile Fast design methods based on online agent model algorithm, significantly
The solution efficiency for accelerating aerofoil profile parameter designing saves a large amount of time and monetary cost, and the aerofoil profile for being very suitable for aircraft is fast
Speed design accesses high-precision parameter optimization as a result, solving and carries out aerofoil profile parameter optimization using global intelligent search algorithm
In the process calculation amount it is huge and it is time-consuming too long the problem of.
2. a kind of aerofoil profile Fast design method based on online agent model algorithm, feature according to claim (1)
It is:
The step (1) the following steps are included:
(a) determine Airfoil Design variable: use maximum opposite camber, the relative position of maximum camber and maximum relative thickness as
Three variables of Airfoil Design；
(b) it determines Airfoil Design target: selecting to maximize the lift coefficient of aerofoil profile and minimize resistance coefficient as Airfoil Design
Two targets；
(c) restriction range of design parameter is determined.
3. a kind of aerofoil profile Fast design method based on online agent model algorithm, feature according to claim (1)
It is:
The step (4) the following steps are included:
(a) target value obtained to step (3) carries out Pareto sequence, is one to be divided into one kind by ranking results, is labeled as " 1 "
Class, ranking results are not one to be divided into one kind, are labeled as " 0 " class；
(b) as feature, corresponding category is denoted as target, building decision tree classification the sampling point variable for obtaining step (2)
Model is as agent model.
4. a kind of aerofoil profile Fast design method based on online agent model algorithm, feature according to claim (1)
It is:
The step (5) the following steps are included:
(a) cross and variation is carried out using the individual in current population, generates new sampled point；
(b) simulation calculation is replaced using currently available agent model, new sampled point parameter is input to agent model, is calculated
To the class label of corresponding parameter point.
5. a kind of aerofoil profile Fast design method based on online agent model algorithm, feature according to claim (1)
It is:
It is the class label obtained according to step (5) that one sampled point of selection in step (6), which carries out simulation calculation, to label
Sampled point for " 1 " class is randomly ordered, selects the parameter of first sampled point, solves Airfoil Aerodynamic Performance using CFD software and makees
For the target value of this sampled point.
6. a kind of aerofoil profile Fast design method based on online agent model algorithm, feature according to claim (1)
It is:
The step (8) the following steps are included:
(a) target value of sampled point carries out Pareto sequence in the population obtained to step (7), is one to be divided by ranking results
One kind, is labeled as " 1 " class, and ranking results are not one to be divided into one kind, is labeled as " 0 " class；
(a) as feature, corresponding category is denoted as target, online instruction the variable of sampled point in the population for obtaining step (7)
Practice renewal agency model.
7. a kind of aerofoil profile Fast design method based on online agent model algorithm, feature according to claim (1)
It is:
Individual target value carries out non-dominated ranking in the population obtained in step (10) to step (7), in ranking results
Non-domination solution is final as a result, the variate-value of non-domination solution is final parameter optimization as a result, the target value of non-domination solution is
Corresponding optimal aeroperformance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910612597.8A CN110334449A (en) | 2019-07-04 | 2019-07-04 | A kind of aerofoil profile Fast design method based on online agent model algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910612597.8A CN110334449A (en) | 2019-07-04 | 2019-07-04 | A kind of aerofoil profile Fast design method based on online agent model algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110334449A true CN110334449A (en) | 2019-10-15 |
Family
ID=68143361
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910612597.8A Pending CN110334449A (en) | 2019-07-04 | 2019-07-04 | A kind of aerofoil profile Fast design method based on online agent model algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110334449A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111348175A (en) * | 2020-01-13 | 2020-06-30 | 北京航空航天大学 | Special airfoil profile matched with high-universality large-loading small-sized unmanned aerial vehicle |
| CN111737928A (en) * | 2020-06-24 | 2020-10-02 | 西北工业大学 | Airfoil type steady aerodynamic optimization design method considering geometric uncertainty factors |
| CN112329140A (en) * | 2020-10-30 | 2021-02-05 | 北京理工大学 | Method for optimizing aerodynamics of variant aircraft based on improved position vector expectation improvement degree |
| WO2021077855A1 (en) * | 2019-10-24 | 2021-04-29 | 南京航空航天大学 | Helicopter rotor airfoil determination method and system |
| WO2021142916A1 (en) * | 2020-01-15 | 2021-07-22 | 深圳大学 | Proxy-assisted evolutionary algorithm-based airfoil optimization method and apparatus |
| CN114297934A (en) * | 2021-12-30 | 2022-04-08 | 无锡雪浪数制科技有限公司 | Model parameter parallel simulation optimization method and device based on proxy model |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106021808A (en) * | 2016-06-07 | 2016-10-12 | 西北工业大学 | Low span chord ratio aerofoil type designing method considering three-dimensional effect |
-
2019
- 2019-07-04 CN CN201910612597.8A patent/CN110334449A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106021808A (en) * | 2016-06-07 | 2016-10-12 | 西北工业大学 | Low span chord ratio aerofoil type designing method considering three-dimensional effect |
Non-Patent Citations (1)
| Title |
|---|
| 孙俊峰等: "基于Kriging模型的旋翼翼型优化设计研究", 《空气动力学学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021077855A1 (en) * | 2019-10-24 | 2021-04-29 | 南京航空航天大学 | Helicopter rotor airfoil determination method and system |
| JP2022508097A (en) * | 2019-10-24 | 2022-01-19 | 南京航空航天大学 | Method and system for determining the airfoil of a helicopter rotor |
| EP3926507A4 (en) * | 2019-10-24 | 2022-05-18 | Nanjing University of Aeronautics and Astronautics | Helicopter rotor airfoil determination method and system |
| JP7092421B2 (en) | 2019-10-24 | 2022-06-28 | 南京航空航天大学 | Method and system for determining the airfoil of a helicopter rotor |
| US11423201B2 (en) | 2019-10-24 | 2022-08-23 | Nanjing University Of Aeronautics And Astronautics | Method and system for determining helicopter rotor airfoil |
| CN111348175A (en) * | 2020-01-13 | 2020-06-30 | 北京航空航天大学 | Special airfoil profile matched with high-universality large-loading small-sized unmanned aerial vehicle |
| WO2021142916A1 (en) * | 2020-01-15 | 2021-07-22 | 深圳大学 | Proxy-assisted evolutionary algorithm-based airfoil optimization method and apparatus |
| CN111737928A (en) * | 2020-06-24 | 2020-10-02 | 西北工业大学 | Airfoil type steady aerodynamic optimization design method considering geometric uncertainty factors |
| CN111737928B (en) * | 2020-06-24 | 2022-01-25 | 西北工业大学 | Airfoil type steady aerodynamic optimization design method considering geometric uncertainty factors |
| CN112329140A (en) * | 2020-10-30 | 2021-02-05 | 北京理工大学 | Method for optimizing aerodynamics of variant aircraft based on improved position vector expectation improvement degree |
| CN112329140B (en) * | 2020-10-30 | 2022-07-15 | 北京理工大学 | Method for optimizing aerodynamics of variant aircraft based on expected improvement degree of improved position vector |
| CN114297934A (en) * | 2021-12-30 | 2022-04-08 | 无锡雪浪数制科技有限公司 | Model parameter parallel simulation optimization method and device based on proxy model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110334449A (en) | A kind of aerofoil profile Fast design method based on online agent model algorithm | |
| Li et al. | Review of design optimization methods for turbomachinery aerodynamics | |
| CN107480335B (en) | A kind of hypersonic vehicle Iterative Design method | |
| Benini et al. | High-lift multi-element airfoil shape and setting optimization using multi-objective evolutionary algorithms | |
| CN112084727A (en) | Transition prediction method based on neural network | |
| Fan | An inverse design method of diffuser blades by genetic algorithms | |
| CN112597702A (en) | Pneumatic modeling generation type confrontation network model training method based on radial basis function | |
| Sánchez-Moreno et al. | Robustness of optimisation algorithms for transonic aerodynamic design | |
| Angelini et al. | Identification of losses in turbomachinery with machine learning | |
| CN117454721B (en) | Wind power plant wake superposition effect evaluation method and medium based on digital simulation experiment | |
| Yao et al. | Prediction of aerodynamic forces at the tip of the compressor blades based on multi-scale 1DCNN combined with CBAM | |
| Zhu et al. | Design of an RBF surrogate model for low Reynolds number airfoil based on transfer learning | |
| CN105760478A (en) | Large-scale distributed data clustering method based on machine learning | |
| CN108664742A (en) | The Multipurpose Optimal Method of nacelle Parametric designing | |
| Allen et al. | Helicopter rotor blade multiple-section optimization with performance considerations | |
| CN117313576A (en) | Bayesian optimization method for analyzing importance degree of airfoil physical quantity | |
| Marx et al. | A machine learning based approach of performance estimation for high-pressure compressor airfoils | |
| Rui et al. | A novel approach for modelling of an injector powered transonic wind tunnel | |
| Wilkinson et al. | Approximating urban wind interference | |
| Thurman et al. | Comparison of Prediction Modeling Methodologies for Aeroacoustic Characterization of Hovering sUAS Rotors | |
| CN113947028A (en) | RBCC health management method based on XGboost and Datawig machine learning | |
| Ivanov et al. | Fast Compressor Map Computation by Utilizing Support Vector Machine and Response Surface Approximation | |
| Lamb | Probabilistic performance-based geometric tolerancing of compressor blades | |
| Shi et al. | An adaptive sequential sampling strategy-based multi-objective optimization of aerodynamic configuration for a tandem-wing UAV via a surrogate model | |
| Ryan | Robust multi-objective optimization of hypersonic vehicles under asymmetric roughness-induced boundary-layer transition |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191015 |