Abstract: As an advanced metal additive manufacturing technology, selective laser melting (SLM) is capable of fabricating metal components rapidly with excellent mechanical properties and complex geometries. One of key foundations for producing high-performance components using SLM is a relatively uniform distribution of metal powder particles in compact powder layers, which is affected by powder flowability, powder spreading speed, and layer thickness. Here, the discrete element modeling for obtaining particle properties, analyzing effects of spreading processes on powder layer quantity and efficient simulations on the multi-layer spreading process are performed. The rolling friction coefficient and surface energy density of tungsten particles are obtained by comparing experimental and numerical results of repose angle and relative density of the tungsten powder pile. Effects of rolling friction coefficient and surface energy density on particle flowability are revealed, and all physical parameters of tungsten particles required in the discrete element modeling are obtained. Effects of powder layer thickness and spreading speed on the layer quality are quantitatively examined based on the discrete element modeling of single-layer spreading process. Through evaluation on packing density, coordination number distribution, layer surface roughness and layer uniformity, the spreading process window is determined for a powder layer consisting of closely and uniformly distributed metal powder particles. By identifying completely unfused metal powder particles in the melted powder bed, a new discrete element model capable of efficiently simulating the multi-layer spreading process in a physical way is established. In the new discrete element model, rough surfaces of fused metal parts are characterized accurately, and completely unfused metal powder particles are represented by movable spherical particles which enables the elimination of unphysical limitation of the movement of completely unfused metal powder particles. Having these advantages, the new discrete element model significantly improves the efficiency and fidelity of simulations on the multi-layer spreading process.