A multi-task learning method for large discrepant aerodynamic data

ZHANG Jun; ZHANG Guangbo; CHENG Yanqing; HU Liwei; XIANG Yu; WANG Wenyong

doi:10.7638/kqdlxxb-2021.0222

English Version Journal Login

您当前的位置：

首页 >

文章列表页 >

A multi-task learning method for large discrepant aerodynamic data

更新时间：2026-05-28

- A multi-task learning method for large discrepant aerodynamic data
- Vol. 40, Issue 6, Pages: 64-72(2022)
- 作者机构：
  
  1. 电子科技大学,成都,611731
  2. 中国空气动力研究与发展中心,绵阳,621000
- 作者简介：
- 基金信息：
- DOI：10.7638/kqdlxxb-2021.0222
  CLC： TP181
- Published：2022
- 稿件说明：
移动端阅览
张骏, 张广博, 程艳青, 等. 一种气动大差异性数据多任务学习方法[J]. 空气动力学学报, 2022, 40(6): 64−72.

ZHANG Jun, ZHANG Guangbo, CHENG Yanqing, et al. A multi-task learning method for large discrepant aerodynamic data[J]. 2022, 40(6): 64-72.
张骏, 张广博, 程艳青, 等. 一种气动大差异性数据多任务学习方法[J]. 空气动力学学报, 2022, 40(6): 64−72. DOI： 10.7638/kqdlxxb-2021.0222

ZHANG Jun, ZHANG Guangbo, CHENG Yanqing, et al. A multi-task learning method for large discrepant aerodynamic data[J]. 2022, 40(6): 64-72. DOI： 10.7638/kqdlxxb-2021.0222.

摘要

与传统方法相比，基于深度学习的空气动力学建模方法建模速度快、精度高。但是传统深度学习采用的全连接神经网络或卷积神经网络往往没有考虑输入数据本身的差异对预测结果的影响，而飞行器的外形特征参数和飞行状态参数在数据类型上存在较大差异。在同时使用这两种参数预测气动特性时，如果忽视这些差异性，预测结果的精度势必会损失。受到多任务学习和集群网络方法的启发，提出了一种基于多任务学习的翼型外形参数与飞行状态参数联合建模方法：大差异性多任务学习网络（LD-MTL）。该方法首先将数据集划分为多个任务，随后将整个学习网络分为多个集群，分别根据不同的任务学习所预测的气动性能相关知识，最终对每个集群所学习到的相关知识进行融合，得到预测结果。通过对比实验，证明了在进行气动大差异性数据建模时，本文提出的结构能更好地反映数据差异性对模型预测精度的影响程度，有更高的预测精度，且能对此差异性进行量化分析。

Abstract

Compared to traditional aerodynamic modeling methods

the deep-learning-based aerodynamic modeling has higher modeling efficiency and precision. However

previous deep learning methods with fully connected network (FCN) or convolutional neural network (CNN) ignored the influence of input data discrepancies. The shape characteristic parameters and flow state parameters of aircraft are greatly discrepant with different types

which will cause different degrees of impacts on the predicted results. When these two types of parameters are used to predict aerodynamic characteristics together

if we ignore the data discrepancy between them

the accuracy of prediction will be lost. Inspired by the multi-task learning method

we propose Large-Discrepancy Multi-task Learning Network(LD-MTL). Our method firstly divides the dataset into multiple tasks

and then splits the whole learning network into multiple clusters to learn relevant knowledge of the predicted aerodynamic performance according to different tasks. Finally

the relevant knowledge learned from individual clusters are then fused into the final prediction result. Through comparative tests

it is shown that our method can better reflect and quantify the influence of data discrepancies and have a higher prediction accuracy when modeling with largely discrepant aerodynamic data.

关键词

Keywords

references

ZHANG Y, SUNG W J, MAVRIS D N. Application of convolutional neural network to predict airfoil lift coefficient[C]//2018 AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, Kissimmee, Florida. Reston, Virginia: AIAA, 2018. doi: 10.2514/6.2018-1903

PATANKAR S V. Numerical heat transfer and fluid flow[M]. CRC Press, 2018. doi: 10.1201/9781482234213

陶文铨. 数值传热学[M]. 西安: 西安交通大学出版社, 2001.

张涵信, 庄逢甘. 与物理分析相结合的计算流体力学[C]//钱学森技术科学思想与力学论文集, 2001: 138-153.

QIAO J L, HAN Z H, DING Y L, et al. Sonic boom prediction method for supersonic transports based on augmented Burgers equation[J]. Acta Aerodynamica Sinica, 2019, 37(4): 663-674 (in Chinese).

XU C Z, QIAO J L, NIE H, et al. Numerical investigation on aerodynamic performance of a standard model CHN-T1 using an unstructured flow solver[J]. Acta Aerodynamica Sinica, 2019, 37(2): 291-300. (in Chinese)

HAN Z H, ZHANG Y, XU C Z, et al. Aerodynamic optimization design of large civil aircraft wings using surrogate-based model[J]. Acta Aeronautica et Astronautica Sinica, 2019, 40(1): 522398. (in Chinese)

HU L W, ZHANG J, XIANG Y, et al. Neural networks-based aerodynamic data modeling: a comprehensive review[J]. IEEE Access, 2020, 8: 90805-90823. DOI:10.1109/ACCESS.2020.2993562

LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521(7553): 436-444. DOI:10.1038/nature14539

ZHANG Z J, DURAISAMY K. Machine learning methods for data-driven turbulence modeling[C]//22nd AIAA Computational Fluid Dynamics Conference, Dallas, TX. Reston, Virginia: AIAA, 2015 doi: 10.2514/6.2015-2460

LI J H, LI B, XU J Z, et al. Fully connected network-based intra prediction for image coding[J]. IEEE Transactions on Image Processing, 2018, 27(7): 3236-3247. DOI:10.1109/TIP.2018.2817044

PARISH E J, DURAISAMY K. A paradigm for data-driven predictive modeling using field inversion and machine learning[J]. Journal of Computational Physics, 2016, 305: 758-774. DOI:10.1016/j.jcp.2015.11.012

MARCATO A, BOCCARDO G, MARCHISIO D L. A computational workflow to study particle transport in porous media: coupling CFD and deep learning[M]. Computer Aided Chemical Engineering. Amsterdam: Elsevier, 2020: 1759-1764. doi: 10.1016/b978-0-12-823377-1.50294-9

HE L, QIAN W Q, WANG Q, et al. Applications of machine learning for aerodynamic characteristics modeling[J]. Acta Aerodynamica Sinica, 2019, 37(3): 470-479. (in Chinese)

HE L, ZHANG X C, QIAN W Q, et al. Unsteady aerodynamics modeling method based on long short-term memory neural network[J]. Flight Dynamics, 2021, 39(5): 8-12. (in Chinese)

CHEN H, QIAN W Q, HE L. Aerodynamic coefficient prediction of airfoils based on deep learning[J]. Acta Aerodynamica Sinica, 2018, 36(2): 294-299. (in Chinese)

YILMAZ E, GERMAN B. A convolutional neural network approach to training predictors for airfoil performance[C]//18th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, Denver, Colorado. Reston, Virginia: AIAA, 2017. AIAA 2017-3660. doi: 10.2514/6.2017-3660

COLLOBERT R, WESTON J. A unified architecture for natural language processing: deep neural networks with multitask learning[C]//25th international conference on Machine learning - ICML '08, Helsinki, Finland. New York: ACM Press, 2008: 160-167. doi: 10.1145/1390156.1390177

DENG L, HINTON G, KINGSBURY B. New types of deep neural network learning for speech recognition and related applications: an overview[C]//2013 IEEE International Conference on Acoustics, Speech and Signal Processing, Vancouver, BC, Canada. IEEE, 2013: 8599-8603. doi: 10.1109/ICASSP.2013.6639344

GIRSHICK R. Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision (ICCV), Santiago, Chile. IEEE, 2015: 1440-1448. doi: 10.1109/ICCV.2015.169

WHITE C, USHIZIMA D, FARHAT C. Fast neural network predictions from constrained aerodynamics datasets[C]//Proc of the AIAA Scitech 2020 Forum, Orlando, FL. Reston, Virginia: AIAA, 2020 doi: 10.2514/6.2020-0364

DAKUA S P. Performance divergence with data discrepancy: a review[J]. Artificial Intelligence Review, 2013, 40(4): 429-455. DOI:10.1007/s10462-011-9289-8

ZHANG L B, HUANG S L, LIU W, et al. Learning a mixture of granularity-specific experts for fine-grained categorization[C ] //2019 IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea (South). IEEE, 2019: 8330-8339. doi: 10.1109/ICCV. 2019.00842 https://openaccess.thecvf.com/content_ICCV_2019/papers/Zhang_Learning_a_Mixture_of_Granularity-Specific_Experts_for_Fine-Grained_Categorization_ICCV_2019_paper.pdf

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Numerical and experimental study on aerodynamic design for the rear wing of an electric sports car

Research on the characteristics of water droplets impacting moving superhydrophobic surfaces and its regulation

Optimization of hot-air anti-icing structures based on neural networks and genetic algorithms

Numerical simulation on gust responses of high aspect ratio aircraft

Analysis of damping and cross dynamic derivative characteristics for a quasi-HyTRV lifting body

Related Author

No data

Related Institution

No data

AI问答

⁰