Abstract
This paper investigates the influence of process parameters on the surface roughness of NiTi samples fabricated using laser powder bed fusion. The investigation uses a comprehensive full factorial design of experiments across wide ranges of scanning speed and laser power. The goal is to establish a direct correlation between surface finishings and manufacturing parameters. The coefficients obtained from the analytical fittings provide valuable insights into how roughness changes with processing parameters, highlighting the major change that is only observed with a significant variation in laser power. By leveraging machine learning algorithms to analyse experimental data, the limitations of traditional analytical methods are highlighted. The machine learning models trained on a subset of experimental data successfully predict the roughness of various NiTi specimens, enabling more efficient optimization of surface quality in additive manufacturing. The greater efficiency of machine learning models is emphasized, offering a significant advantage over conventional approaches and facilitating easier deployment of predictive tools. Moreover, an analysis of the results indicates that input energy density is not generally a reliable predictor of surface quality for the fabricated samples.