Prediction of Bead geometry in Gas Metal Arc Welding by Statistical Regression Analysis

Abstract

Manufactures in fabrication industry have long depended on Gas Metal Arc welding as one of the most reliable and economical techniques for joining parts. The need for reducing costs and addressing a skills shortage on the shop floor, often limits productivity in the manufacturing industry involving welding. Developing a welding procedure for instance, involves a lot of time and cost, and currently there is a gap in welding industry where few available methodologies can predict even the basic features of weld geometry based on input parameters. To provide a methodology for prediction, the aim of this thesis focusses on developing a statistical model for the weld inputs and outputs in the Gas Metal Arc Welding process to predict the various geometries of the weld bead. To study the effect of welding process parameters—such as wire feed speed, voltage, travel speed and gas type—on the resultant bead geometry such as bead width, penetration, reinforcement height, reinforcement area and penetration area a factorial design of experiment was used. Low carbon electrode (ER 70S-6) of two different diameters was used, and a total of 242 welds were made with 121 for each wire diameter. Two cross sections were cut from each weld bead and the geometries were measured, and a linear regression analysis was performed to develop a statistical model for each of the bead geometry based on experimental data. Analysis of variance (ANOVA) indicated the significant squared and interaction variables for each of the bead geometry with 95% confidence interval. The trends of geometry for each diameter varied with gas type. Residual analysis revealed that all assumptions inherent in the regression analysis were satisfied. Finally the statistical models were validated in bead on plate, fillet and V-groove joint positions. A total of 8 fillet tests and 5 groove tests were performed for both the wire types and it was found that predicted values were in good agreement with the measured values for bead on plate and fillet conditions, whereas welds with a V-groove joint geometry had a significantly under-predicted penetration area due to increased heat transfer and faster cooling down of weld with a higher equivalent sample thickness.