Global solutions for nonlinear systems using qualitative reasoning (Q1277335)

The paper is devoted to the problem of determining the order and the structure of nonlinear models, used for nonlinear dynamic system identification. Particularly it considers the dependence of the optimal solution on starting values of unknown system parameters, which are of crucial importance when the system behavior is chaotic. The paper explores how qualitative information can be used to improve the performance of global optimization procedures. A nonlinear parameter estimation reasoner is proposed for finding the values which match an ordinary differential equation model to observed data. In such a way it becomes possible to choose intelligently starting values for the unknown parameters and to avoid local extrema in the regression landscape. Via Monte Carlo simulation, it is shown that the proposed nonlinear parameter estimation reasoner is robust and effective. An application case is considered and a computer program is described. The proposed paper could be of interest for engineers and specialists, working in the applied area of nonlinear system identification and especially, application of practical methods of computer science in nonlinear dynamic system modeling and design.