基于物理信息神经网络的跑道结冰预测方法

taptap点点手机网页 ›› 2025, Vol. 43 ›› Issue (3): 38-44.

基于物理信息神经网络的跑道结冰预测方法

1. 同济大学 a. 道路与交通工程教育部重点实验室；b. 民航飞行区设施耐久与运行安全重点实验室，上海

201804；

2. 上海济熠智能科技有限公司，上海

201805

收稿日期:2025-04-14 修回日期:2025-05-21 出版日期:2025-07-12 发布日期:2025-07-12
通讯作者: 刘诗福（1993—），男，江西吉安人，副教授，博士，研究方向为道路与机场工程
作者简介:邢馨元（2001—），女，辽宁沈阳人，硕士研究生，研究方向为道路与机场工程
基金资助:
上海市青年科技启明星项目（24QB2703000）

A runway icing prediction method based on physics-informed neural network

1a. Key Laboratory of Road and Traffic Engineering of Ministry of Education; 1b. Key Laboratory of Infrastructure Durability

and Operation Safety in Airfield of CAAC, Tongji University, Shanghai 201804, China;

2. Shanghai Jiyi Intelligent Technology Co., Ltd., Shanghai 201805, China

Received:2025-04-14 Revised:2025-05-21 Online:2025-07-12 Published:2025-07-12

摘要/Abstract

摘要：

跑道结冰威胁航空器地面运行安全，准确预测道面温度与结冰状态至关重要。为解决传统跑道结冰预测方

法在处理复杂工况、数据稀缺与物理一致性方面的不足，本文提出基于物理信息神经网络（PINN，physics informed neural network）的跑道结冰预测方法，将多层结构热传导与水冰相变机理嵌入深度神经网络，在稀疏数据下实现温度场与水冰状态的高精度预测，并结合自主设计的结冰模拟实验数据，将有限差分法（FDM，

finite difference method ）与 PINN 求解结果进行对比。结果表明，数据-物理联合驱动的 PINN 温度平均预测

误差较 FDM 降低约 90%，仅为 0.21 ℃，具备从有限数据重建全域温度场的能力。此外，本文还分析了盐分降

低冰点、延缓结冰与抑制冰层生长的作用机制。本研究可为跑道结冰预测提供新的技术路径。

关键词:

物理信息神经网络（PINN）, 跑道结冰, 热传导, 相变, 有限差分法（FDM）

Abstract:

Runway icing poses a threat to the safety of aircraft ground operation, accurate prediction of pavement temperature

and icing state is essential. To overcome the limitations of traditional runway icing predication methods in handling

complex scenarios, limited data, and physical inconsistency, this study proposes a runway icing prediction method

based on physics-informed neural network (PINN). The model embeds multilayer structure heat conduction and

water-ice phase change mechanisms into a deep neural network, enabling precise prediction of temperature fields

and water-ice state under limited data. Experimental data from a self-designed icing simulation are used to compare the solution results of PINN with the finite difference method (FDM). Results show that the jointly driven data physics PINN reduces average prediction error of temperature by about 90% compared to FDM, with just 0.21 ℃,

which is able to reconstruct full-field temperature field from limited data. Furthermore, the study analyzes the

mechanisms of salinity lowers the freezing point, delays icing, and suppresses ice growth. These findings can provide a new technological path for runway icing prediction.

Key words:

physics-informed neural network (PINN), runway icing, heat conduction, phase change, finite difference method

(FDM)

中图分类号:

V351.11

邢馨元 a, b , 刘诗福 a, b , 凌建明 a, b , 陶泽峰 a, b, .

基于物理信息神经网络的跑道结冰预测方法

[J]. taptap下载安装安卓学报, 2025, 43(3): 38-44.

XING Xinyuana, b , LIU Shifua, b , LING Jianminga, b , TAO Zefenga, b, .

A runway icing prediction method based on physics-informed neural network

[J]. Journal of Civil Aviation University of China, 2025, 43(3): 38-44.

参考文献 25

	[1]
	ICAO. Annual report-2018 air transport statistics[R]. Montreal: ICAO,
	2018: 1-12.
[2]	王立文, 郭贤, 陈斌, 等．多传感器跑道积冰预警无线数据采集
	系统设计(英文)[J]．机床与液压, 2019, 47(6): 38-45．
[3]	陈斌, 高大枘, 汤凯峰, 等．基于 LADRC 的跑道积冰主动探测装
	置温度精准控制[J]．控制理论与应用, 2024, 41(9): 1610-1618．
	[4]
	DAN H C, TAN J W, DU Y F, et al． Simulation and optimization of road
	deicing salt usage based on water-ice-salt model[J]． Cold Regions Science and Technology, 2020, 169: 102917．
	[5]
	NUIJTEN A D W． Runway temperature prediction, a case study for Oslo
	Airport, Norway[J]． Cold Regions Science and Technology, 2016, 125:
	72-84．
	[6]
	NUIJTEN A D W, H覫YLAND K V． Modelling the thermal conductivity
	of a melting snow layer on a heated pavement[J]． Cold Regions Science
	and Technology, 2017, 140: 20-29．
	[7]
	FUJIMOTO A, SAIDA A, FUKUHARA T． A new approach to modeling
	vehicle-induced heat and its thermal effects on road surface tempera－
	ture[J]． Journal of Applied Meteorology and Climatology, 2012, 51(11):
	1980-1993．
	[8]
	FUJIMOTO A, TOKUNAGA R A, KIRIISHI M, et al． A road surface
	freezing model using heat, water and salt balance and its validation by
	field experiments[J]． Cold Regions Science and Technology, 2014, 106/
	107: 1-10．
	[9]
	CHEN J Q, SUN C Q, SUN X, et al． Finite difference model for predict－
	ing road surface ice formation based on heat transfer and phase transi－
	tion theory[J]． Cold Regions Science and Technology, 2023, 207: 103772．
[10]	CHEN J Q, WANG H, XIE P Y． Pavement temperature prediction: the－
	oretical models and critical affecting factors[J]． Applied Thermal Engi－
	neering, 2019, 158: 113755．
[11]	LIU W, RAO Z. Pavement icing forecasting based on long short-term
	memory network[C]//2019 5th International Conference on Environmental Science and Material Application, Xi′an, China, 2020: 052070.
[12]	WANG S H, WANG T L, PEI X, et al． Highway icing time prediction
	with deep learning approaches based on data from road sensors[J]． Sci－
	ence China Technological Sciences, 2023, 66(7): 1987-1999．
[13]	DILORENZO T, HABIBZADEH-BIGDARVISH O, LI D F, et al． Rigid
	pavement icing: misting tests on a model pavement column under simu－
	lated cold fronts inside a freezer[J]． International Journal of Pavement
	Engineering, 2023, 24(2): 2044036．
[14]	陈斌, 刘悦, 尹开浪, 等. 冰雪天气下基于 LSTM 的跑道温度数
	据-机理联合预测[J]. 北京航空航天大学学报, 2024, 50(7): 2184-
	2194.
[15]	陈斌, 刘悦, 李庆真, 等．冰雪天气下基于 MFOA-KELM 残差修
	正的跑道温度混合预测[J]．北京航空航天大学学报, 2022, 48(11):
	2153-2164．
[16]	余波, 甘子玉, 张森林, 等. 基于物理信息神经网络预测 2D/3D 非
	稳态温度场及热源[J/OL]. 工程力学, 113, [2025-05-21].http://kns.
	cnki.net/kcms/detail/11.2595.O3.20231013.1641.006.html.
[17]	赵暾, 周宇, 程艳青, 等．基于内嵌物理机理神经网络的热传导方
	程的正问题及逆问题求解[J]．空气动力学学报, 2021, 39(5): 19-26．
[18]	MADIR B E, LUDDENS F, LOTHOD譩 C, et al． Physics informed neural
	Networks for heat conduction with phase change[EB/OL]． (2024-10-18)
	[2025-05-20]. https://arxiv．org/abs/2410.14216v1．
[19]	CAI S Z, WANG Z C, WANG S F, et al． Physics-informed neural net－
	works for heat transfer problems[J]． Journal of Heat Transfer, 2021, 143
(6)	: 060801．
[20]	XING Z B, CHENG H, CHENG J． Deep learning method based on physics-informed neural network for 3D anisotropic steady-state heat con－
	duction problems[J]． Mathematics, 2023, 11(19): 4049．
[21]	CHU F Q, LI S X, ZHAO C J, et al． Interfacial ice sprouting during salty
	water droplet freezing[J]． Nature Communications, 2024, 15(1): 2249．
[22]	王海瑞, 王兴鹏, 李朝阳, 等．南疆咸水结冰与融化过程中盐分及离
	子的变化[J]．环境化学, 2022, 41(4): 1392-1400．
[23]	吴赣昌. 层状路面体系温度场分析[J]. 中国公路学报, 1992(4): 17-25.
[24]	RAISSI M, PERDIKARIS P, KARNIADAKIS G E． Physics-informed
	neural networks: a deep learning framework for solving forward and in－
	verse problems involving nonlinear partial differential equations[J]． Journal of Computational Physics, 2019, 378: 686-707．
[25]	LING J M, XING X Y, ZHANG J, et al． An experimental validation based study of airport pavement icing mechanisms in saline environ－
	ments and the development of a simplified prediction model[J]． Applied
	Sciences, 2024, 14(19): 8867．