Statistical analysis in gtaw welding application using full factorial design experiment

Abstract

Electrical and functional parameters determination for welding application requires preliminary tests in industry and research fields. However, this process turns out to be very expensive, slow and, for research purposes, with low reliability. Thus, the development of a statistical analysis that allows obtaining these parameters properly is very important. The present work shows and experimental methodology to apply the metal arc welding with tungsten electrode and inert gas (GTAW), used to determine the optimum torch speed and current level to achieve defect-free welding. With the results obtained in the cause-effect diagram (Ishikawa), the effective interpretation of the root cause of the defects is allowed, reducing the number of experiments, then the tree diagram gives guidelines to define the experimental matrix. The validation of the experimentation is performed by designing a complete factorial experiment, the effectiveness provided by coefficient of determination estimation or correlation level known is 94.24% in cord penetration and 85.03% in cord width. Residuals and predicted values graph are used to validate the specimens error distribution.