Research On Applications Of Stochastic Subspace Identification And Bayesian Modal Analysis Method

Stochastic subspace identification and Bayesian method can both identify the structural modal parameters and quantify their uncertainties,which has become a research hotspot in the field of modal parameter identification.However,in the engineering applications of the two methods,the impact of the sampling during time for dynamic acceleration on the identification results is still unclear;the modal identification accuracy,uncertainty quantification,and identification efficiency of the two methods lack comparative research;the two method fail to be full automated.Based on the above reasons,this paper takes a numerical model and several actual engineering structures as the research objects,and conducts in-depth research on the engineering application of the two methodsFirst of all,the theoretical formulas of the covariance-driven stochastic subspace identification(COV-SSI)and the fast Bayesian FFT modal parameter identification method are derived.Based on a six-degree-of-freedom(6-DOF)spring-mass numerical model and Guangzhou New TV Tower(GNTVT)as research objects,the influence of different acceleration sampling during time on the identification results of the two methods was explored.The results show that:as the increase of the dynamic acceleration sampling during time,the modal parameter identification results of both methods are more reliable,and the uncertainties are gradually decreasing;considering the time cost and the reliability of the identification results comprehensively,it is recommended to set the dynamic acceleration sampling during time as 10-30 minutes in engineering applicationsFurthermore,taking the 6-DOF spring-mass numerical model,HCT Building in Canada,GNTVT in China,and TKB Bridge in Hong Kong as research objects,the comparison research of the two methods has been carried out in terms of modal identification accuracy,uncertainty quantification,and identification efficiency,etc The results show that:both methods can identify the structural modal parameters accurately,and quantify their uncertainties effectively,and the identification results meet the engineering requirements;when the excitation is not white noise,there may be some spurious modes identified via COV-SSI,and the uncertainty quantified under colored-noise excitation are almost larger than those under white-noise,while the Bayesian method is not disturbed;Bayesian methods are more sensitive to changes in modal order,and COV-SSI is more conservative in quantification of uncertainties;in actual structures with a large number of measurement degrees of freedom and a long sampling during time,Bayesian method has higher identification efficiency.Finally,aiming at the problem that COV-SSI requires human interaction and the calculation is time-consuming and labor-intensive,a stabilization diagram containing uncertainty features of the modal parameters is reconstructed.A convolutional neural network is used to automatically identify the constructed stabilization diagram to achieve the COV-SSI fully automatically for modal parameters identification and uncertainty quantification.The proposed method is validated with a 6-DOF spring-mass numerical model,HCT Building in Canada and TKB Bridge in Hong Kong.The results show that the proposed method is very robust and accurate.