提出一种新的由结构自由振动响应识别结构阻尼比的方法。对于单自由度振动系统，推导出响应历程与时轴所围各面积之间的确定性关系后，利用各面积之间的关系来确定自由衰减振动的阻尼比系数。与传统的对数衰减率法比较，它具有抗噪声干扰能力强、精度高、稳定性好及简便实用等特点。对于多自由度振动系统，先将结构上某测点的自由振动响应表达为一理论解析式x（t），以eat相乘x（t），预给a初值范围，通过牛顿二分法（或黄金分割法）搜索a值，直至eatx（t）做等幅振荡，则该阶模态的各模态参数得以确定。从总响应中扣除该阶模态对总响应的贡献后，重复这一过程，则可识别出响应信号中各阶模态的模态阻尼比。仿真计算与岳阳洞庭湖斜拉桥拉索实测试验结果表明了本文方法的有效性和实用性。

斜拉桥拉索模态参数（固有频率、阻尼比）在索力测试、拉索减振、实时控制等方面起着重要作用。文中利用随机减量技术从岳阳洞庭桥环境激励下拾取的拉索的加速度响应中分离出自由衰减振动加速度响应信号，将获取的加速度响应信号表达为一理论形式，综合运用参数识别、最优估计理论，识别出拉索的模态参数，与理论值吻合良好。本文方法简单，试验容易实施，具有工程实用价值。数字仿真与工程测试结果表明了方法的有效性和可行性。

主要研究基于神经网络技术的南京长江大桥桥域船撞力识别。在建立了南京桥4#桥域的空间有限元分析模型后，对4#桥域进行了动力特性分析，基fMATllAB平自建立了“船撞力响应”神经网镕，自沃辛试验船撞力曲线计算得到了船对桥域的撞击响应并作为神经网络训练的样本。仿真计算结果表明，自桥域域顶的响应可有效识别船舶撞击力的大小、方向和撞击位置，该神经网镕可用于有、无噪声干扰的船撞力识别，并且具有良好的窖错性和鲁棒性。

A now approach to estimating the frequency response function (FRF) is proposed in this paper. Wo will dernonstate this using the feature that the third-order spectrum is insensitive to Gaussian while nosie (GWN). The theoretical analysis and results of digital simulation and specimen tost show tha estimate H is cioser to the true FRF value tha estimates H1 H2 H3 and H3 under the contamination of GWN with a zero-mean value.

This paper presents an application of Hilbert-Huangtransform(HHT) analysis to health moni-toring for the Nanjing Yangtze River Bridge (NYRB), and a new method based on the empirical mode de-composition (EMD) to eliminate local strong disturbance from strain signals of main truss members of the NYRB. Firstly, measured vibration signals are decomposed into several intrinsic mode functions (IMFs) and a residue through EMD. Secondly, the signal amplitude and instantaneous frequency as a function of time, and frequency-time distribution of the amplitude designated as the Hilbert-Huang amplitude spectrum are ob-tained. Measured strain data of the NYRB with the disturbance of interphone are purified. Results of process-ing and analysis of measured signals of NYRB by using HHT as the train passing show the feasibility and va-lidity of the method in this paper. It also has demonstrated that the original data can be reconstructed with only four IMFs.