Comparison Between RSM and ANN Models To Predict Carbon Content Equivalent In a Tig Weld

Abstract

Prediction of process parameters beyond design of experiment is a

limitation for some traditional models like the response surface

methodology. In this study a comparative analysis between the

response surface methodology and the artificial neural network

algorithm to predict carbon content present in welding process of

mild steel was presented. The RSM model explored its numerical and

graphical techniques to predict the carbon content. The results

obtained showed that the RSM prediction did not fit perfectly into

the observed values; the ANN model was also used to predict the

carbon content. The ANN employed the process of training,

validation and testing with the help of hidden neurons. The ANN

model predictions fit perfectly into the plot of the observed values.

The ANN is a better predictive model compared to the RSM.