Modeling and prediction of surface roughness at the drilling of SLM-Ti6Al4V parts manufactured with pre-hole with optimized ANN and ANFIS

Abstract

In this study, modeling of the surface roughness (Ra) with artificial neural networks (ANN) in drilling the desired diameter of Ti6Al4V alloy produced by Selective Laser Melting (SLM) method is discussed. Drill type, cutting speed and feed were determined as evaluation parameters. Multilayer perceptron (MLP), Learning Vector Quantization (LVQ) and fuzzy-Mamdani methods were used in the study. In the estimation of Ra, a MAPE value of 0.206 % for MLP and 8 % for Mamdani was obtained. 100 % success was achieved in the classification made with LVQ. The ANN model accurately predicted Ra with excellent nonlinear mapping ability and an error of 4.9 %. It has been demonstrated that the presented models for drilling SLM-Ti6Al4V parts produced with unsupported front holes instead of drilling full holes can determine the appropriate drilling parameters for the desired Ra. The forecast results and presented models can be applied in drilling process planning for target industrial applications.