Neural network for composition design machine and tractor aggregate

Agrotechnical technologies are distinguished by their variety of types, conditions, methods of execution and other factors. At the same time, being divided into various types of processes, they are performed by appropriate equipment and machine-tractor parking. The machine and tractor fleet are multifunctional in terms of technical characteristics and purpose, but the use of a certain brand of machine with a unit for any, even the same type of operations can be irrational in terms of economic indicators. The problem of forming a rational, and even more so the optimal composition of a machine-tractor aggregate is extremely complex, primarily due to the fact that it is multi-parametric, both in terms of initial data and output parameters, which leads to uncertainties in accounting for the interactions of parameters of their joint influence on the technology indicators. The method of designing a machine-tractor aggregate used to this day is significantly outdated, because it was created back in 1923. At the same time, modern achievements of information technology, in particular intelligent systems, make it possible to successfully solve problems of this class. However, in the Russian agro-industrial complex, such examples aimed at improving the production operation of the fleet of agricultural machinery are practically absent. In this regard, the goal of this work was determined, which was to create a neural network for designing the most rational composition of a machine-tractor aggregate for agricultural technology processes under given conditions. The following tasks were considered: 1) justification of input and output parameters for the design of a machine-tractor unit that performs the process of plowing the soil under given conditions; 2) determination of the type, tools and formation of a training sample of the neural network; 3) setting up a neural network in the MATLAB environment; 4) checking the adequacy of the neural network on test examples. The result of the work is a tuned neural network that allows you to calculate the productivity, fuel consumption and costs of the soil plowing process based on the initial characteristics of tractors and units. Keywords: MACHINE-TRACTOR AGGREGATE; NEURAL NETWORK; HOURLY PRODUCTIVITY; SPECIFIC FUEL CONSUMPTION; THE COST OF PLOWING THE SOIL