Efficient texture prediction for additive manufacturing

Abstract Microstructure features including grain morphology and texture are key factors in determining the properties of additively manufactured metallic components. In a digital twin of additive manufacturing, as it is still challenging to directly obtain microstructure features with sensors, microstructure control relies on the predictions from mechanistic models. However, available mechanistic models for texture prediction are too computationally expensive for digital twin applications. Here, we propose a cellular automata model, which is up to two orders-of-magnitude faster than traditional models. By adopting an exact temporal integration and a multi-level capture algorithm, a large time step can be employed without compromising the simulation accuracy. The proposed model is validated with a 316L steel case and three NiTi cases, where a good agreement is achieved between the simulations and the experiments. Our findings reveal the preferential orientations are selected by the vertical and the inclined temperature gradients from multi-pass temperature profiles, leading to different grain morphologies and textures. With a significant reduction in computational cost while maintaining accuracy, this approach marks a crucial step towards a practical implementation of digital twin for additive manufacturing.