Microstructure prediction in low and high strain deformation of Al6063 using artificial neural network and finite element simulation

The final microstructure of materials under interactive effects of critical elements of deformation: strain, strain-rate, and temperature, often follows complex trajectories. Capturing the existing process-structure linkages is fundamental for controlling product outcomes, yet it calls for establishing the constitutive relationships that describe material behavior. In this paper, a backpropagation Artificial Neural Network (ANN) is proposed for microstructure prediction for a wide range of strain, strain-rate, and temperature conditions. Microstructural changes in Al6063 are experimentally examined using (i) quasi-static universal testing apparatus, Drop Weight Impact Tester (DWIT), and Split Hopkinson Pressure Bar (SHPB) tests for low strain regimes at elevated temperatures, and (ii) Plane Strain Machining (PSM) for high strain, high strain-rate, and the accompanied temperature rise conditions. Two ANNs are established to predict microstructure responses, grain and subgrain sizes, in low and high strain regimes, respectively. Additionally, the stress-strain results obtained from the low strain regime are used to calculate the Johnson–Cook (J-C) material model constants, which are then incorporated in the finite element (FE) simulation along with the developed ANN algorithm, to predict microstructure response for different cutting conditions. The performance of the ANNs and the FE simulations was evaluated using statistical indices. The comparative assessment of the models’ outcomes indicates close agreements with the experimental results in both low- and high-level deformations. The accurate predictions from PSM conditions can potentially be applicable for microstructural prediction of the machined surface.