Content area
Full text
Abstract
In this investigation, four input process parameters viz. arc rotational speed, eccentricity, wire feed rate and ratio of wire feed rate to welding speed and output parameters viz. bead height, bead width and depth of penetration have been studied. The experiments were conducted on square butt joint plate of 5083-H111 aluminum alloy with arc rotation using full factorial design. The back-propagation neural network has been applied for predicting the weld bead geometry. Accuracy analysis also has been performed to find the percentage error in predicting the weld bead geometry. It is observed that back-propagation neural network can predict these output parameters with good accuracy.
Keywords: Aluminum alloy 5083-Hill, arc rotation mechanism, arc rotational speed, back- propagation neural network, eccentricity, square butt joint
1. Introduction
Pulsed Gas Metal Arc Welding (GMAW-P) is one of the most widely used processes in manufacturing industries for joining aluminum alloys. It is found that GMAW-P usually results in finger type penetration and this can be avoided by using an arc rotation mechanism. The mechanical properties of the weld joint depend on the weld bead geometry such as bead width, bead height and depth of penetration which can be controlled by the input process parameters. Some researchers have attempted arc rotation mechanism, wire bending method and arc oscillation method for improving the weld bead geometry. Rao (2004) developed an arc rotation mechanism for improving the weld bead geometry and studied the effect of arc rotational speed at constant eccentricity on output parameters. Kumar et al. (2011) investigated the effect of eccentricity and arc rotational speed on output parameters and developed nonlinear regression model for predicting the weld bead geometry. Kang and Na (2003) developed electromagnetic arc oscillation for narrow groove gas metal arc welding and studied the arc shape and bead characteristics. It was found that magnetic arc oscillation resulted in uniform and sufficient penetration to both groove faces. Several methods have been tried by various investigators to predict the weld bead geometry. Some researchers have applied statistical method for prediction of weld bead geometry. Rao et al. (2009) presented nonlinear regression model and observed that wire feed rate has maximum effect on penetration and convexity. Palani and Murugan (2006) developed second order polynomial equation to represent the response...