The component-mode synthesis method is effective for reducing the degrees of freedom of a finite element model for dynamic analysis of a complex structure. However, in the resulting computational mode1 unacceptable errors can exist due to poor modeling. The mode1 can be corrected by applying system identification methods. By localising the parameter errors it is possible to use the updating procedure of the second author. The modal transformation performed within this updating leaves the parameters unchanged.
As a result, the corrected component equations and synthesised component-mode model are derived. The general identification procedure contains many parameters to be identified for a complex system. Minimising the objective function with respect to these parameters results in a highly non-linear system of equations. In order to circumvent this, the tlvo-level identification procedure is developed and applied. If the measurement errors of the component quantities are independent of each other, the quadratic objective function can be decomposed on the basis of these components. By introducing model and goal coordination variables, the identified parameters as subsystem units are corrected at the first level and the second level includes the co-ordination units. A parallel iterative algorithm is presented in detail.