Research And Application For Freqeuncy-domain Modal Parameter Identification With Consideration Complex Ambient Noise

Frequency-domain modal parameter identification methods and software implementation for modal test with stochastic noise or harmonic components interference are investigated in this dissertation.The main work and conclusions are as follows.The simulation methods for multiple-input multiple-output(MIMO)system response,including numerical integrating method,state-space method,frequency response function method(FRF)and digital filter method,are studied systematically.Improvement of these methods are proposed,meanwhile the efficiency,accuracy and condition of each method are given respectively,which lays a numerical verification foundation for further study of vibration signal processing and modal parameter identification.An improved method of modal parameter identification based on input–output Fourier data under the measurement noise and a fast maximum likelihood estimation method are presented.The accuracy of estimated damping can be improved obviously by using these two methods to optimize the identification results of the least squares complex frequency algorithm.For a structure excited by both artificial force and ambient noise,applying Hilbert transformation to the power spectrum density(PSD)of the noise response can yield a new function,which can be combined with frequency response functions obtained through artificial excitation to make the results of modal parameter identification better.In experimental modal analysis,two methods,i.e.harmonic components reconstruction and advanced complex mode indicator function(CMIF),are proposed to eliminate the harmonic components in FRFs.The former is achieved by the harmonic components reconstruction in impulse response.The latter,as an improvement of CMIF method,separates the harmonic components in modal space by using the pseudo-inverse matrix of mode shape and participation factor to compute the enhanced FRF.The methods for harmonic detection and removal are studied in detail.First,the high-order cumulants of harmonic and white noise is derived,respectively.Then a new function named spectral variance(SV)is proposed to detect the harmonic components,and a phase-synchronous time averaging(PSTA)technique is presented to remove the harmonic components.Finally,a modified frequency-special domain decomposition(FSDD)method is presented for detecting and eliminating harmonic components during the parameter identification process.A modal parameter identification software named as N-SMA is developed by the combined programming technology of C# and Matlab platform.The software is suitable for EMA and OMA with interference of the test surroundings.Three application cases are carried out to assess the identification quality of the software under the stochastic noise or harmonic components interference.