Uncertainty quantification of modal properties of Rainbow Bridge from multiple-setup OMA data

Operational modal analysis (OMA) has been increasingly applied to identify the modal parameters of a constructed structure, due to its high economy in implementation. However, due to the absence of loading information, the identified modal parameters are often associated with significantly higher uncertainty compared to their counterparts in free or forced vibration tests. Quantifying the identification uncertainty, and hence precision, is therefore especially relevant in OMA. On the other hand, it is also necessary to manage the uncertainty during the planning stage for an ambient vibration test. For example, to achieve a certain identification precision, how long should the data be collected? Contributing to uncertainty quantification and management in OMA, this paper presents work on full-scale vibration testing on a suspension bridge. Eight triaxial accelerometers were deployed for the test. Each was paired with a synchronised data logger capable of storing data locally in a distributed manner. ‘Uncertainty laws’, which are closed-form asymptotic expressions explaining identification uncertainty in terms of test configurations, were applied for planning the test. Four setups were carried out in the ambient vibration test to cover 26 measurement locations. Modal identification is challenged by the low signal/noise level due to the heavy double-deck girder and the first two vertical modes occurring at very close frequencies. A recently developed Bayesian multiple-setup approach is used to identify the modal properties in terms of their most probable value and identification uncertainty. The results are compared with those obtained by a conventional method. The test configuration is also assessed based on the computed uncertainties and uncertainty laws. Lessons learnt are discussed. • Ambient vibration test of Rainbow Suspension Bridge in Tokyo. • Multiple setups were designed based on ‘Uncertainty laws’. • Operational modal analysis was performed by a recently developed Bayesian multiple-setup approach. • Identification uncertainties of modal parameters were quantified. • Test configuration was assessed based on the computed uncertainties and uncertainty laws.