The results of field studies of the yield of spring barley are considered in order to use them to build a neural network model. The main stages of obtaining the model are described. An accuracy of about 2% was obtained, which makes it possible to use the model for research and practical activities. The model was verified, which showed new ways of conducting scientific research on the yield of grain crops.

Modern farm machinery is equipped with on-board computers that execute complicated tasks. The main efforts are aimed at selecting and justifying methods for adequate hardware-software translation of functional programming from one computer system to another, while increasing its stability to failures of computer hardware. The development of approaches to the selection of procedures for detecting errors determines the degree of their relative information value when analyzing the programs progress. As it follows from the analysis of error types in functional programming, it is established that errors occur at the stages of forming the technical task, when developing algorithms, when programming, and when configuring programs. Errors have different frequency and different weight in terms of the impact on the functioning of on-board computers used in agricultural machinery. The article contains a detailed list of errors that must be detected when debugging working programs of on-board computers. Based on expert assessments, the errors are divided into 33 different types and formed into 6 groups. By combining errors, it is possible to minimize the time spent on performing static debugging of the functional programming of on-board computers. Using the hierarchy analysis method, matrices of paired comparisons of control directives with error types in each group are formed. Thus, the relationship between groups of errors, control directives, and values of error types is established. The first thing, programs debugging is performed using the control directives that have the highest degree of significance, since they can be used to identify and eliminate the most dangerous errors.