Research On Tensor Subspace Identification Of Time-varying System For Bridge Structures

In recent years,with the flourish development of structural health diagnosis technology and its wide application in long-span bridges,the research on bridge system identification based on health monitoring signals is booming.The system identification method based on monitoring signals is easy to obtain high-precision modal information from various excitation environments,and has become the key of bridge condition assessment.Subspace system identification methods have attracted more and more attention in bridge structural system identification research field because of less human intervention and good robustness.However,the current subspace system identification methods have low computational efficiency,and can not meet the needs of long-term identification of bridges.Moreover,the complex environment of the bridge leads to the vibration signal with nonlinear,non-stationary and strong noise characteristics,which directly affect the accuracy of results obtained from the subspace system identification methods.Therefore,in view of the deficiency of the research on the system identification of long-span bridges at home and abroad,the scientific research ideas which contain theoretical research,experimental verification and engineering application were adopted to study three problems in the identification of bridge systems: nonlinear and non-stationary testing of signals,signal adaptive noise reduction and rapid system identification with high-precision.This study mainly completed the work of the following aspects:1.On the basis of extensive reading lots of domestic and foreign literatures,the development history and research status of the signal nonlinear and non-stationary testing,signal adaptive noise reduction and system identification methods in the nonlinear system identification research of bridges are analyzed and summarized.The deficiency of the present study is put forward.Then the main research content of this paper and the test and project which supports the study in this paper are introduced.2.The recurrence plots and recurrence quantification analysis methods are introduced in detail.The actual signal physical meaning contained in the recurrence quantification indicators are summarized and analyzed.By combining with the signal substitution technique and introducing the idea of probability and statistics,the nonlinear and non-stationary indexes of bridge signals are established.On this basis,the improved recurrence plots method is set up,and an adaptive unified nonlinear and non-stationary test theory and method for bridge signals is put forward.The validity of the method is verified by numerical simulation signals and measured signals of bridges.3.The current commonly used signal adaptive decomposition methods are summarized and the advantages and disadvantages of each method are analyzed.On this basis,the problem of adaptively determining the decomposition number of intrinsic modal functions in the variational modal decomposition method is solved by means of EMD and principal component analysis.Based on the uncorrelated characteristics of noise components,multi-scale principal component analysis is introduced to realize the adaptive noise reduction of bridge dynamic test signals.By using root mean square error,mean square absolute error,signal-to-noise ratio and other indicators,the superiority of the proposed method in noise reduction performance of nonlinear and non-stationary signals is compared and analyzed.Furthermore,the stability diagram method is used to verify the practicability of the proposed method in reducing noise and improving the accuracy of system identification results.4.The theory and method of subspace system identification based on two-dimensional matrix operation are studied.On this basis,by introducing the time dimension,the twodimensional matrix is extended to three-dimensional tensor,and the time-varying Hankel tensor is established.By introducing tensor expansion and tensor fast parallel decomposition theory,the fast estimation of system matrix based on tensor operation is realized.Through combining with the stability diagram method,a system identification method named tensor subspace system identification for bridges with high accuracy and high speed in different excitation environments is established.Using the measured signals of the model bridge,the comparative analysis is proformed and the results show that the proposed method is superior to the traditional sliding window subspace system identification method in recognition efficiency and accuracy.5.Firstly,taking the curved cable-stayed model bridge with different damaged components as the research object,on the premise that the input signal is unknown,the tensor subspace system identification method is applied to identify the nonlinear variation of structural modal frequency.The effects of different damaged components on the structural dynamic characteristics are analyzed.Taking the seismic test of cable-stayed model bridge as the research object,the damage degree of concrete tower under earthquake excitations with different PGA level is studied on the premise that the input signal is known.The nonlinear variation of bridge damage degree caused by linear variation of PGA of ground motion is discussed.Finally,the measured signal of a long-span cable-stayed bridge is taken as the research object.By using the tensor subspace system identification method,the variation law of modal frequencies of the main girder of cable-stayed bridges is studied from two time scales which contains every minutes in 24 hours during one day and every day and night in 1 day during one month.The variation characteristics of weak nonlinearity and weak nonstationarity of the fundamental and high-order frequencies of the main girder modes are analyzed.