Multi-fidelity deep neural network surrogate model for aerodynamic shape optimization

2021 ◽  
Vol 373 ◽  
pp. 113485
Author(s):  
Xinshuai Zhang ◽  
Fangfang Xie ◽  
Tingwei Ji ◽  
Zaoxu Zhu ◽  
Yao Zheng
Keyword(s):  
Neural Network ◽  
Shape Optimization ◽  
Surrogate Model ◽  
Deep Neural Network ◽  
Aerodynamic Shape Optimization ◽  
Aerodynamic Shape

Aerodynamic Shape Optimization Design of the Swept Wing Based on the Kriging Surrogate Model

2013 ◽  
Vol 444-445 ◽  
pp. 1277-1282
Author(s):  
Dan Wang ◽  
Jun Qiang Bai ◽  
Jun Hua ◽  
Zhi Wei Sun ◽  
Lei Qiao
Keyword(s):  
Shape Optimization ◽  
Surrogate Model ◽  
Optimization Design ◽  
Cfd Simulation ◽  
Free Form ◽  
Design System ◽  
Aerodynamic Shape Optimization ◽  
Swept Wing ◽  
Aerodynamic Shape ◽  
Optimization Search

The aerodynamic shape optimization design system was established in this paper. In the system, the RANS equation was used for solving the flowing; the free form deformation (FFD) method was used for the geometry parameterization, and the genetic algorithm was used for the optimization search. For the reducing of the time cost, the Kriging model was used for the surrogate model instead of the CFD simulation during the optimization design. The aerodynamic shape design of a swept wing was presented which used the system, and the result indicated that the 14% drag coefficient was reduced at the cruise conditions, which proved the validity of the system.

Surrogate model for aerodynamic shape optimization of a tractor-trailer in crosswinds

2012 ◽  
Vol 226 (10) ◽  
pp. 1325-1339 ◽  
Author(s):  
Xu Gong ◽  
Zhengqi Gu ◽  
Zhenlei Li
Keyword(s):  
Fluid Dynamics ◽  
Computational Fluid Dynamics ◽  
Shape Optimization ◽  
Drag Coefficient ◽  
Surrogate Model ◽  
Aerodynamic Drag ◽  
Aerodynamic Shape Optimization ◽  
Aerodynamic Shape ◽  
Design Variables ◽  
Aerodynamic Drag Coefficient

A surrogate model-based aerodynamic shape optimization method applied to the wind deflector of a tractor-trailer is presented in this paper. The aerodynamic drag coefficient of the tractor-trailer with and without the wind deflector subjected to crosswinds is analyzed. The numerical results show that the wind deflector can decrease drag coefficient. Four parameters are used to describe the wind deflector geometry: width, length, height, and angle. A 30-level design of experiments study using the optimal Latin hypercube method was conducted to analyze the sensitivity of the design variables and build a database to set up the surrogate model. The surrogate model was constructed based on the Kriging interpolation technique. The fitting precision of the surrogate model was examined using computational fluid dynamics and certified using a surrogate model simulation. Finally, a multi-island genetic algorithm was used to optimize the shape of the wind deflector based on the surrogate model. The tolerance between the results of the computational fluid dynamics simulation and the surrogate model was only 0.92% when using the optimal design variables, and the aerodynamic drag coefficient decreased by 4.65% compared to the drag coefficient of the tractor-trailer installed with the original wind deflector. The effect of the optimal shape of the wind deflector was validated by computational fluid dynamics and wind tunnel experiment.

Aerodynamic Shape Optimization and Knowledge Mining of Centrifugal Fans Using Simulated Annealing Coupled With a Neural Network

2006 ◽  
Author(s):  
Kazuyuki Sugimura
Keyword(s):  
Neural Network ◽  
Simulated Annealing ◽  
Shape Optimization ◽  
Leading Edge ◽  
Optimization Method ◽  
Collaborative Optimization ◽  
Design Parameters ◽  
Centrifugal Impeller ◽  
Aerodynamic Shape Optimization ◽  
Aerodynamic Shape

An aerodynamic shape optimization method suitable for “inexpensive” centrifugal impellers and diffusers has been developed. The shapes are parameterized using non-uniform rational B-spline curves with special attention being paid to the blade’s edge profiles. A hybrid algorithm combining simulated annealing and a neural network is employed for collaborative optimization. The simulated annealing and neural network take turns in controlling the optimization processes, not only for maximizing the efficiency of global exploration, but also for minimizing the risks of automation failures or of reaching an incorrect optimum. A statistical analysis was also conducted using the neural network to extract design knowledge. By applying the proposed method to a centrifugal impeller and diffuser design problem, we obtained innovative shapes for the leading edge of the impeller and the trailing edge of the diffuser. Important design parameters related to the new shapes were identified through the design space analysis.

Sign in / Sign up

Export Citation Format

Share Document