The ability of Powder Bed Fusion (PBF) to create complex geometries across a wide range of materials makes PBF a widely used powder-based metal additive manufacturing (AM) process in various industries for advanced applications. However, compared to conventional manufacturing processes, the metal parts printed by PBF exhibit lower surface quality due to soot and spatter particles arising from laser-powder interaction. To minimize spatter and soot generation during the build, PBF systems are equipped with cross-flow nozzles that are designed to flow inert gas across the build platform. It is desired that these gas flow systems have the ability to remove most of the spattered powder from the build chamber, but do not erode the freshly spread layer of powder on the to-be-printed surface to ensure high-quality manufactured parts. The onset of particle bed erosion can be characterized by the critical Shields number. Once the critical Shields number is known for the metal powders and system of interest, the flow of inert gas in the build chamber can be optimized to ensure the build process is efficient and clean. This work proposes a Shields number-based method for obtaining engineering design guidance for PBF gas flow systems to optimize the spatter removal process. A combined experimental and Computational Fluid Dynamics (CFD) study was performed to provide design guidance for these cross-flow systems. All experiments were conducted using a small, closed-loop wind tunnel, with built-in flexibility, capable of testing a number of cross-flow configurations. A high-speed camera captured the threshold of particle movement at a variety of operating conditions for various metal powders used in metal AM including aluminum alloy AlSi10Mg, nickel-based superalloy Inconel 718, titanium alloy Ti-6Al-4V, steel alloy 4340, and 316L stainless steel. Time-averaged flowfield measurements of the gas flow inside the test section were made using particle tracking velocimetry (PTV) and a hot-wire air flow meter at the same conditions. Using these experimental measurements and attendant CFD simulations, CFD predictions of wall shear stress can be used to calculate the Shields number at the condition of incipient movement as identified experimentally.