Migration Synchronous Genetic Algorithm for Reverse Engineering

Abstract

Much research adopts various evolution computation technologies to identify system parameters and structures of highly nonlinear S-system model. They, always focus on evolution skills and neglect that the choice of performance index is the key for learning. A suitable performance index not only provides good searching direction but also reduces computation time. In this study, a migration synchronous genetic algorithm (MSGA) is proposed for achieving global optimal search. Twenty eight performances of concentration- or slope-error-based indexes for parameter identification is examined and discussed. When the chosen performance candidates are used for structure identification, only one- or two-steps pruning-operation is necessary. The pruning threshold is set to be 10(-15) to ensure a safely pruning action is guaranteed positively.