In this work, a concept of using surface roughness data as an evaluation tool of the process variation in a commercial Laser Powder Bed Fusion (L-PBF) machine is demonstrated. The interactive effects of powder recoating, spatter generation, gas flow and heat transfer are responsible for the intra-build quality inconsistency of the L-PBF process. Novel specimens and experiments were designed to investigate how surface roughness varies across the build volume and with the progression of a build. The variation in roughness has a clear and repeatable pattern due to the strong impact of the orientation of inclined surface to the laser origin. The effects of other factors such as exposure sequence of specimens, build height, and recoating process are less prominent and are difficult to isolate. A neural network regression model was built upon the large dataset in measured Ra values. The neural network model was applied to predict distribution of roughness within the build volume under hypothetical processing conditions. Connections between the predicted variation in roughness and underlying physical mechanisms are discussed. The present work has value for machine qualification and modifications which lead to the manufacturing of parts with better consistency in quality. The detailed variation observed in surface roughness can be used as a reference for designing experiments to optimise processing parameters in order to minimise the roughness of inclined surfaces.

在这项工作中，展示了使用表面粗糙度数据作为商业激光粉末床熔合（L-PBF）机器中工艺变化的评估工具的概念。粉末重涂，飞溅物产生，气流和传热的相互作用是造成L-PBF工艺内部质量不一致的原因。设计了新颖的标本和实验，以研究表面粗糙度在整个构建体体积以及构建过程中如何变化。由于倾斜表面的方向对激光原点的强烈影响，粗糙度的变化具有清晰且可重复的图案。其他因素（例如试样的暴露顺序，成型高度和重涂过程）的影响不太明显，并且很难隔离。在测得的Ra值的大型数据集上建立了神经网络回归模型。应用神经网络模型来预测假想加工条件下构建体积内粗糙度的分布。讨论了粗糙度的预测变化与潜在物理机制之间的联系。当前的工作对于机器的鉴定和修改具有价值，这使得零件的制造具有更好的质量一致性。观察到的表面粗糙度的详细变化可以用作设计实验的参考，以优化处理参数，以最大程度地减小倾斜表面的粗糙度。讨论了粗糙度的预测变化与潜在物理机制之间的联系。当前的工作对于机器的鉴定和修改具有价值，这使得零件的制造具有更好的质量一致性。观察到的表面粗糙度的详细变化可以用作设计实验的参考，以优化处理参数，以最大程度地减小倾斜表面的粗糙度。讨论了粗糙度的预测变化与潜在物理机制之间的联系。当前的工作对于机器的鉴定和修改具有价值，这使得零件的制造具有更好的质量一致性。观察到的表面粗糙度的详细变化可以用作设计实验的参考，以优化处理参数，以最大程度地减小倾斜表面的粗糙度。