Artificial neural network and regression-based models for prediction of surface roughness during turning of red brass (C23000)

In this study, models based on artificial neural networks (ANN) and regression analysis were developed to predict the surface roughness during the turning of red brass (C23000), using a high-speed steel (HSS) tool. The full factorial design approach was used for experimentation to achieve a high level of confidence. The influence of cutting parameters (cutting speed, feed rate and depth of cut) on the surface roughness was also investigated using ANOVA. The ANN model having hyperbolic tangent sigmoid (tansig) and linear (purelin) transfer functions was used for the hidden and output layers respectively. The regression model based on the power law was also developed. It was found that at a speed of 840 m/sec and depth of cut of 0.1 mm, the surface roughness changed by 11.3% when changing the feed rate by 5.6%. However, the surface roughness changed by only 6.8% when changing the velocity by 6% at the feed rate of 0.12 mm and depth of cut 0.1 mm. A similar trend was observed for different feed rates, speeds and depths of cut. It was concluded that the feed rate was the most significant factor influencing the surface roughness, followed by the depth of cut and cutting speed. The models developed were compared using statistical methods: coefficient of determination (R2 ) and mean absolute percentage error (MAPE). R 2 was found to be 0.99784 and 0.9969 for the ANN and regression models, respectively. Similarly, MAPE was found to be 1.4243% and 4.8161% for the ANN and regression models respectively. It was concluded that the ANN model is a superior choice over the multiple regression model for the prediction of surface roughness. The accuracy of the ANN can be attributed to its ability to capture the nonlinearities involved in the turning operation