The Reiterated Neural Network Parametric Identification of Nonlinear Dynamic Models of Objects

We posed a problem and developed an algorithm for the neural network parametric identification of nonlinear dynamic models of objects with a computational experiment, the formation of training samples on its basis and the subsequent sequential training of two neural networks. Neural networks perform bijective mapping of object parameters into the original model to the output variables of the second neural network. Sequential training or sequential bijective identification of a neural network consists in preliminary training of the first neural network based on experimental data and using its synaptic coefficients to train the second neural network. An example of parametric identification of a nonlinear dynamic model will be a 600 W high pressure sodium lamp. The computational experiment was carried out in the MATLAB environment. The computational experiment technique and the results of experimental studies are presented in the article. Taking into account the good approximating ability of neural networks, the proposed algorithm can be considered as an effective method for parametric identification of nonlinear models.