Comparative analysis of the application of different types of neural networks to the recognition of one-dimensional signals

In the processes of determining the properties of materials and structures based on the study of the response to a given dynamic impact, the problem of analysing a one-dimensional time signal and its classification arises. One of the effective approaches to solving it is the use of artificial neural networks with generalized properties of approximation and data filtering. The paper investigates the effectiveness of using fully connected, recurrent and convolutional neural networks to problems of impact indentation to determine the strength properties of metals and elastic moduli of layered structures of non-rigid highways.