Frequency Domain Modal Parameter Identification And Software Implementation

The research in this dissertation is focused on frequency-domain modal parameters identification method for MIMO in deterministic and stochastic framework, operational modal analysis, uncertainty calculation of model parameters and software implementation of algorithm. The main work and conclusions include as follows.Modal parameter identification method based on right matrix fraction description model is studied deeply in deterministic framework. The performance of orthogonal functions is compared in s and Z domain. Fast implementation methods are investigated in Z domain. A modal parameter identification method based on left matrix fraction description of FRF is proposed.Owing to the fact that FRF cannot be estimated exactly when only a limited test data is available, a modal parameters identification method starting directly from IO spectra is studied. Two algorithms based on common denominator and left matrix fraction description are presented according to parametric model of FRF. A simulation case of GARTEUR model is employed to validate the algorithm. The results show that the algorithm based on left matrix fraction description can get better identification, especially for closely spaced mode.A frequency-domain modal parameters identification method based on maximum likelihood estimation is investigated in stochastic framework. This method uses right matrix fraction description model of FRF in discrete time domain. The noise covariance matrix is adopted as weighting function. First, the least square estimation is implemented to get the initial value of modal parameters. Then, the iterative optimization of Newton-Gauss algorithm is carried out to get more precise identification result. According to the Cramer-Rao lower bounder inequality, statistical information of the modal parameters is obtained with increasing a little calculation. A simulation case of a GARTEUR model and an application case of an automobile chassis are employed to validate the method.Due to the measurement noise, the uncertainty of model parameters is inevitable. The uncertainty can be described by using the variance of model parameters. For CMIF, FDPR and polyLSCF, the variance estimation procedure is proposed by first-order sensitivity analysis of the modal parameters to the perturbations of measured FRF. The Monte Carlo results show that first-order sensitivity analysis can reach to high accuracy under small noise.Modal parameters identification approach based on positive power spectral density is studied for ambient excitation. According to the same expression between positive power spectral density and FRF, EMA algorithms can be applicable for ambient excitation. The positive time lag points, exponential window and response date length which have influence on the identification results are discussed. Processing methods for multi-setup measurement data are presents for engineering application. Focusing on the requirement of automatic modal analysis in condition monitoring, a method based on fuzzy clustering is proposed for the analysis of stabilization diagram. Automatic modal analysis is achieved by selecting the clusters grouped by physical poles.A modal parameter identification software named as N-Broband is developed in VC++ platform. The software is suitable for EMA and OMA with broband identification feature. Correlation analysis between experiment and FEA can be performed in N-Broband. An application case of vibration table is carried by N-Broband.