MULTI-TIME-SCALE SYSTEMS MODEL ORDER REDUCTION VIA GENETIC ALGORITHMS WITH EIGENVALUE PRESERVATION

Abstract

A novel substructure (dominant eigenvalue) preserving genetic algorithm approach for model order reduction (MOR) of multi-time-scale systems is presented in this paper. The new technique is formulated based on genetic algorithms (GAs), sub-optimization and estimation. The GA predicts the elements of an upper triangular matrix form of the system state matrix A, defined in state space representation along with the elements of B, C, and D matrices. The GA procedure is based on maximizing the fitness function corresponding to the reciprocal response deviation between the full order model and the reduced order model. The proposed GA model order reduction method is compared to well-known reduction techniques such as the Balanced Schur Decomposition (BSD), proper orthogonal decomposition (POD), and state elimination through balanced realization. Simulation results validate the robustness of the new technique for MOR with eigenvalue preservation.