Carbon Alloying of Metal Matrix Composites Based on Fe – Cr – Mn – Mo – N – C Alloys During Their Manufacturing by the Aluminobarothermic Variant of the SHS Method

Abstract

Metal matrix composites based on Fe – Cr – Mn – Mo – N – C system and obtained by the aluminobarothermic variant of self-propagating high-temperature synthesis (SHS) are studied. The possibility of uniform carburizing of a melt obtained by aluminobarothermic synthesis in the cooling crucible is shown. An artificial neural network model is suggested, which makes it possible to predict the carbon content in the studied composite during carburization in the cooling crucible of an SHS reactor (the average approximation error is 9 – 14% depending on the training method). The results of training of the artificial neural network model using the Adam optimization algorithm and the Levenberg–Marquardt method are compared. It is shown that under the conditions of a limited set of initial data, it is effective to use a perceptron with one hidden layer containing three target neurons and one displacement neuron.