摘要： Relying on trial-and-error methods to determine the optimal processing parameters which maximize the density of parts produced using selective laser melting (SLM) technique is costly and time consuming. With a given SLM machine characteristics (e.g., laser power, scanning speed, laser spot size, and laser type), powder material, and powder size distribution, the present study proposes a more systematic strategy to reduce the time and cost in finding optimal parameters for producing high-density components. In the proposed approach, a circle packing design algorithm is employed to identify 48 representative combinations of the laser scanning speed and laser power for a commercial Nd:YAG SLM system. For each parameter combination, finite element heat transfer simulations are performed to calculate the melt pool dimensions and peak temperature for 316L stainless steel powder deposited on a 316L substrate. The simulated results are then used to train the artificial neural networks (ANNs). The trained ANNs are used to predict the melt pool dimensions and peak temperature for 3600 combinations of the laser power and laser speed in the design space. The resulting processing maps are then inspected to determine the particular parameter combinations which produce stable single scan tracks with good adhesion to the substrate and a peak temperature lower than the evaporation point of the SS 316L powder bed. Finally, the surface roughness measurements are employed to confirm the parameter settings which maximize the SLM component density. The experimental results show that the proposed approach results in a maximum component density of 99.97 %, an average component density of 99.89%, and a maximum standard deviation of 0.03%.