基于声发射的铝蜂窝板超高速撞击损伤模式识别方法

发布时间：2018-04-25 15:30

本文选题：空间碎片 + 超高速撞击　；参考：《航空学报》2017年05期

【摘要】：为通过声发射技术识别铝合金蜂窝板超高速撞击(HVI)的损伤状态,提出一种基于神经网络的损伤模式识别方法。通过超高速撞击实验获取声发射信号,结合精确源定位技术、时频分析技术、小波分析技术及模态声发射技术,提出了10个与损伤相关的特征参数,通过非参数检验分析其与损伤的关系,设计了一种基于贝叶斯正则化BP神经网络的超高速撞击损伤模式识别方法。建立最优网络模型,通过不同参数组合识别能力分析,优选出2种特征参数组合,通过非同源样本对其损伤模式识别能力进行验证。结果表明:传播距离与损伤模式无关,却是识别损伤模式的重要参数;125~250kHz频域的自动加窗小波能量比会降低损伤模式的识别能力;采用贝叶斯正则化的BP神经网络可以较好地识别蜂窝板超高速撞击损伤模式,参数组合为传播距离、上升时间、持续时间、截止频率、4个自动加窗小波能量比及小波能量熵,共9个参数,对任意选取非同源样本识别错分率仅为9.38%。
[Abstract]:In order to identify the damage state of the hypervelocity impact (HVI) of aluminum alloy honeycomb panel by acoustic emission technology, a method of damage pattern recognition based on neural network is proposed. The acoustic emission signals are obtained by ultra high speed impact test, combined with the precise source location technology, time frequency analysis, small wave analysis and modal acoustic emission technology, 10 of which are presented. The relationship between damage and damage is analyzed by nonparametric test. A model identification method for hypervelocity impact damage based on Bayesian regularization BP neural network is designed. The optimal network model is established. Through the analysis of different parameters combination recognition ability, 2 combination of characteristic parameters is optimized and the non homologous sample is used. The results show that the propagation distance is independent of the damage mode, but it is an important parameter to identify the damage mode, and the automatic adding window wavelet energy ratio in the 125~250kHz frequency domain can reduce the recognition ability of the damage mode, and the BP neural network with Bayesian regularization can identify the hypervelocity impact damage of the honeycomb plate. The parameter combination is the propagation distance, the rising time, the duration, the cut-off frequency, the 4 automatic window wavelet energy ratio and the wavelet energy entropy, which are 9 parameters, and the error rate is only 9.38%. for the arbitrary selection of non homologous samples.

【作者单位】：哈尔滨工业大学航天学院;
【基金】：国家“十二五”空间碎片专项(K0203210) 中央高校基本科研业务费专项资金(HIT.NSRIF.2015029)~~
【分类号】：V528

【相似文献】