METHOD OF MODEL REDUCTION FOR ELASTIC MULTIBODY SYSTEMS

In this paper we propose a new method of linear model order reduction for elastic multibody systems. One of the drawbacks of traditional reduction methods is a lack of possibility to tune a reduced model for certain transfer paths and certain frequency ranges. Our approach makes it possible to overcome these problems. The method relies on the idea of line fitting of model transfer functions. We apply the method to a model of flexible bar, evaluate properties of the reduced order model in time and frequency domains, and compare the results with the results of widely used Craig-Bampton approach. The tests show that the proposed method generates reduced models with lower order, higher accuracy, and better condition number of system matrices, but it takes more time for the construction of coordinate transformation matrix. However, the computational cost of our procedure for moderate dimensional models remains low.