The core of this paper is the methodology of the dynamicalmodels’ simplification for the real time simulation. The simplified simulation models are based on neuro-fuzzymodelling approach, which was originally designed for predictive control-orientedmodelling of nonlinear dynamical systems. The two ways of the neuro-fuzzymodelling utilization are presented. First, the training of the predictive dynamical neuro-fuzzymodel and, second, the training of the statical approximation of the right-hand side of the system’s state space description. We demonstrate the results on the examples of nonlinear spring damper system and double pendulum.
real time simulation, LOLIMOT, neuro-fuzzy model, identification of dynamic systems, state space