基于短时傅里叶变换的快速贝叶斯模态参数识别方法  

Fast Bayesian modal parameter identification method based on short-time Fourier transform

作　　者：杜航 徐海巍[1] 楼文娟[1] DU Hang;XU Haiwei;LOU Wenjuan(College of Civil Engineering and Architecture,Zhejiang University,Hangzhou 310058,China)

机构地区：[1]浙江大学建筑工程学院,浙江杭州310058

出　　处：《建筑结构学报》2023年第5期305-314,334,共11页Journal of Building Structures

基　　金：国家自然科学基金项目(51978614);浙江省自然科学基金项目(LY19E080026)。

摘　　要：为反映结构非平稳响应信号的时频特性,提出基于短时傅里叶变换的快速贝叶斯模态参数识别(FBST)方法。该方法采用短时傅里叶变换代替傅里叶变换进行模态参数识别,使模态参数的识别同时具有时频特性,同时能够给出识别结果的不确定性。利用时域分解解耦技术,将多自由度多模态响应信号转变为单自由度单模态响应信号以提升计算效率,推导得到高信噪比下负对数似然函数的表达式。采用数值算例验证了FBST方法在时变频率和阻尼比识别上的有效性。在此基础上,针对某大跨柔性光伏支架结构气弹模型的风洞试验数据和某高层建筑风振实测加速度响应数据,利用FBST方法识别了对应结构的阻尼比、频率,并与连续小波变换和Hilbert-Huang变换等经典方法的识别结果进行对比。数值算例分析结果显示,对于时变、非时变信号,FBST方法均能识别与理论值较为一致的阻尼比和频率结果。对于大跨柔性光伏支架结构的气弹试验数据和高层建筑实测加速度响应,FBST方法识别得到的频率结果与连续小波变换以及Hilbert-Huang的结果较为一致,而识别出的阻尼比存在较大变异系数。In order to describe the time-frequency characteristics of non-stationary response signals,a modal identification method based on fast Bayesian and short-time Fourier transform(STFT)was proposed.This method uses short-time Fourier transform to replace classical Fourier transform to identify modal parameters,so that the characteristics of modal parameters in both time and frequency domain can be obtained.Using the time-domain decomposition decoupling(TDD)technology,the multi-DOF multi-modal response signal was converted into a single-DOF single-modal response signal,thus improving the computational efficiency.The expression of NLLF(negative log-likelihood function)was presented under high signal-to-noise ratios.Numerical examples were used to verify the validity of the proposed fast Bayesian STFT method in identification of time-varying frequencies and damping ratios.In addition,the aeroelastic wind tunnel test data of a large-span flexible photovoltaic support structure and field measured wind-induced acceleration responses of a high-rise building were identified with the proposed method to derive the damping ratio and frequency of the corresponding structures.The results were compared with the ones obtained from the classical continuous wavelet transform and Hilbert-Huang transform,which shows that the identified damping ratio and frequency based on FBST are consistent with theoretical ones for both time-dependent and time-independent simulation signals.For aeroelastic test data of the large-span flexible photovoltaic support structure and field measured acceleration data of the high-rise building,the FBST provides frequencies consistent with ones from the continuous wavelet transform and Hilbert-Huang transform,while the obtained damping ratio shows a large coefficient of varation.

关 键 词：模态参数识别 时域分解 短时傅里叶变换 快速贝叶斯 时频特性 非平稳过程 

分 类 号：TU311.3[建筑科学—结构工程]