M.V. Shcherbakov - Ph.D. (Eng.), Assistant Professor, Department of Computer-aided design and exploratory design, State Educational Institution «Volgograd Technical University». E-mail: maxim.shcherbakov@vstu.ru
D.A. Skorobogatchenko - D.Sc. (Eng.), Assistant Professor, Department of of Economics and Management of road services, State Educational Institution «Volgograd State University of Architecture and Civil Engineering». E-mail: Skor2004@rambler.ru
Tran Hung Anh - Master of Engineering and Technology, State Educational Institution «Volgograd Technical University». E-mail: Hunganh.tr@gmail.com
N.L. Shcherbakova - Ph.D. (Eng.), Assistant Professor, Department of experimental physicist, State Educational Institution «Volgograd Technical University». E-mail: natalia.shchrbakova@gmail.com

The problem of identification of dynamical systems in various productions is an important task in many domains. It should be at least three problems associated with the identification of models of complex systems. The first is the possibility of storing information about the behavior of the system in near real time and transfer that data to the data center. The second - the methods for measuring physical characteristics when measured frequency values, not the amplitude. Finally, the third issue concerns finding effective authentication mechanisms in nature. One such universal approaches to solving these problems is the development of the third generation of neural networks, which are called - pulsed neural networks. However, pulsed neural networks have a number of disadvantages that are associated with excessive accumulation of output pulses in the operation of multi-impulsive neural networks as well as the lack of sensitivity of the model to the input and output pulse in the training patterns. In an attempt to create a mathematical model which excludes the listed deficiencies. An algorithm for parametric optimization of pulsed neural network of this structure. We consider the situation and the availability of the desired model time impulse response of the neuron. If there is no pulse synchronous correction procedure is proposed, in which the initial value of the weights of presynaptic terminals added a small value, leading to the emergence of at least one pulse. The application of the model for different test problems.