基于独立分量分析的桥梁状态特征提取与识别研究

项目名称： 基于独立分量分析的桥梁状态特征提取与识别研究

项目编号： No.50878081

项目类型： 面上项目

立项/批准年度： 2009

项目学科： 金属学与金属工艺

项目作者： 晏班夫

作者单位： 湖南大学

项目金额： 35万元

中文摘要： 独立分量分析（ICA）以非高斯源信号为研究对象，在统计独立的假定前提下，对观测到的混合信号进行盲分离（BSS），从而把隐含在混合信号中的各独立信源信号较好地分离出来。该方法不需要对系统输入具有先验知识，仅由系统的输出响应便可以完成识别。对独立分量分析的基本理论进行了分析研究，包括各种独立性判据和一些流行的算法。援引一个经典动力模型为研究对象，采用四种不同ICA算法，对模型在不同阻尼、不同噪声干扰情况下的数值模拟位移响应进行识别，通过频率、振型和阻尼比等模态动力参数的对比分析，选出表现最稳定的SOBI算法来进行后续研究；分别运用有限元分析软件ANSYS和场内试验建立同样的钢简支梁模型，利用SOBI算法对模型的加速度响应数据进行识别；结果表明，无论是仿真分析还是场内试验，SOBI算法识别出来的模态动力参数与理论值都比较吻合；开展场外试验，利用上面表现稳定的SOBI算法对衡湘大桥在环境激励下的动力响应进行识别，并与频率分解方法（FDD）和随机子空间法（SSI）进行对比试验。研究表明，应用ICA对环境激励下的实桥模态识别精度还有赖于随机减量信号及互相关信号的质量。

中文关键词： 桥梁；独立分量分析；动力参数；模态识别；SOBI

英文摘要： Independent component analysis (ICA) is an innovative method developed in the field of signal processing recently. This method takes the non-gaussian source signals as the research object and assumes that the signals are statistical independent. Through taking the observed mixed signals to blind source separation (BSS), ICA separate the implicit independent source signals from the mixed signals. This method does not require the prior knowledge of the input and it can conduct the output-only based identification.The concept of virtual source is exploited and the vibration system modals are regarded as the source.The basic theory of ICA is introduced, including various independence criterions and some popular algorithms. Four different ICA algorithms are employed for modal identification based on numerical study on the dynamic responses of a structure with different damping and noise level. Through the comparison of identification results of the previous four ICA algorithms, the SOBI is then employed to extract the frequency, damping and mode of the system under ambient vibration. The capabilities of SOBI and the other identification methods such as stochastic subspace identification (SSI) and frequency domain decomposition (FDD) are compared and assessed by an ambient vibration test on a continuous box girder bridge. The results demonstrate that the performance of SOBI in modal parameter identification is determined by the qualitie of the RD data and the cros-corlation function.

英文关键词： Bridge;Independent Component Analysis(ICA); Dynamic Parameters; Modal Identification; SOBI