Accompanying aerodynamic configuration optimization design method based on point adding strategy
By constructing a dynamic downgrade model in the aerodynamic appearance optimization design, predicting the initial value of the iteration of the accompanying field, the problem of large amount of calculation of the accompanying field solving is solved, the efficiency of the aerodynamic appearance optimization design is improved, and it has wide applicability.
Patent Information
- Application Number
- CN202410990595.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-08
AI Technical Summary
In pneumatic appearance optimization design, the calculation amount of accompanying field solution is too large, which affects the optimization design efficiency.
A dynamic downgrade model is constructed using a point-added strategy, and the iterative initial value is predicted by accompanying field samples, reducing the amount of calculation of accompanying equations and improving optimization design efficiency.
It significantly accelerates the solution of accompanying equations in the aerodynamic appearance optimization design process, improves design efficiency, and has wide applicability, and can be combined with any accompanying equation solver without modification.
Smart Images

Figure CN118917082A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of aerodynamic shape optimization design, and in particular relates to a method for optimizing aerodynamic shape based on a point adding strategy. Background Art
[0002] In aerodynamic shape optimization design, the gradient method is the most common type of algorithm in engineering. It constructs a search direction with the help of the gradient of the objective function with respect to the shape parameters to find the final shape that meets the optimal aerodynamic performance. A key challenge of the gradient method is how to quickly and accurately obtain the gradient of the objective function with respect to the shape parameters. The calculation efficiency of the gradient significantly affects the efficiency of the optimization design.
[0003] The finite difference method is a traditional method for solving gradients in a relatively simple way. It obtains the change value of the objective function by making a small perturbation to the design variables. The ratio of the change in the objective function to the perturbation of the design variables is approximately equal to the gradient of the objective function with respect to the design variables. However, the accuracy of this method is sensitive to the size of the perturbation, and the computational cost increases with the number of design variables. The complex variable method is a type of gradient solution method for complex functions mentioned in "Using Complex Variables to Estimate Derivatives of Real Functions" published in the journal "SIAM Review" in 1998. It is the same as the finite difference method. It also solves the gradient by perturbing each design variable. Similarly, its computational complexity is proportional to the dimension of the design variable. In addition, the adjoint method is currently a more efficient method for solving gradients in the field of aerodynamic optimization design. It can obtain the gradient of the objective function with respect to any number of design variables by solving the adjoint equation only once. However, according to the article "Dynamic-mode-decomposition-based gradient prediction for adjoint-basedaerodynamic shape optimization" published in the journal "Aerospace Science and Technology" in April 2024, the computational cost of solving the adjoint field is equivalent to that of solving the flow field. This shows that in a gradient-based aerodynamic shape optimization design process, each gradient acquisition still requires a computational cost equivalent to that of solving the flow field, so the computational cost of the adjoint field accounts for a considerable proportion in the optimization process. Therefore, the development of an efficient adjoint field prediction method is of great significance to improving the optimization efficiency in engineering. Summary of the invention
[0004] In order to solve the problem of excessive computational complexity in adjoint field solutions in the current aerodynamic shape optimization design based on adjoint equations, the present invention adopts a point-adding strategy to construct a dynamic order reduction model to predict the initial value of adjoint field iteration to accelerate the adjoint field iterative solution. This method is independent of the solution method itself and can be very conveniently grafted onto any adjoint equation solver. It can also be combined with other adjoint equation accelerated convergence measures and has a wide range of applicability.
[0005] To achieve the above object, the present invention provides the following technical solutions:
[0006] An adjoint aerodynamic shape optimization design method based on the point-adding strategy collects adjoint field samples in the aerodynamic shape optimization design process based on the adjoint equation, constructs a dynamic adjoint field reduction model, and uses the model to predict the adjoint field corresponding to the aerodynamic shape in the optimization process. The predicted value is then used as the initial value of the adjoint field iteration, thereby improving the efficiency of aerodynamic shape optimization design by accelerating the solution of the adjoint equation.
[0007] Furthermore, the method specifically comprises the following steps:
[0008] Step 1: Determine the incoming flow state (incoming flow Mach number and angle of attack), optimization objectives and constraints (such as geometric constraints and aerodynamic constraints) of the object to be optimized, and generate a grid suitable for CFD calculation.
[0009] Step 2: Based on the initial shape data of the object, the CST parameterization method is used to obtain the initial design variable vector x0.
[0010] Step 3: Use dynamic mesh technology to generate the computational mesh corresponding to the current design variables, use the RANS flow field control equation to calculate the corresponding flow field solution, and obtain aerodynamic parameters such as drag coefficient and lift coefficient.
[0011] Step 4: If it is the first step of optimization, set all the values in the initial adjoint vector Λ to 0, and solve the adjoint equation according to the adjoint method using the same pseudo-time advancement method as the flow field solution to obtain the objective function gradient with respect to the design variable and the first adjoint field sample Λ1;
[0012] If it is the second step of optimization, set the initial adjoint vector equal to Λ1, solve the adjoint equation according to the adjoint method, obtain the gradient of the objective function with respect to the design variable and the second adjoint field sample Λ2;
[0013] If it is the nth step of optimization, n>2, add the adjoint field samples obtained in the previous step to the sample matrix, reduce the original adjoint equation to a non-homogeneous linear system of equations, use the least squares method to solve the non-homogeneous linear system of equations, and then obtain the initial adjoint vector of the nth step, continue to solve the adjoint equation, and obtain the objective function with respect to the design variable gradient and the nth adjoint field sample Λ n ;
[0014] Step 5: For constrained optimization design, SUMT technology is used for optimization, in which the objective function and aerodynamic constraints on the gradient of the design variables are solved using the method in step 4 until convergence.
[0015] Furthermore, the adjoint equation is: Where R is the flow field residual vector, Q is the flow field conservation variable vector, and J is the objective function.
[0016] Furthermore, in step 4, the sample matrix is A=[Λ1, Λ2,…, Λ n-1 ], assuming Λ n =Ab, where b is an n-1 dimensional column vector, and the original adjoint equation is reduced to a non-homogeneous linear system of equations The least square method is used to solve the non-homogeneous linear equations to obtain b, and then according to Λ n =Ab obtains the initial adjoint vector of the nth step, continues to solve the adjoint equation, obtains the objective function with respect to the design variable gradient and the nth adjoint field sample Λ n .
[0017] Furthermore, it is only necessary to construct a reduced-order model through the adjoint field samples in the aerodynamic shape optimization design process based on the adjoint equation, so as to predict the initial value of the adjoint field iteration corresponding to the aerodynamic shape.
[0018] Furthermore, it can be grafted onto any adjoint equation solver without making any changes to the solution method.
[0019] Compared with the prior art, the present invention has the following beneficial effects:
[0020] (1) The present invention only needs to construct a reduced-order model through the companion field samples in the aerodynamic shape optimization design process based on the companion equation, so as to more accurately predict the initial value of the companion field iteration corresponding to the aerodynamic shape, greatly accelerate the solution of the companion equation in the optimization process, and thus effectively improve the efficiency of the aerodynamic shape optimization design based on the companion equation.
[0021] (2) The present invention can be easily grafted onto any adjoint equation solver without making any changes to the solution method, and has a wide range of applicability. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 It is the comparison of pressure coefficient distribution on the airfoil surface before and after optimization.
[0023] Figure 2 This is the appearance comparison before and after optimization.
[0024] Figure 3 It is the change history of the total number of pseudo-time iterations required to solve the equation during the optimization process.
[0025] Figure 4 It is the convergence process of the drag coefficient.
[0026] Figure 5 This is the appearance comparison before and after optimization.
[0027] Figure 6 It is the change history of the total number of pseudo-time iterations required to solve the equation during the optimization process.
[0028] Figure 7 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0029] The present invention will be further described in detail below in conjunction with the accompanying drawings.
[0030] The present invention is a method for optimizing the aerodynamic shape based on a point-adding strategy. The method collects samples of the accompanying field in the process of optimizing the aerodynamic shape based on the accompanying equation, constructs a dynamic accompanying field reduction model, predicts the accompanying field corresponding to the aerodynamic shape in the optimization process through the model, and then uses the predicted value as the initial value of the iteration to accelerate the solution of the accompanying equation, thereby improving the efficiency of the aerodynamic shape optimization design based on the accompanying equation.
[0031] The present invention obtains more accurate initial values of adjoint field iterations through existing adjoint field samples to accelerate the solution of adjoint equations in the process of aerodynamic shape optimization design. Figure 7 As shown, the specific implementation steps of this method are as follows:
[0032] (a) Determine the incoming flow state (incoming flow Mach number and angle of attack), optimization objectives and constraints (such as geometric constraints and aerodynamic constraints) of the object to be optimized, and generate a grid suitable for CFD calculation.
[0033] (b) Based on the initial shape data of the object, the CST parameterization method (Class-shape function transformation) is used to obtain the initial design variable vector x0.
[0034] (c) The dynamic mesh technology is used to generate the computational mesh corresponding to the current design variables, and the RANS flow field control equation is used to calculate the corresponding flow field solution to obtain aerodynamic parameters such as the drag coefficient and lift coefficient.
[0035] (d) If this is the first step of optimization, set all values in the initial adjoint vector Λ to 0, and solve the adjoint equation using the same pseudo-time advancement method as the flow field solution according to the adjoint method. (where R is the flow field residual vector, Q is the flow field conservation variable vector, and J is the objective function), obtain the objective function gradient with respect to the design variable and the first adjoint field sample Λ1;
[0036] If it is the second step of optimization, set the initial adjoint vector equal to Λ1 and solve the adjoint equation according to the adjoint method Obtain the objective function gradient with respect to the design variable and the second adjoint field sample Λ2;
[0037] If the optimization is for the nth step (n>2), the companion field samples obtained in the previous step are added to the sample matrix A = [Λ1, Λ2, ..., Λ n-1 ], assuming Λ n =Ab, where b is an n-1 dimensional column vector, the original adjoint equation can be reduced to a non-homogeneous linear system of equations The least square method is used to solve the above equations to obtain b, and then according to Λ n =Ab obtains the initial adjoint vector of step n and continues to solve the adjoint equation Get the objective function with respect to the design variable gradient and the nth adjoint field sample Λ n .
[0038] (e) For constrained optimization design, the SUMT technique (sequential unconstrained minimization technique) is used for optimization, where the objective function and the aerodynamic constraints on the gradient of the design variables are solved using the method in step (d) until convergence.
[0039] Embodiment 1:
[0040] The NACA0012 airfoil was selected for drag reduction optimization design using an accompanying aerodynamic shape optimization design method based on a point adding strategy according to the present invention. The design state is free flow Mach number Ma=0.7, angle of attack α=2.0°, and the constraints are that the lift coefficient after optimization is not less than the initial value, and the airfoil area is not less than 90% of the initial value. Eight CST parameters were used as design variables, and the initial lift coefficient and drag coefficient were 0.419 and 0.0117, respectively. The initial optimization step length was set to 10% of the first initial CST parameter, and the optimal solution was found after 9 optimizations. The comparison of pressure coefficient distribution and shape changes on the airfoil surface before and after optimization are shown in the figure below. Figure 1 , Figure 2 As shown in Figure 2, the leading edge of the airfoil becomes thinner after optimization, resulting in a significant weakening of the shock wave intensity and a 94% reduction in the drag coefficient. The total number of pseudo-time iterations required to solve the equations during the optimization process is shown in Figure 2. Figure 3 As shown, compared with the original adjoint method, the method of the present invention can reduce the total number of pseudo-time iterations required for solving the adjoint equation by 41%, and the optimization efficiency is significantly improved.
[0041] Embodiment 2:
[0042] The laminar drag reduction optimization design is performed using an accompanying aerodynamic shape optimization design method based on a point adding strategy of the present invention, with the NACA0012 airfoil as the initial shape, the design state is the free flow Mach number Ma=0.5, the angle of attack α=2.0° and the Reynolds number Re=5000, and the constraints are that the lift coefficient after optimization is not less than the initial value, and the airfoil area is not less than 90% of the initial value. Eight CST parameters are used as design variables, and the initial lift coefficient and drag coefficient are 0.038 and 0.0584 respectively. The initial optimization step length is set to 5% of the first initial CST parameter. After 37 optimizations, the optimal solution is found, such as Figure 4 As shown in the figure, the drag coefficient is reduced by 7.7% after optimization, and the lift-to-drag ratio is improved by nearly 100%. Figure 5 As shown in , the leading edges of the upper and lower boundaries of the airfoil are thinner, while the thickness of the lower airfoil increases. The total number of pseudo-time iterations required to solve the equations during the optimization process is shown in Figure 6 As shown, compared with the original adjoint method, the method of the present invention can reduce the total number of pseudo-time iterations required for solving the adjoint equation by 68%, and the optimization efficiency is significantly improved.
[0043] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A method for optimizing the aerodynamic shape based on the point adding strategy, characterized in that: The accompanying field samples in the aerodynamic shape optimization design process based on the adjoint equation are collected, and a dynamic accompanying field reduction model is constructed. The accompanying field corresponding to the aerodynamic shape in the optimization process is predicted by this model. The predicted value is then used as the initial value of the accompanying field iteration. The efficiency of aerodynamic shape optimization design is improved by accelerating the solution of the adjoint equation.
2. The method for optimizing the aerodynamic shape based on the point adding strategy according to claim 1 is characterized in that: The specific steps include: Step 1: Determine the incoming flow state, optimization objectives and constraints of the object to be optimized, and generate a grid suitable for CFD calculation. Step 2: Based on the initial shape data of the object, the CST parameterization method is used to obtain the initial design variable vector x0. Step 3: Use dynamic mesh technology to generate the computational mesh corresponding to the current design variables, use the RANS flow field control equation to calculate the corresponding flow field solution, and obtain aerodynamic parameters such as drag coefficient and lift coefficient. Step 4: If it is the first step of optimization, set all the values in the initial adjoint vector Λ to 0, and solve the adjoint equation according to the adjoint method using the same pseudo-time advancement method as the flow field solution to obtain the objective function gradient with respect to the design variable and the first adjoint field sample Λ1; If it is the second step of optimization, set the initial adjoint vector equal to Λ1, solve the adjoint equation according to the adjoint method, obtain the gradient of the objective function with respect to the design variable and the second adjoint field sample Λ2; If it is the nth step of optimization, n>2, add the adjoint field samples obtained in the previous step to the sample matrix, reduce the original adjoint equation to a non-homogeneous linear system of equations, use the least squares method to solve the non-homogeneous linear system of equations, and then obtain the initial adjoint vector of the nth step, continue to solve the adjoint equation, and obtain the objective function with respect to the design variable gradient and the nth adjoint field sample ΛΛ n ; Step 5: For constrained optimization design, SUMT technology is used for optimization, in which the objective function and aerodynamic constraints on the gradient of the design variables are solved using the method in step 4 until convergence.
3. The method for optimizing the aerodynamic shape based on the point adding strategy according to claim 2 is characterized in that: The adjoint equation is: Where R is the flow field residual vector, Q is the flow field conservation variable vector, and J is the objective function.
4. The method for optimizing the aerodynamic shape based on the point adding strategy according to claim 2 is characterized in that: In step 4, the sample matrix is A = [Λ1, Λ2, ..., Λ n-1 ], assuming Λ n =Ab, where b is an n-1 dimensional column vector, and the original adjoint equation is reduced to a non-homogeneous linear system of equations The least square method is used to solve the non-homogeneous linear equations to obtain b, and then according to Λ n =Ab to obtain the initial adjoint vector of the nth step, continue to solve the adjoint equation, obtain the gradient of the objective function with respect to the design variable and the nth adjoint field sample Λ n .
5. The method for optimizing the aerodynamic shape based on the point adding strategy according to claim 1 is characterized in that: It is only necessary to construct a reduced-order model through the adjoint field samples in the aerodynamic shape optimization design process based on the adjoint equation, so as to predict the initial value of the adjoint field iteration corresponding to the aerodynamic shape.
6. The method for optimizing the aerodynamic shape based on the point adding strategy according to claim 1 is characterized in that: It can be grafted onto any adjoint equation solver without making any changes to the solution method.
