MULTI-3: A GPU-enhanced meshfree simulation framework for multi-track, multi-layer, and multi-material laser powder bed fusion processes

Abstract

Multi-material laser powder bed fusion (MM-LPBF), especially with materials of contrasting properties, presents both exciting potential and significant challenges in additive manufacturing. Detailed modeling is essential for further development of these processes due to the difficulty of in-situ monitoring and control of inter-material interfaces, given the small spatiotemporal scales involved. To address this need, we present MULTI-3, a high-fidelity, GPU-accelerated computational framework designed to simulate MM-LPBF processes, including multiple tracks, layers, and materials. MULTI-3 combines a hybrid meshfree approach, leveraging a modified discrete element method (DEM) for efficient powder application and a stabilized smoothed particle hydrodynamics (SPH) technique to capture melt pool dynamics. The framework’s GPU-accelerated runtime enables the completion of a single-track LPBF simulation with modest resolution in about 29 min using a single consumer-grade graphics card. We demonstrate MULTI-3’s capabilities through a series of LPBF simulations with 316L and CuCr1Zr, employing varied deposition patterns and geometries to analyze melt pool behavior and morphology across different processing conditions. Results from these numerical experiments indicate that: (1) the SPH-DEM approach effectively addresses material mixing and interface challenges in MM-LPBF, primarily due to its Lagrangian formulation; (2) diffusion effects on interfacial material concentration remain negligible at the powder scale within the millimeter range of processing; and (3) high-fidelity, meshfree simulations of MM-LPBF processes involving multiple tracks and layers are currently achievable only through parallel computing.