Identification of a complex control object with frequency characteristics obtained experimentally with its dynamic neural network model

We solve the identification problem for “input-output” transmission channels of complex control objects that represent technological systems with continuous production in chemical, metal processing, mining, paper, and other industries, with neural networks. The resulting model represented with a dynamical neural network models the behavior of the technological object and lets us find the object's output, including outputs for periodic test influences. By the resulting complex frequency response, with the method of least squares we find the parameters of the channel's transfer function. We show an example of identification for “input-output” channels with a delay of the imitational object under random noise at the input.