Modelling of large scale linear discrete time interval systems using K-means algorithm

This paper proposes a new method for the reduction of discreet time interval systems based on pattern recognition algorithm. The proposed method always yields a stable reduced order model for a stable original higher order system. As the method does not have random assumptions in it, the method gives a unique reduced order model. The denominator polynomial is formulated using K-Means algorithm and numerator polynomial is formulated by matching the ration of like powers of original system with that of the reduced order model. Numerical examples were presented and compared with existing methods in the literature.