Predictive Modeling of Surface Roughness for Turning of Al-6061 Using Artificial Neural Network Model

TLDR: In this article, the influence of input variables over the output parameters such as surface roughness, cutting force, and temperature on turned sample of aluminum was investigated and a predictive neural network model was developed using the experimental results obtained from the full factorial study.

Abstract: Achieving outstanding quality with minimum wastage has been an ever-standing thumb rule in manufacturing industry, for which many statistical approaches are continuously examined. This work intends to study the influence of input variables over the output parameters such as surface roughness, cutting force, and temperature on turned sample of aluminum. Four different values for each input variables such as 510–900 rpm (Spindle speed), 0.135–0.28 mm/rev (Feed rate), 0.2–1.7 mm (Depth of cut) are chosen for the present experimental investigation. Artificial intelligence is implemented in the present work and a predictive neural network model is developed using the experimental results obtained from the full factorial study. A model of 3-5-3-1 configuration is created and trained with the experimental data, which is found to have a mean absolute percentage error (MAPE) of 5.24% and mean squared error (MSE) as 0.035, for the test data. Also, the developed model is compared to a multiple regression model and found to be more accurate in predicting the surface roughness of the turned sample. Moreover, the surface roughness is found to be predominantly influenced by feed rate followed by depth of cut and cutting speed.