Method of operation of the module of the management system for segments of assemblies and objects of the agro-industrial complex using neural networks to predict their condition

Abstract

The research objective is to analyze the application of the system for evaluating individual units (equipment objects) of production at an agro-industrial complex enterprise. The complexity of using forecasting methods consists in their time costs to obtain a more accurate forecast. The objective included obtaining an acceptable and sufficient accuracy when using a neural network trained on a small amount of data. Neural networks have already shown themselves well in various industries, such as IT and network technologies, and autopilots. An object at an agro-industrial complex enterprise is a technical device which normal operation ensures the operation of the production cycle. Modern production requires digitalization. The operation of equipment should be controlled. But there should also be a system that allows assessing the condition of the equipment object. The use of neural networks that allow extracting useful information and arrays of statistical data provides an opportunity for research in the field of their application for solving management problems in the agro-industrial complex. This paper discusses the research findings in the field of application of neural networks in the agro-industrial complex for predicting the state of an object. It should be taken into account that the object is a piece of equipment in the production cycle, and it may be influenced by external factors which is expressed in one of the characteristics. The results are presented as a numerical indicator of the state of the object based on neural network data. The results may be used as an option for the use of modern neural networks in the agro-industrial complex in the work of control and management, and dispatching tasks. Further research is aimed at studying the patterns of operation of individual devices. A method for controlling the operation of units of production facilities of the agro-industrial complex was studied. The methods of mathematical modeling and the apparatus of neural networks were used. As part of working with a small data set, the neural network showed good results when evaluating a particular object. Root mean square error was 0.121; standard deviation: 0.091.