In Line Nondestructive Testing for Sheet Metal Friction Stir Welding

As automotive designs add more aluminum to lightweight their vehicles, friction stir welding (FSW) will likely become a principal joining process in the industry. FSW is a solid-state joining process which avoids many of the traditional problems of welding aluminum alloys such as hot cracking, porosity and solidification shrinkage. These attributes enable high preforming friction stir welded joints of cast, 5XXX, 6XXX, 7XXX or mixed aluminum alloy combinations. Although FSW technologies have advanced to support high volume applications and have been applied in current automotive parts, its inability for nondestructive evaluation (NDE) increases the cost to manufacture friction stir welded parts. Current state of the art NDE methods for FSW are either ultrasound or radiographic technologies which add complexity to manufacturing lines and additional costs to FSW production. Many have researched ways to reduce NDE costs by using measured forces of the FSW process. These methods have included trained neural networks (NN) that result in accurate defect predictions that can be applied in an industrial setting. Although NN provide an alternative solution to traditional NDE methods, they require large amounts of training and can only inspect welds that share exact welding parameters and machinery that were included in the training. An ideal FSW NDE method would reduce costs and be able to be applied on multiple welding machines and with a variety of parameters. The cost of a cited generalized force based stochastic NDE method in an industrial setting will be validated by an automotive production example here in.