Bayesian operational modal analysis of a long-span cable-stayed sea-crossing bridge

期望最大化贝叶斯模态识别算法在跨海斜拉桥运营模态分析中的应用

Sea-crossing bridges have attracted considerable attention in recent years as an increasing number of projects have been constructed worldwide.Situated in the coastal area,sea-crossing bridges are subjected to a harsh environment(e.g.strong winds,possible ship collisions,and tidal waves)and their performance can deteriorate quickly and severely.To enhance safety and serviceability,it is a routine process to conduct vibration tests to identify modal properties(e.g.natural frequencies,damping ratios,and mode shapes)and to monitor their long-term variation for the purpose of early-damage alert.Operational modal analysis(OMA)provides a feasible way to investigate the modal properties even when the cross-sea bridges are in their operation condition.In this study,we focus on the OMA of cable-stayed bridges,because they are usually long-span and flexible to have extremely low natural frequencies.It challenges experimental capability(e.g.instrumentation and budgeting)and modal identification techniques(e.g.low frequency and closely spaced modes).This paper presents a modal survey of a cable-stayed sea-crossing bridge spanning 218 m+620 m+218 m.The bridge is located in the typhoon-prone area of the northwestern Pacific Ocean.Ambient vibration data was collected for 24 h.A Bayesian fast Fourier transform modal identification method incorporating an expectation-maximization algorithm is applied for modal analysis,in which the modal parameters and associated identification uncertainties are both addressed.Nineteen modes,including 15 translational modes and four torsional modes,are identified within the frequency range of[0,2.5 Hz].

目的:由于地理位置特殊,跨海大桥周围的环境非常复杂,进而导致跨海桥梁的模态特征复杂多变。本文旨在应用期望最大化贝叶斯快速傅里叶变换(FFT)算法对跨海斜拉桥进行运营模态分析。创新点:1.通过使用期望最大化贝叶斯FFT算法,使得基于贝叶斯的运营模态分析速度更快且收敛性更高;2.成功识别了2.5Hz以内的19阶模态的自然频率、阻尼比以及振型,同时得到了识别参数的不确定性大小。方法:通过应用贝叶斯模态识别算法对某跨海斜拉桥的运营模态数据进行分析,并研究模态参数及其不确定性。结论:应用期望最大化贝叶斯FFT算法能够高效地识别2.5 Hz以内的19阶模态的自然频率、阻尼比和结构振型,并能得出参数识别的不确定性大小。