OPERATIONAL MODAL ANALYSIS USING EM ALGORITHM AND MODAL-FORM STATE-SPACE MODEL WITH AUTO MODEL ORDER REDUCTION

With advantages such as simplicity of testing, low cost and practical feasibility, etc., operational modal analysis has been widely used in engineering practice. In state-space model-based modal analysis, determination of the model order is one of the key problems to obtain accurate and stable modal parameters. A method for auto model order reduction and modal identification is proposed using expectation-maximization (EM) algorithm and modal-form state-space model. By transforming the general state-space model into a special modal-form, not only the parameter space is simplified, but the modal responses can be estimated. Modal contribution ratio which represents the importance of each mode to the total response is used as a criterion for auto order reduction in EM algorithm. Meanwhile, combined with spectrum analysis and damping ratio threshold, spurious modes are eliminated for the selection of practical structural modes. Through validation analysis of synthetic and field data, it is shown that the method is practically feasible and effective.