Performance of Bayesian Regularization Neural- Network in Prediction of Surface Roughness

Surface roughness is one of the essential quality characteristics that must be precisely controlled. Artificial neural network modeling is a method of surface roughness prediction. In this study, two different networks are used for surface roughness prediction in turning processes. Experimental data for turning of 6061- T6511 Aluminum alloy, obtained from literature were employed to train the ANN models. These ANNs were trained by Levenberg – Marquardt and Bayesian regularization algorithms. Results show that the Bayesian regularization network has better prediction accuracy in comparison with Levenberg-Marquardt algorithm. A comparison of ANN model with regression model was also carried out.

کلیدواژه ها:

Bayesian regularization ، Neural network ، Surface roughness prediction