Study Of Bridge Deflection Separation Based On SVD And Eiegnvalue Analysis

Health monitoring of bridge structure can not only evaluate the safety state of thestructure, but also alarm potential safety hazard of it early，so as to avoid catastrophicaccidents. Deflection is one of the key parameters of bridge structural healthmonitoring,and it is an important index to evaluate the safety and determine thepotential bearing capacity of bridge structure, because it can accurately reflect theworking performance under various factors. Bridge deflection data obtained bybridge health monitoring is the superposition of deflection which in a variety ofenvironmental factors and long-term deformation forces.Environmental factorscontained vehicle load、the crowd、the day temperature, the annual temperature andlong-term deformation caused by prestress loss and gradient shrinkage ofconcrete.Obtaining accurate deflection component data of each factors and judgingwhether it in the allowable range, thus to evaluate the safety of bridge structure anddetermine the potential bearing capacity.In this paper, the approaches using the multi-scale processing method, analysingthe effect of various factors on the time scale, combining with signal processingtechnology and putting forward to establish the system of bridge deflection signalseparation system are applied. Firstly, through the analysis of the high-frequencysignal components in the signal containing deflection (usually the vehicle loaddeflection signal), proposed the separation by using FIR filter; secondly consideringthe bridge frequency deflection signal contains periodic signal components,respectively, using Singular value decomposition algorithms and eigenvalue analysisto separate.Singular value decomposition and eigenvalue analysis is one of the mostimportant theories of matrix theory and linear algebra.Based on the theory of lowfrequency deflection signal contains periodic signal components, two algorithms ofthe deflection separation model were constructed, by solving the model, realize theseparation of low frequency deflection signal. Because, before the separation of lowfrequency deflection signal, should accurately identify the cycle interval, which contains a cycle. Therefore, we proposed using delay singular value ratio spectrummethod for low frequency deflection signal detection.And the detection results show,that the method can greatly recognize its period interval.In this paper, by using Midas software,the simulation calculation of deflectionunder the action of various factors was carried out, and it was also used for separation.Correlation coefficients were all above0.9, showing that the effect is better.Compared the separation results with the results of EMD+ICA method, the algorithmis better than the later. Finally, the algorithm proposed in this paper used to separatethe real signal, correlation coefficient of temperature deflection is above0.9, and thecorrelation coefficient of the long-term deflection is above0.89, which showed thatthe separation effect is good.