Instruments and Systems: Monitoring, Control, and Diagnostics Annotation << Back Using neuron networks with fractional order modeling of complex control objects E.B. KARIKOV G. RUBANOV This article describes the process of modeling of complex control objects on the example of the grate cooler “Volga-75”. Performed an analysis of the object. Selected set of input and output signals to simulate the heat transfer process. Been made the analysis of existing approaches to modeling based on recurrent neural networks. Performed modification of the NARX-network by using fractional-order delay elements. In the System MATLAB Simulink done testing this approach on the control object. Performed configuring network. Obtained the model of heat transfer in the grate cooler and assess the adequacy of the model. Research results have shown that the use of dynamic recurrent neural networks with fractional delay elements in the process of developing a mathematical model of the control object can significantly accelerate this process, the obtained models are suitable for solving the synthesis of control laws and the structure of the automation system. Keywords: grate cooler, mathematical model, neural networks, NARX network, fractional order models. Contacts: Email: rubanov@intbel.ru Pp. 18-25.