Improvement of a Reduced Torsional Model by Means of Parameter Identification

Turbogenerator sets in operation may be excited to transient torsional vibrations by dynamic electrical moments at the generator due to short-circuits or faulty synchronization. For the solution of the torsional vibration problem it is essential to find an appropriate torsional model of the original system. A common approach is to model the torsional system finely by the finite element method which normally results in a very accurate mechanical model with many degrees of freedom (DOF). However for some applications it is desirable to have a torsional model with a reduced number of DOF which reproduces the original system exactly only in the lower eigenfrequencies and modes. This paper describes a method which allows finding a most accurate reduced torsional model with discrete masses and springs from a finite element model with many DOF. The results for the eigenfrequencies, the modes, and internal moments due to a short-circuit excitation of a 600 MW turbogenerator set are presented. They are compared with other reduction methods.