Theoretical And Experimental Study Of Frequency Domain Algorithms For Operational Modal Analysis

Experimental modal analysis is a process of system identification by extracting modal parameters,which is an important way to evaluate the dynamic characteristics of structures.In this thesis,the traditional experimental modal analysis methods are studied in theory and simulation,and the BSS(Blind Source Separation)technology in the field of blind signal processing is introduced into the identification of experimental modal parameters.The experimental modal analysis method under the framework of blind source separation is introduced.The advantages of non parametric modal analysis method in calculation efficiency,identification accuracy and stability are verified by simulation and experimental analysis.In this thesis,the frequency domain methods and time domain methods of classical modal analysis are studied.The scope of application and influencing factors of mobility circle fitting method,the Levy method,the ITD method and the time series method for ARMA model,are studied by changing the structural parameters and the noise of the external environment.The results show that the time domain method has high accuracy and good stability for modal parameter identification and sensitivity for noise,which is more suitable for the condition of high SNR(Signal to Noise Ratio)or denoising of response signal in advance.Because of using the spectral averaging,the frequency domain method is less affected by noise.However,the accuracy of the frequency domain method is affected by the frequency resolution,and the calculation amount is large when the frequency resolution is high.Classical experimental modal analysis methods need to establish a mathematical model of the parameters.Such algorithms are difficult to determine the modal order,which have sensitivity for noise and large calculation.As a non parametric signal processing method,the BSS is applied in the field of mechanical vibration,which can solve the problems of the above-mentioned parametric modal parameter extraction method and realize the modal parameter extraction without measurable input.Based on the modal superposition method,the relationship between BSS and operational modal analysis is established.And the assumption of uncorrelation between modal coordinates in the case of weak damping is given.The influence factors of the accuracy of modal parameter recognition for the FMSV(Frequency domain Minimum Spectral Variance algorithm)and its extended algorithm are analyzed.Through the simulation examples of multi degree of freedom system and simply supported beam,the influence of different system damping,SNR and weighting matrix power on the accuracy of the algorithm is investigated,and the parameters of the FMSV are optimized to further improve the performance of the algorithm.In order to verify the simulation results,the method of modal parameter extraction based on BSS is applied to calculate the vibration test data of diesel engine raft frame and installation platform of diesel engine active damping device.The natural frequency,damping ratio and mode shape of the measured structure are obtained by the optimized FMSV algorithm and compared with those of the commercial test software's default method,which verifies the validity and reliability of the FMSV algorithm in the experimental modal analysis.According to the simulation and experimental results,FMSV,compared with the traditional experimental modal analysis method does not rely on the mathematical model or iteration,and takes up less storage without model ordering.Not only can it ensure the accuracy,stability and robustness to noise of modal parameter identification under the premise of high calculation efficiency,but also be applied to parameter identification of dense modal system,which makes it a very effective modal analysis algorithm in practical application.