Neural Network Modeling of Microstructure Formation in an AlMg6/10% SiC Metal Matrix Composite and Identification of Its Softening Mechanisms under High-Temperature Deformation

Abstract

The paper investigates the rheological behavior and microstructuring of an AlMg6/10% SiC metal matrix composite (MMC). The rheological behavior and microstructuring of the AlMg6/10% SiC composite is studied for strain rates ranging between 0.1 and 4 s−1 and temperatures ranging from 300 to 500 °C. The microstructure formation is studied using EBSD analysis, as well as finite element simulation and neural network models. The paper proposes a new method of adding data to a training sample, which allows neural networks to correctly predict the behavior of microstructure parameters, such as the average grain diameter, and the fraction and density of low-angle boundaries with scanty initial experimental data. The use of neural networks has made it possible to relate the thermomechanical parameters of deformation to the microstructure parameters formed under these conditions. These dependences allow us to establish that, at strain rates ranging from 0.1 to 4 s−1 and temperatures between 300 to 500 °C, the main softening processes in the AlMg6/10% SiC MMC are dynamic recovery and continuous dynamic recrystallization accompanied, under certain strain and strain rate conditions at 300 and 350 °C, by geometric recrystallization.