Digital Multiphase Material Microstructures for Image-Based AI Methods

The availability, cost, and producibility of components using additive manufacturing (AM) have changed how parts are designed and manufactured. These changes have highlighted the need for real-time control of production parameters to adjust product quality and to reduce or eliminate post-processing. Improved quality can be achieved by controlling the as-produced microstructure. The need for improved part quality is particularly pressing for aerospace products made from titanium alloys, due to their high material and production costs. Since mechanical properties and material microstructure are closely related, microstructure prediction for additively manufactured products has been intensely pursued. Titanium alloy microstructures with a low ratio of columnar grains and a high ratio of more fatigue-resilient equiaxed grains are preferable for aerospace applications. In addition to controlling microstructure, improved process control can mitigate the formation of porosities and residual stresses during part fabrication. The objective of this research is to provide a fast method to produce digital multiphase microstructures that will support an AI (artificial intelligence)-based deep learning model for predicting the microstructure of a multiphase Ti6Al4V alloy produced by powder bed fusion (PBF). Digitally developed microstructures, through polycrystalline generation, can be used as training samples for a deep learning algorithm (DLA).