Machine learning approach for prediction analysis of aluminium alloy on the surface roughness using CO2 laser machining

Abstract

A comprehensive analysis investigated the impact of cutting speed, nozzle diameter, gas pressure and the addition of SiC and ZrO2 particles on the surface quality of aluminum alloy 6062. The correlation between experimental and predicted values was established using deep neural network (DNN), support vector machine regression and response surface methodology. To validate the models, root mean squared error and mean absolute error were computed for four hidden layers with the DNN approach. The surface roughness was significantly affected by the higher cutting speed (3000 mm/min) and lower nitrogen gas pressure (10 bar). The results from the developed models closely matched experimental data. Additionally, the study analyzed the impact of laser parameters on crack width due to rapid thermal changes. The scanning electron microscopy, energy-dispersive X-ray spectroscopy and optical microscopy were utilized to examine the laser-cut surface's microstructure for crack formation analysis.