Identification Of Structural Parameters Using The Continuous Wavelet Transform Of Acceleration Measurements

Parameters identification is great of importance in dynamic research area speciously for structurehealth monitoring and damage diagnosis. However， previous research work on parametersidentification mainly focuses on linear time-invariant (LTI) structure and its dynamic behavior.Recently the research on linear time-varying (LTV) structure identification has drawn more and moreattention and a number of problems about theory and application for LTV systems need to be solved.The wavelet transform, namely “Mathematical microscope”, has been widely used in signalprocessing for its special multi-scale analysis. It has been proved that CWT theory is more accurateand easier than other methods on the identification of dynamic parameters of LTV structures.In this project, the structure dynamic parameters are identified based on two wavelet methods byusing the measurement acceleration data. The first identification algorithm is based on wavelet ridgesextraction, which allows the estimation of the natural frequencies, viscous damping ratios and modeshapes, by using either the modulus or the phase of the wavelet transform coefficients. Regarding asthe second one, by using the analyzing wavelet’s localized nature and the moment conditions, theequations of motion can be translated to algebraic equations in terms of the acceleration responses.Hence one can identify the structural dynamic parameters only by solving these linear algebraicequations. The first identification approach can only be drove by the free decay responses of a lineardynamical system and the second one can be drove by both the free decay responses and the forcedmeasurement data. Therefore the priority was given to the latter one in this dissertation.A lumped mass-damper-string model of two degrees of freedom is simulated by using theidentification methods under different kinds of system input. The performance of the proposedmethods and the influence of the observation noise as well as the sampling density of the scaleparameters are discussed in numerical examples. The identification formulations of single degree offreedom (SDOF) system and multi-degree of freedom (MDOF) system are deduced. Simulationresults demonstrate that the identification methods are of effectiveness and accuracy for identifyingthe smoothly, periodically and abruptly time-varying parameters of LTV system.