Estimation of surface roughness of direct metal laser sintered AlSi10Mg using artificial neural networks and response surface methodology

Abstract

Direct metal laser sintering (DMLS) is a metal-specific additive manufacturing (AM) technique that has grown in efficiency and precision due to compelling advancements in high-power lasers and fiber optics. This study examines the surface roughness of AlSi10Mg specimens manufactured additively using the DMLS technique. First, DMLS experiments were conducted with a range of control variables, including laser power, laser speed, orientation, and post-heat treatment temperatures. Later, surface roughness prediction models were developed using machine learning techniques and statistical methods such as artificial neural networks (ANN) and response surface methodology (RSM). The ANN model with an architecture of 4-9-9-1 is identified as the optimal network. The predictions of the ANN models were compared to those of the RSM models, and performance was quantified using the correlation coefficient (R-value) between predictions and the experimental data. The R-value of 0.96218 with experimental data and the least mean absolute percentage error (MAPE) of 0.9804% indicated that ANN predictions were more accurate than the RSM model estimates. Conclusive results prove that the developed ANN model accurately estimated the relationship between DMLS process parameters and surface roughness.