Finte Elemet Model Updating Using Multiobjective Optimization Technique

1 INTRODUCTION Dozens of finite element model updating methods have been studied and suggested but still it remains as a difficult problem. That is because model updating is an inverse process and often contains highly non-linear characteristics. Also the updated result is not unique and depends heavily on the objective function and parameters used in the updating process. Usually, objective function is set as the weighted sum of output differences. But the selection of weighting factors is very difficult since the relative importance among the measured data is not obvious but specific for each problem. For example, the fundamental natural frequency difference may be over-weighted in the objective function such that the other updated differences such as 2nd natural frequency, mode shapes are not fit satisfactorily or vice versa. Then, the time-consuming optimization process should be solved again until an appropriate result is derived. In this work, multiobjective optimization technique is introduced in model updating to extremizeseveral objective terms simultaneously. Together with error localiza-tion, an appropriate choice of updating parameters gives an improved result in model updating with physical significance. When the selected parameters are inadequate, then it was found that the updated model is unrealistic. A method to guide the parameter selection is also suggested.