Prediction of Optimal Process Parameters in Hardfacings Obtained by Using Submerged Arc Welding Technique

Abstract

In this study, hardfacings were obtained by using with submerged arc welding powders including different high carbon ferrochromium (4-16 wt.%) and currents from 400 A to 550 A. In order to determine the optimal process parameters in hardfacings, we presented two neural network based predictions with ANN algorithm for chromium and carbon percentages, secondary dendrite arms spacing, cooling rates, macrohardness, and wear loss. The results of ANN performance were presented at two sets of FeCr (wt.%) and net heat inputs in detail. Similarly, two ANN architectures were preferred to obtain the most accurate results. The performance evaluations of the networks were carried out by using Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and the Coefficient of Correlation, R2, for all models. The models having architecture of 2-15-3 and 2-23-4 were found to be optimal after these criteria. The results showed that the ANNs which helped to decrease number of experimental tests had an acceptable degree of accuracy and great reliability.