第一作者胡异丁男，讲师，1974年生

基于延时随机子空间方法的非白噪声环境激励结构模态参数识别

胡异丁1，李丹2，任伟新3，李子兵4

(1.五邑大学信息工程学院，广东江门529020； 2. 新加坡国立大学土木与环境工程学院，新加坡117576；3.合肥工业大学土木与水利工程学院，合肥230009；4.安徽省高速公路试验检测科研中心，合肥230601)

摘要：为了消除非白噪声环境激励在结构模态参数识别结果中产生的虚假模态，引入扩展状态模型，从状态空间方程的角度论证了虚假模态产生的原因；然后，针对一类具有典型性和代表性的(自相关函数在纵坐标轴附近一定范围内有非零值的)非白噪声环境激励信号，在传统随机子空间算法的基础上引入延时指标，建立延时随机子空间方法。数值算例表明延时随机子空间方法能够有效地剔除非白噪声环境激励在模态参数识别结果中产生的虚假模态，放宽了传统模态参数识别方法对环境输入的白噪声假设。

关键词：结构模态参数识别；非白噪声环境激励；随机子空间方法；延时指标

基金项目：国家自然科学基金资助项目(51078357)

收稿日期：2014-01-08修改稿收到日期：2014-04-24

中图分类号:TB123; TN911.6文献标志码：A

Modal parameter identification of structures under non-white noise ambient excitations using delay-index-based stochastic subspace method

HUYi-ding1,LIDan2,RENWei-xin3,LIZi-bing4(1. School of Information Engineering, Wuyi University, Jiangmen 529020, China;2. National University of Singapore, Department of Civil and Environmental Engineering, Singapore 117576;3. School of Civil Engineering, Hefei University of Technology, Hefei 230009, China;4. Anhui Expressway Engineering Test and Research Center, Hefei 230601, China)

Abstract:In order to eliminate spurious modes caused by non-white noise ambient inputs, an augmented state space model was introduced to explain how the spurious modes arise in the modal parameter identification of structures. For a kind of typical non-white ambient excitations, whose autocorrelation function values are nonzeroes near the vertical axis, the delay-index-based stochastic subspace method was proposed by introducing a delay index in the traditional stochastic subspace method. Numerical examples showed that this improved method can eliminate spurious modes due to non-white noise inputs the white noise assumption of ambient excitations in the traditional modal analysis methods.

Key words:structural modal parameter identification; non-white noise ambient excitations; stochastic subspace method; delay index

环境激励结构模态参数识别方法[1-2]利用车辆、行人、风等日常工作荷载作为系统的输入，仅根据结构的响应完成模态参数识别。其中，频域类包括峰值拾取法(Peak Picking Technique, PP)[3]，频域分解法(Frequency Domain Decomposition, FDD)[4]，Polymax法[5]等；时域类包括ARMA自回归滑动平均模型法[6]，NExT(National Excitation Technique)[7]，随机子空间方法(Stochastic Subspace Identification，SSI)[8-12]等。这些方法都假定未知的环境输入为平稳白噪声随机过程，而实际的自然环境激励可能包含一些由交通、阵风甚至地震波等引起的频率成分，表现出明显的非平稳、非白噪声特性[13-16]。……