Prediction of surface roughness in finish milling of 6061 aluminium alloy using artificial neural network and extreme learning machine

Tóm tắt

Surface roughness is a key indicator of surface quality in milling processes. This study presents a comparative evaluation of polynomial regression, Artificial Neural Network (ANN), and Extreme Learning Machine (ELM) models for predicting surface roughness during finish milling of 6061 aluminium alloy. Experimental data were obtained using a Taguchi L27 design with cutting speed, feed rate, and depth of cut as inputs. Model generalisation was further assessed through six additional experiments with parameter combinations outside the original design. All models achieved high accuracy within the experimental domain; however, predictive performance decreased under unseen conditions. ANN and ELM showed better generalisation than polynomial regression. Although their performance difference was moderate, ELM provided more consistent prediction accuracy across validation tests. The results highlight the importance of external validation and offer practical guidance for model selection in milling with limited data.