基于迁移学习的结冰风洞试验冰形预测方法

任宇鹏; 王强; 屈经国; 彭博; 岳静; 易贤

doi:10.13224/j.cnki.jasp.20230169

基于迁移学习的结冰风洞试验冰形预测方法

doi: 10.13224/j.cnki.jasp.20230169

任宇鹏^{1, 3,},
王强^{2, 3},
屈经国^{1, 3},
彭博¹,
岳静¹,
易贤^{2, 3,*, ,}

1.
西南石油大学计算机与软件学院，成都 610500
2.
空天飞行空气动力科学与技术全国重点实验室，四川绵阳 621000
3.
中国空气动力研究与发展中心结冰与防除冰重点实验室，四川绵阳 621000

基金项目: 国家自然科学基金重点基金（12132019）；国家重大科技专项（J2019-Ⅲ-0010-0054）；国家自然科学面上基金（12172372）

详细信息

作者简介:
任宇鹏（1999-），男，硕士生，研究方向为深度学习与结冰预测。 E-mail：3483621160@qq.com

通讯作者:
易贤（1977-），男，研究员，博士，研究方向为飞机结冰与防/除冰。E-mail：yixian_2000@163.com

中图分类号: V211.41
计量
- 文章访问数: 361
- HTML浏览量: 195
- PDF量: 38
- 被引次数: 0
出版历程
- 收稿日期: 2023-03-20
- 网络出版日期: 2025-05-22

Ice shape prediction method for icing wind-tunnel experiment based on transfer learning

REN Yupeng^{1, 3
,},
WANG Qiang^{2, 3},
QU Jingguo^{1, 3},
PENG Bo¹,
YUE Jing¹,
YI Xian^{2, 3,*
, ,}

1.
School of Computer Science and Software Engineering，Southwest Petroleum University，Chengdu 610500，China
2.
State Key Laboratory of Aerodyniamics，Mianyang Sichuan 621000，China
3.
Key Laboratory of Icing and Anti/De-icing，China Aerodynamics Research and Development Center，Mianyang Sichuan 621000，China

摘要

摘要:
针对目前缺乏有效预测高精度风洞试验冰形手段的问题，提出了一种结合迁移学习与神经网络的方法来实现对风洞试验冰形的预测。该方法基于数值模拟冰形数据样本进行训练，获得预训练模型；引入结冰风洞试验冰形数据样本对预训练模型进行微调，获得最终的预测模型。模型以U-Net和多层感知机为主要架构，以翼型数据和结冰气象参数作为输入，以二维结冰冰形作为输出。结果表明：提出的方法能够实现结冰风洞试验冰形的准确预测，在主要几何特征上与风洞试验冰形吻合度较高，大部分结果的相对误差不超过15%。该方法为在地面条件下研究航空飞行器的冰形特征规律提供了新的手段。
- 风洞试验 /
- 冰形预测 /
- 迁移学习 /
- U-Net模型 /
- 深度学习
Abstract:
To address the lack of effective means to predict the ice shape of high-precision wind tunnel experiment, a method combining transfer learning and neural networks was proposed to predict ice shape of wind-tunnel experiment. According to this method, a pre-trained model was obtained by training based on numerical simulation ice shape data samples at first. Secondly, ice shape data samples from icing wind-tunnel experiment were introduced to fine-tune the pre-trained model, ultimately obtaining the final prediction model. The model adopted the U-Net and multilayer perceptron as the main architecture, with airfoil data and icing meteorological parameters as the input, and 2-dimensional ice shape as the output. The results showed that the proposed method can achieve accurate prediction of ice shape in icing wind-tunnel experiment, which was very close to the ice shape in wind-tunnel experiment in terms of main geometric features. The relative error of most results was not more than 15%. This method could provide a new means for studying the characteristics of aircraft icing under ground conditions.
- wind-tunnel experiment /
- ice shape prediction /
- transfer learning /
- U-Net model /
- deep learning

HTML

图 1 影响结冰的主要因素

Figure 1. Main factors affecting icing

下载: 全尺寸图片幻灯片

图 2 冰形预测模型网络架构

Figure 2. Network structure of ice shape prediction model

下载: 全尺寸图片幻灯片

图 3 冰形对比示例图

Figure 3. Example diagram of icing shape contrast

下载: 全尺寸图片幻灯片

图 4 二维冰形几何特征

Figure 4. Geometric features of two-dimensional ice shape

下载: 全尺寸图片幻灯片

图 5 模型的预测结果对比

Figure 5. Comparison of prediction results of model

下载: 全尺寸图片幻灯片

表 1 结冰参数取值

Table 1. Values for icing parameters

结冰参数	取值
α/（°）	3.5
v/（m/s）	58.1, 67.1, 102.8
T/℃	−27.96, −19.96, −16.66, −13.64, −11.11, −10.3, −8.08, −7.78, −6.41, −4.75
LWC/（g/m³）	0.4, 0.5, 0.55, 1, 1.3, 1.6
MVD/μm	15, 20, 25, 30, 40
t/min	3, 6, 7, 8, 12

下载: 导出CSV

表 2 测试集工况参数取值

Table 2. Parameters values of test set

算例	α/ （°）	v/ （m/s）	T/℃	LWC/ （g/m³）	MVD/ μm	t/ min
1	3.5	102.8	−11.11	0.55	20	7
2	3.5	67.1	−8.08	1.06	30	6
3	3.5	67.1	−4.75	1	20	12
4	3.5	102.8	−11.11	1.6	30	6

下载: 导出CSV

表 3 微调模型误差

Table 3. Error of fine-tune model %

算例	${S_{\text{u}}}$	${S_{\text{l}}}$	${h_{{\text{sp}}}}$	${S_{{\text{ice}}}}$	${h_{\text{u}}}$	${h_{\text{l}}}$	${\theta _{\text{u}}}$	${\theta _{\text{l}}}$
1	9.82	2.69	14.22	19.52	6.42	7.41	3.69	10.66
2	13.81	20.22	12.11	10.55	5.82	4.14	2.59	2.98
3	8.32	37.65	6.57	13.53	19.63	6.33	1.16	5.29
4	23.61	12.39	5.62	3.88	10.93	7.99	6.48	2.14

下载: 导出CSV

参考文献(28)

[1]	LYNCH F T, KHODADOUST A. Effects of ice accretions on aircraft aerodynamics[J]. Progress in Aerospace Sciences, 2001, 37(8): 669-767. doi: 10.1016/S0376-0421(01)00018-5
[2]	BRAGG M. Aircraft aerodynamic effects due to large droplet ice accretions[C]//34th Aerospace Sciences Meeting and Exhibit. Reno, US: AIAA, 1996: 932.
[3]	易贤, 朱国林, 王开春, 等. 翼型积冰的数值模拟[J]. 空气动力学学报, 2002, 20(4): 428-433. YI Xian, ZHU Guolin, WANG Kaichun, et al. Numerically simulating of ice accretion on airfoil[J]. Acta Aerodynamica Sinica, 2002, 20(4): 428-433. (in Chinese YI Xian, ZHU Guolin, WANG Kaichun, et al. Numerically simulating of ice accretion on airfoil[J]. Acta Aerodynamica Sinica, 2002, 20(4): 428-433. (in Chinese)
[4]	SCHUCHARD E, MELODY J, BASAR T, et al. Detection and classification of aircraft icing using neural networks[C]//38th Aerospace Sciences Meeting and Exhibit. Reno, US: AIAA, 2000: 361.
[5]	李小龙, 洪冠新. 一种基于神经网络的机翼结冰冰型预测方法[C]//中国航空学会第22届飞行力学与飞行试验学术交流会. 成都: 中国航空学会, 2006: 98-103. LI Xiaolong, HONG Guanxin. A neural network-based prediction method for wing icing shape[C]//22nd Academic Exchange Conference on Flight Mechanics and Flight Test of the Chinese Society of Aeronautics and Astronautics. Chengdu: CSAA, 2006: 98-103. (in Chinese LI Xiaolong, HONG Guanxin. A neural network-based prediction method for wing icing shape[C]//22nd Academic Exchange Conference on Flight Mechanics and Flight Test of the Chinese Society of Aeronautics and Astronautics. Chengdu: CSAA, 2006: 98-103. (in Chinese)
[6]	柴聪聪, 易贤, 郭磊, 等. 基于BP神经网络的冰形特征参数预测[J]. 实验流体力学, 2021, 35(3): 16-21. CHAI Congcong, YI Xian, GUO Lei, et al. Prediction of ice shape characteristic parameters based on BP nerual network[J]. Journal of Experiments in Fluid Mechanics, 2021, 35(3): 16-21. (in Chinese CHAI Congcong, YI Xian, GUO Lei, et al. Prediction of ice shape characteristic parameters based on BP nerual network[J]. Journal of Experiments in Fluid Mechanics, 2021, 35(3): 16-21. (in Chinese)
[7]	OGRETIM E, HUEBSCH W, SHINN A. Aircraft ice accretion prediction based on neural networks[J]. Journal of Aircraft, 2006, 43(1): 233-240. doi: 10.2514/1.16241
[8]	张强, 高正红. 基于神经网络的翼型积冰预测[J]. 飞行力学, 2011, 29(2): 6-9. ZHANG Qiang, GAO Zhenghong. Prediction of ice accretions based on the neural net[J]. Flight Dynamics, 2011, 29(2): 6-9. (in Chinese ZHANG Qiang, GAO Zhenghong. Prediction of ice accretions based on the neural net[J]. Flight Dynamics, 2011, 29(2): 6-9. (in Chinese)
[9]	CHANG Shinan, LENG Mengyao, WU Hongwei, et al. Aircraft ice accretion prediction using neural network and wavelet packet transform[J]. Aircraft Engineering and Aerospace Technology, 2016, 88(1): 128-136. doi: 10.1108/AEAT-05-2014-0057
[10]	屈经国, 王强, 彭博, 等. 基于多模态融合的任意对称翼型结冰预测方法[J]. 航空动力学报, 2024, 39(1): 20220143. QU Jingguo, WANG Qiang, PENG Bo, et al. Icing prediction method for arbitrary symmetric airfoil using multimodal fusion[J]. Journal of Aerospace Power, 2024, 39(1): 20220143. (in Chinese QU Jingguo, WANG Qiang, PENG Bo, et al. Icing prediction method for arbitrary symmetric airfoil using multimodal fusion[J]. Journal of Aerospace Power, 2024, 39(1): 20220143. (in Chinese)
[11]	ADDY H E. Ice accretions and icing effects for modern airfoils[M]. Cleveland, US: National Aeronautics and Space Administration, Glenn Research Center, 2000.
[12]	陈晓圆. 飞机结冰多参数影响研究[D]. 南京: 南京航空航天大学, 2015. CHEN Xiaoyuan. Study on multi-parameter influence of aircraft icing[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2015. (in Chinese CHEN Xiaoyuan. Study on multi-parameter influence of aircraft icing[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2015. (in Chinese)
[13]	DAI Wenyuan, YANG Qiang, XUE Guirong, et al. Boosting for transfer learning[C]//Proceedings of the 24th International Conference on Machine Learning. Corvalis, US: Association for Computing Machinery, 2007: 193-200.
[14]	GARCKE J, VANCK T. Importance weighted inductive transfer learning for regression[C]// Machine Learning and Knowledge Discovery in Databases: European Conference. Nancy, France: Springer Berlin Heidelberg, 2014: 466-481.
[15]	SALAKEN S M, KHOSRAVI A, NGUYEN T, et al. Seeded transfer learning for regression problems with deep learning[J]. Expert Systems with Applications, 2019, 115: 565-577. doi: 10.1016/j.eswa.2018.08.041
[16]	ZHU Yin, CHEN Yuqiang, LU Zhongqi, et al. Heterogeneous transfer learning for image classification[C]//Proceedings of the 25th AAAI Conference on Artificial Intelligence, San Francisco, US: AAAI Press, 2011: 1304-1309.
[17]	QUATTONI A, COLLINS M, DARRELL T. Transfer learning for image classification with sparse prototype representations[C]//2008 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway, US: IEEE, 2008: 1-8.
[18]	TIRINZONI A, SESSA A, PIROTTA M, et al. Importance weighted transfer of samples in reinforcement learning[C]//International Conference on Machine Learning. Stockholm, Sweden: PMLR, 2018: 4936-4945.
[19]	GAMRIAN S, GOLDBERG Y. Transfer learning for related reinforcement learning tasks via image-to-image translation[C]//International conference on machine learning. Long Beach, US: PMLR, 2019: 2063-2072.
[20]	何磊, 钱炜祺, 易贤, 等. 基于转置卷积神经网络的翼型结冰冰形图像化预测方法[J]. 国防科技大学学报, 2021, 43(3): 98-106. HE Lei, QIAN Weiqi, YI Xian, et al. Graphical prediction method of airfoil ice shape based on transposed convolution neural networks[J]. Journal of National University of Defense Technology, 2021, 43(3): 98-106. (in Chinese HE Lei, QIAN Weiqi, YI Xian, et al. Graphical prediction method of airfoil ice shape based on transposed convolution neural networks[J]. Journal of National University of Defense Technology, 2021, 43(3): 98-106. (in Chinese)
[21]	RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]//18th International Conference on Medical Image Computing and Computer-Assisted Intervention. Munich, Germany: Springer International Publishing, 2015: 234-241.
[22]	AMIRI M, BROOKS R, BEHBOODI B, et al. Two-stage ultrasound image segmentation using U-Net and test time augmentation[J]. International Journal of Computer Assisted Radiology and Surgery, 2020, 15(6): 981-988. doi: 10.1007/s11548-020-02158-3
[23]	PUNN N S, AGARWAL S. Inception U-Net architecture for semantic segmentation to identify nuclei in microscopy cell images[J]. ACM Transactions on Multimedia Computing, Communications, and Applications, 2020, 16(1): 1-15.
[24]	IBTEHAZ N, RAHMAN M S. MultiResUNet: Rethinking the U-Net architecture for multimodal biomedical image segmentation[J]. Neural Networks, 2020, 121: 74-87. doi: 10.1016/j.neunet.2019.08.025
[25]	WRIGHT W B, RUTKOWSKI A. Validation results for LEWICE 2.0: NASA/CR-1999-208690[R]. Cleveland, US: NASA Glenn Research Center, 1999.
[26]	KINGMA D P, BA J. Adam: a method for stochastic optimization[R]. San Diego, US: International Conference on Learning Representations, 2015.
[27]	YOSINSKI J, CLUNE J, BENGIO Y, et al. How transferable are features in deep neural networks?[C]//Proceedings of the 27th International Conference on Neural Information Processing Systems. Montreal, Canada: Curran Associates, Inc., 2014: 3320-3328.
[28]	WRIGHT W. A summary of validation results for LEWICE 2.0[C]//37th Aerospace Sciences Meeting and Exhibit. Reno, US: AIAA, 1999-249.