Mechanical properties of metallic parts are closely related to defects and microstructure. Additive manufacturing of metals (sometimes considered as micro-welding [1]) involves a melt pool whose shape and local thermal gradients are responsible for the formation of defects such as porosity and microstructure. Many studies have been dedicated to process parameters optimization in order to reduce the amount of defects, which at the first order drive important material parameters such failure properties [2]. In addition many studies focus on process parameters optimization to better control thermal history and melt pool geometry in order to better control phase mixture and residual stresses [3-5], and microstructure [6].

This contribution focuses on fast estimation of microstructure as most of existing numerical methods rely on very fine scale, which implies significant computation time that hinders the development of optimization loops at the scale of the entire process. For instance Phase Field (PF) approaches [7] consider the detailed geometry of individual dendrites, and even though Cellular Automaton (CA) techniques [6] enable to reduce computation time significantly, crystal growth is controlled by pixel by pixel, which involves large numbers of degrees of freedom and limits the approach to small size samples.

This contribution focuses on a numerical method to estimate microstructure at the scale of the entire process within sufficiently short computation to perform optimization. The proposed approach relies on (i) a fast estimation of temperature and thermal gradients in the melt pool vicinity [3], (ii) solidification maps [8] to determine the region of equiaxed grains with random crystal orientations from which epitaxial columnar competitive growth takes place, and (iii) a new algorithm relying on Voronoi-Laguerre tessellations to generate the corresponding microstructure [9,10]. The competitive growth and resulting crystallographic texture are accounted for by using classical criteria based on easy growth direction of dendrites (i.e., <100> for cubic lattices), while the morphological texture is obtained by considering that the solidification front evolves along the local thermal gradient. Various examples of inverse pole figures will be presented, as well as statistical analysis of a large number of microstructures obtained for different draws of random crystal orientation of equiaxed grains.