Artificial Neural Network Modeling of Ti-6Al-4V Alloys to Correlate Their Microstructure and Mechanical Properties

Abstract

The heat treatment process of Ti-6Al-4V alloy alters its microstructural features such as prior-beta grain size, Widmanstatten alpha lath thickness, Widmanstatten alpha volume fraction, grain boundary alpha lath thickness, total alpha volume fraction, alpha colony size, and alpha platelet length. These microstructural features affect the material's mechanical properties (UTS, YS, and %EL). The relationship between microstructural features and mechanical properties is very complex and non-linear. To understand these relationships, we developed an artificial neural network (ANN) model using experimental datasets. The microstructural features are used as input parameters to feed the model and the mechanical properties (UTS, YS, and %EL) are the output parameters. The influence of microstructural parameters was investigated by the index of relative importance (IRI). The mean edge length, colony scale factor, alpha lath thickness, and volume fraction affect UTS more. The model-predicted results show that the UTS of Ti-6Al-4V decreases with the increase in prior beta grain size, Widmanstatten alpha lath thickness, grain boundaries alpha thickness, colony scale factor, and UTS increases with mean edge length.