Control of weld quality in large welded populations

Abstract

The use of factorial experimental design during weld procedure development, as a statistical tool for the control of weld quality in large welded populations, is investigated. As an example, it is applied to single pass welds made in the laboratory using the spacer to tube attachment geometry present in AGR boilers. The welds have been deposited using a robotic cold wire TIG technique. A preliminary welding trial has been made to identify a potential working point welding condition and the significant welding parameters. With this experience, a replicated 2 5 factorial welding trial has been designed and evaluated experimentally. Transverse sections have been examined to determine weld penetration quality features. Rapid data accumulation and processing has been achieved by using a digital graphics tablet and a desk top computer. The means and variances of each quality feature have been calculated. Using a standard factorial analysis program the significant effects have been identified and incorporated into control equations. Using these data and having established minimum acceptance criteria, defect rates have been calculated as a function of weld position. The control equations have also been used to revise working point welding conditions, to achieve acceptable weld quality. These equations have also been used in conjunction with the reproducibility measurements to estimate the amount of weld quality variability that can be attributed to process control. The use of these estimates to influence the selection of welding equipment both in the laboratory and on the shop floor is also discussed.