Multivariate Data Analysis of Gas-Metal Arc Welding Process

Gas-metal arc welding is a widely used welding process. The testing of such welds is done offline and in most cases after the welding operation is over. To monitor the progress of the welding run, it is essential to develop multivariate data analysis techniques that can classify the welds into good or bad runs and also be able to predict the quality variables. In this work, popular multivariate data analysis methods such as hierarchical clustering analysis, principal component analysis and partial least squares are used to develop classification and regression models to predict the weld quality based on various parameters. The results indicate that models obtained using these methods are effective in classification and prediction of weld quality and can be further developed for online and industrial uses in weld run monitoring.