基于LSTM和响应分解的冲击载荷识别方法研究  

Impact load identification method based on LSTM and response decomposition

作　　者：黄大伟 陈立昆 高亚东[1] HUANG Dawei;CHEN Likun;GAO Yadong(National Key Laboratory of Rotorcraft Aeromechanics,College of Aerospace Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学航空学院直升机旋翼动力学国防科技重点实验室,南京210016

出　　处：《振动与冲击》2024年第3期69-76,119,共9页Journal of Vibration and Shock

基　　金：直升机旋翼动力学国防科技重点实验室基金(61422202104)。

摘　　要：同一量级的冲击载荷所产生的动响应要远大于静态响应,因此准确识别冲击载荷对于航空器结构件的动强度设计、校核与结构健康监测都具有重要意义。该文章提出的方法主要针对一般线性结构的冲击载荷识别问题,从实测冲击响应应变信号出发,主要解决了冲击载荷与响应信号样本长度不一致这一突出矛盾。首先基于冲击响应信号分解方法来进行振动信号特征提取,然后基于长短期记忆神经网络对载荷和响应信号样本特征进行映射,从而实现冲击载荷识别。通过对挂架模型实测冲击载荷信号进行识别,结果表明4种工况下,该方法识别的冲击载荷的均方根相对误差小于0.6,相关系数大于0.94。结果初步表明,在理想的试验环境中,该方法具备一定的识别精度。Dynamic response generated by the same level of impact load is much larger than static response,so correct identification of impact load is of important significance for dynamic strength design,verification and structural health monitoring of aircraft structural components.Here,the method proposed here mainly focuses on identification of impact loads on general linear structures.Starting from measured impact response strain signals,the prominent contradiction of inconsistent sample lengths between impact load and response signals was solved.Firstly,extracting signal characteristic was done based on impact response signal decomposition method.Then,load and response signal sample features were mapped using long-short term memory neural network to realize impact load recognition.By identifying measured impact load signals of a suspension model,the results showed that under 4 working conditions,RMS relative errors of impact load identified by this method are less than 0.6,correlation coefficients are larger than 0.94;this method has a certain recognition accuracy in ideal test environment.

关 键 词：动力学逆问题 冲击载荷识别 响应分解 振动信号特征提取 长短期记忆神经网络 

分 类 号：TB123[理学—工程力学]