Research On Modal Parameters Identification Based On Multi-scale Chirplet Sparse Signal Decomposition Method

Modal parameters identification of vibrating structure is widely applied innational defense industry, aeronautics and astronautics, mechanical engineering andso on. According to whether the system’s parameters change over time, the vibrationsystem can be divided into time-varying system and the time-invariant system.Although the time-invariant system modal parameter identification method is alreadymature， it is still need further research for the time-invariant vibration modalparameters identification under complex environment. For time-varying vibrationsystem, as its vibration response signal is non-stationary, it is difficult to identify themodal parameters. Hence, the related research is less in domestic and overseas atpresent.As a new kind of time-frequency analysis method, the Multi-scale ChirpletSparse Signal Decomposition (MCSSD) method is proposed in recent years. Thismethod can adaptively extract the signal component with the largest energy in theoriginal vibration signal, and get its instantaneous frequency accurately. It has strongcapability of noise immunity, and separates the seismic signal into component signals.Hence, it can be used to decouple the vibration response of multiple degree offreedom systems and extract the vibration information.Supported by the national natural science foundation (No.50875078), this thesisapplied the MCSSD method in the field of modal parameter identification, and doesin-depth and systematic research work on modal parameter identification oftime-varying system and time-invariant system.The main research works in this thesis are as follows:(1) The Multi-scale Chirplet Sparse Signal Decomposition method is introduced.As MCSSD can not pick up the component amplitude accurately, a method forimproving the precision of amplitude is put forward. This method uses polynomialfitting to fit the amplitude in each short time segment, then uses Fourier series to fitthe amplitude as a whole. Hence the precision of amplitude is improved, and then, theMCSSD method can be used in the modal parameter identification field.(2) Modal parameters identification method of time-invariant system based onMCSSD is proposed. Aiming at solving the problems such as vibration responsesignal of engineering structure is mixed up with noise and modal coupling of multi-degree-of-freedom system, the MCSSD method is used to decompose theresponse signal of multi-degree-of-freedom system into a number of single moderesponse signals adaptively. Hence decoupling the mode coupling, as well aseliminating the effect of noise. Considering the characteristics of time-invariantsystem response, the least-square identification method is applied to identify themodal parameters. Simulation results of the modal parameter identification of asingle-degree-of-freedom system and multi-degree-of-freedom system are presented,which confirm that compared with EMD method, the proposed method can identifythe modal parameters of time-invariant system effectively.(3) A method for modal parameters identification of time-varying system basedon MCSSD is proposed. The response signal of time-varying structure isnon-stationary and is sensitive to noise. Aiming at those problems, the improvedMCSSD which has strong noise resistance and can decompose non-stationary signaleffectively, is applied to deal with the time-varying structure’s response.Theimproved MCSSD method can adaptively decompose the response signal oftime-varying structure into several single-mode signals， extract the amplitude andfrequency of each single-mode signals，thereby getting the system modal parameters.The simulation results confirmed the effectiveness of the method.(4) A method for structure parameters identification of time-varying singledegree of freedom(SDOF) system based on MCSSD is put forward. On the basis ofmodal parameter identification, the structure parameters can be obtained. Linearchanging, sudden changing and cycle of slow changing modes of stiffness are used tosimulate the slow-varying damage, fracture damage and complex damage of actualstructure under the external loads. The numerical analysis indicated that the proposedmethod can identify the structure parameters effectively, thus providing an importantbasis for structure damage identification.The improved MCSSD method can use the chirplet atoms whose instantaneousfrequency curves are linear lines to adaptively match the signal component with thelargest energy in a multi-component vibration signals. Meanwhile, the instantaneousfrequency of this signal component can be estimated by jointing the piecewise linearfrequencies of the chirplets. In this thesis, the MCSSD method is applied to the modalparameters identification of time-invariant system and time-varying system, thestructure parameters identification of time-varying SDOF system. The simulationresults show that, according to modal frequency, this method can adaptivelydecompose the vibration response signal of multiple degrees of freedom into several single-mode signals, and then the modal parameters of vibration system can be furtheridentified effectively.