On Joint Functioning of the Neural Network and Differential Model

Objectives: The analytical and numerical solutions of the differential equations systems in practice can be the computationally difficult task. The main objective is to improve the methods of physical processes research, speed up the mathematic model calculations. Methods/Analysis: Neural network training; Software realization of an artificial neural network observes the solution, based on the differential model of the hydrodynamic process and uses the values of channel volume consumption, obtained in the dynamic model, to train the connections weights and carry out its own prediction. Findings: The joining of dynamic model and neural network technology allows synthesizing the adaptive parallel system, which have the significant advantage in computational speed in ccomparisons with the traditional differential conception. Data compression by means of the neural network funnel permits to find the main components in the input data and reduce the time for system training. Application/Improvement: The offered method allows to significantly increasing the speed and quality of the technological object studying. It expands the field of application of such combined systems in automatic control, engineering, in the study of processes described by different types of differential equations.