Introducing of linear dependence measure into multipoint projection techniques for model order reduction

The new strategy of multipoint projection for model order reduction (MOR) is proposed that is directed to decrease redundancy of the reduced models. The developed MOR algorithm takes into account the mechanism of possible occurrence of redundancy at the stage of basis generation for projective matrix. The numerical procedure is based on estimation of maximal linear independence while subsequent selection of state vectors for basis generation. The compact low-order models are obtained by the developed adaptive procedure to determine frequency points for basic vectors under current error control. Some experimental results are presented.