The non-linear chaotic model reconstruction for the experimental data obtained from different dynamic system

The non-linear chaotic model reconstruction is the major important quantita tive index for describing accurate experimental data obtained in dynamic an alysis. A lot of work has been done to distinguish chaos from,randomness, t o calculate fractral dimension and Lyapunov exponent, to reconstruct the st ate space and to fix the rank of model. In this paper, a new improved EAR m ethod is presented in modelling and predicting chaotic timeseries, and a su ccessful approach to fast estimation algorithms is proposed. Some illustrat ive experimental data examples from known chaotic systems are presented, em phasising the increase in predicting error with time. The calculating resul ts tell us that the parameter identification method in this paper can effec tively adjust the initial value row ards the global limit value of the sing le peak target Junction nearby. Then the model paremeter can immediately be obtained by using the improved optimization method rapidly, and non-linens chaotic models can nor provide long period superior predictions. Applicati ons of this method are listed to real data from widely different areas.