A large-scale GPU-accelerated DEM–ray-tracing method for particle-resolved thermo-optical-mechanical simulation of selective laser sintering of polymer powders

Abstract

Selective laser sintering (SLS) involves strongly coupled laser transport, particle-scale heat transfer, particle motion, and sintering, making high-fidelity simulation of realistic powder beds computationally demanding. This work presents a large-scale GPU-accelerated DEM–ray-tracing framework for particle-resolved thermo-optical-mechanical simulation of polymer SLS. Deterministic, particle-interface-resolved optical transport is retained throughout the transient DEM simulation, with reflection, refraction, internal attenuation, particle exit, and subsequent transmission to neighboring particles explicitly tracked. An ordered grid-accelerated nearest-intersection procedure, using the analytical ray–sphere solution as the narrow-phase test, is integrated with the DEM neighbor-search structure through a shared GPU-resident background grid. The resulting ray-resolved laser heat input is consistently coupled with particle motion, interparticle heat transfer, sintering displacement, and bond formation. The framework is verified against ray-level optical solutions and validated using powder-bed energy-deposition and single-track sintering experiments. In a serial CPU benchmark designed to isolate algorithmic cost, the analytical intersection treatment reduced the runtime by up to 22.5 times relative to a grid-marching ray tracing implementation. Comparisons with Gaussian volumetric heat-source models show deviations of up to 80.23%, highlighting their limitations in representing heterogeneous energy deposition within discrete powder beds. Full-process simulations involving more than one million particles further demonstrate the capability of the framework to resolve temperature evolution, sintering morphology, and part formation. The proposed method provides a large-scale particle-resolved computational framework for investigating laser–powder interactions and process–structure evolution in polymer SLS.