基于空洞卷积神经网络的随机动载荷识别

中国民航大学学报 ›› 2025, Vol. 43 ›› Issue (5): 58-63.

基于空洞卷积神经网络的随机动载荷识别

中国民航大学安全科学与工程学院，天津 300300

收稿日期:2023-04-09 修回日期:2023-05-20 出版日期:2025-11-17 发布日期:2025-11-17
作者简介:王伟（1977— ），男，河南许昌人，副教授，博士，研究方向为航空发动机控制系统适航审定
基金资助:
中央高校基本科研业务费专项（3122019163）

Random dynamic load identification based on atrous convolutional neural network

College of Safety Science and Engineering, CAUC, Tianjin 300300, China

Received:2023-04-09 Revised:2023-05-20 Online:2025-11-17 Published:2025-11-17

摘要/Abstract

摘要：

为解决随机动载荷识别精度低的问题，本文将空洞卷积神经网络（ACNN，atrous convolutional neural network）

引入随机动载荷识别研究中，提出一种扩张率为 2 的基于一维 ACNN 的随机动载荷识别方法，该方法通过

提高扩张率来增大信号感受野，进而提高载荷识别精度。以 GARTEUR 飞机模型为研究对象进行载荷识别

验证实验，结果表明：无噪声干扰时，用本文方法识别的翼尖激励点上随机动载荷与真实载荷之间的均方

根误差为 0.990 1 N，相关系数为 0.987 4，功率谱密度曲线（PSD，power spectral density）能够较好地吻合；在

不同噪声水平干扰下，本文方法也能有效识别出随机动载荷时间序列。本文方法具有识别精度高、抗干扰能

力强的优点。

关键词:

载荷识别, 空洞卷积神经网络, 扩张率, 随机动载荷

Abstract:

To address the issue of low accuracy in random dynamic load identification, the atrous convolutional neural network (ACNN) is introduced into the research of random dynamic load identification. A random dynamic load

identification method based on one-dimensional ACNN with a dilation rate of 2 is proposed, which increases the

signal receptive field by increasing the dilation rate, thereby improving the accuracy of load identification. The

load identification verification experiment is conducted using the GARTEUR aircraft model as the research object, and the results show that the root mean square error between the random dynamic load on the wingtip excitation point identified by the method in this article and the real load is 0.990 1 N, with a correlation coefficient

of 0.987 4, and the power spectral density (PSD) curve can match well. Under different levels of noise interference, the proposed method can effectively identify the time series of random dynamic load. The proposed

method has the advantages of high recognition accuracy and strong anti-interference ability.

Key words:

load identification, atrous convolutional neural network, dilation rate, random dynamic load

中图分类号:

王伟, 许步锋, 何文博, 邸泽坤, 孙含宇.

基于空洞卷积神经网络的随机动载荷识别

[J]. 中国民航大学学报, 2025, 43(5): 58-63.

WANG Wei, XU Bufeng, HE Wenbo, DI Zekun, SUN Hanyu.

Random dynamic load identification based on atrous convolutional neural network

[J]. Journal of Civil Aviation University of China, 2025, 43(5): 58-63.

参考文献 15

[1]	杨智春, 贾有. 动载荷的识别方法[J]. 力学进展, 2015, 45: 29-54.
[2]	LIU R X, DOBRIBAN E, HOU Z C, et al. Dynamic load identification
	for mechanical systems：a review[J]. Archives of Computational Methods
	in Engineering, 2022, 29(2): 831-863.
[3]	MIAO B R, ZHOU F, JIANG C Y, et al. A comparative study of regularization method in structure load identification[J]. Shock and Vibration, 2018(1): 9204865.
[4]	WANG L J, LIU J W, XIE Y X, et al. A new regularization method for
	the dynamic load identification of stochastic structures[J]. Computers &
	Mathematics with Applications, 2018, 76(4): 741-759.
[5]	贾俊. 基于神经网络的图像轮廓提取算法研究[D]. 重庆: 重庆师范大学, 2009.
[6]	张方, 朱德懋. 基于神经网络模型的动载荷识别[J]. 振动工程学报, 1997, 10(2): 156-162.
[7]	宋雪刚. 基于光纤光栅传感器的动载荷识别算法研究[D]. 南京: 南京航空航天大学, 2018.
[8]	曾俊玮, 季元进, 任利惠, 等. 融合一维卷积神经网络和双向门控循
	环单元的 APM 车辆轮胎径向载荷识别方法[J]. 中国机械工程,2023, 34(3): 359-368.
[9]	YANG H J, JIANG J H, CHEN G P, et al. Dynamic load identification
	based on deep convolution neural network[J]. Mechanical Systems and
	Signal Processing, 2023, 185: 109757.
[10]	夏鹏, 杨特, 徐江, 等. 利用时延神经网络的动载荷倒序识别[J]. 航空学报, 2021, 42(7): 224452.
[11]	杨特, 杨智春, 梁舒雅, 等. 平稳随机动载荷的信号特征提取与深度神经网络识别[J]. 航空学报, 2022, 43(9): 410-420.
[12]	常晓通, 闫云聚, 刘鎏, 等. 基于 Green 函数和正则化的动态载荷识别方法[J]. 振动·测试与诊断, 2014, 34(1): 124-129.
[13]	张燕红. 基于卷积神经网络的人脸识别研究[D]. 天津: 天津大学,2018.
[14]	谭龙, 高昂. 感受野特征增强的 SSD 目标检测算法[J]. 计算机系统应用, 2020, 29(9): 149-155.
[15]	冯振宇, 郝晓帆, 杨青青, 等. GARTEUR 飞机模型的模态实验与模
	型修正技术[J]. 科学技术与工程, 2019, 19(25): 105-111.