平稳随机载荷的信号特征提取与深度神经网络识别  

Feature extraction and identification of stationary random dynamic load using deep neural network

作　　者：杨特 杨智春[1] 梁舒雅 康在飞 贾有[1,2] YANG Te;YANG Zhichun;LIANG Shuya;KANG Zaifei;JIA You(School of Aeronautics,Northwestern Polytechnical University,Xi’an 710072,China;School of Applied Science,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]西北工业大学航空学院,西安710072 [2]太原科技大学应用科学学院,太原030024

出　　处：《航空学报》2022年第9期402-412,共11页Acta Aeronautica et Astronautica Sinica

基　　金：国家自然科学基金(12102353)。

摘　　要：针对线性时不变结构的平稳随机载荷识别问题,从结构的动力学响应求解原理出发,利用小波变换对于信号特征的提取能力与长短期记忆神经网络(LSTM)对于序列问题的强大建模与映射能力,提出了一种针对平稳随机载荷的特征信号识别方法,通过对作用于三自由度振动系统数值模型上的平稳随机动载荷识别,证明了方法的可行性。对一个受2点平稳随机载荷作用的加筋壁板结构模型进行动载荷识别实验,结果表明,用提出的方法识别的动载荷均方根相对误差均小于5%,该动载荷识别方法具有良好的识别能力。A feature signal identification method for stationary random dynamic load is proposed based on the dynamic principle of structures.using Wavelet transform is used to extract the time-frequency characteristics of signals,and Long-Short Term Memory(LSTM)is employed to model and map sequence problems.The feasibility of the method is proved byidentification of stationary random dynamic loads acting on a three-degree-of-freedom vibration system.The dynamic load identification experiment is carried out on a stiffened panel structure model under two-point stationary random loads.The results show that the root mean square error of dynamic load identified by the proposed method is less than 5%,and the method has good identification ability.

关 键 词：平稳随机载荷 小波变换 振动信号特征提取 深度神经网络 动载荷识别 

分 类 号：V215.36[航空宇航科学与技术—航空宇航推进理论与工程] TB535[理学—物理]