Aerospace Instrument-Making Annotation << Back APPLICATION OF NEURAL NETWORKS TO DETERMINE THE TIGHTNESS OF VALVES IN AVIATION SYSTEMS D.O. Pushkarev, D.Yu. Kiselev, Yu.V. Kiselev The paper analyzes the tightness requirements for metal-on-metal connections for a shut-off valveseat pair for aircraft products. It was found that there are no formalized methods for assessing tightness before the testing stage of prototypes. In this paper, a neural network is constructed in order to establish the tightness of products before the testing stage, as well as the selection of optimal requirements for products. For this study, an analysis of the parameters of the products was performed, criteria for assessing tightness were selected, and a neural network was built based on these data. The neural network was verified. The paper shows the features of neural networks, as well as recommendations for the creation, training and improvement of neural networks. Keywords: neural network; tightness; valve; test; machine learning; neuron. DOI: 10.25791/aviakosmos.4.2022.1275 Pp. 34-43.