LLE算法的模态识别及影响因素研究

Modal identification and influence factors of LLE algorithm

局部线性嵌入算法(LLE)是一种实现对高维数据降维的流形学习算法,可基于结构响应数据进行模态参数识别,算法的噪声敏感度和稳定性对参数识别精度具有重要影响。本文以一块复合材料板为研究对象,利用LLE算法对其振动响应数据进行降维处理从而实现模态识别,重点分析了该算法的噪声敏感度和其在不同采样频率下的流形特征稳定性,同时利用模态置信准则(MAC)衡量LLE算法提取得到的振型与有限元振型的相关性。结果表明,利用LLE算法识别的模态参数具有较高的精度,且LLE算法具有较强的抗噪声干扰能力,采样频率对LLE算法的影响与采样定理相一致。

Locally linear embedding(LLE)is a manifold learning algorithm for dimensionality reduction,which can identify modal parameters based on structural response datas.A composite plate has been taken as an example in this paper to analyze the accuracy of modal parameters identified by LLE with different influence parameters,such as,different sampling frequencies,white noise with different magnitudes.The corresponding modal shapes identified by finite element method(FEM)and LLE has been compared by modal assurance criterion(MAC)in this paper to quantify the accuracy of the modal parameters identified by LLE.The results show that LLE algorithm can identify modal parameters with high accuracy,and the algorithm has strong anti-noise ability.And the influence of sampling frequency is consistent with sampling theory.