Signal processing for quality assurance in friction stir welds

Abstract

This paper is based upon the project 'Qualistir' for the on-line quality control of FSW in aluminium. Friction stir welding (FSW) is a solid-state bonding technique. Because it is an automated technique it is a controlledand reliable process. However, with changes in material condition, dimensions or welding parameters, flaws can be generated. These flaws can be conventional ones (lack of penetration or voids), or joint line remnants, which are unique to this welding technique. The latter flaw has been called a 'kissing bond' and is extremely difficult to detect with any non-destructive testing method due to its sub-millimetre size and transparency to NDT. Hence direct ultrasonic detection of these defects by back-reflected energy cannot be achieved reliably. Analysis of ultrasonic noise distribution, however, has provided a means to assess the quality of the weld root, where the joint line remnants can cause the greatest reduction in mechanical properties. A mathematical algorithm comparing the noise level within the weld root to that within the parent metal was developed to provide a quantitative inspection method. Signal processing software imports 3D volumes of the data from the specified areas within the weld and parent plate; it then processes the acquired noise amplitude data to obtain a weld quality indicator. Hence, instead of detecting the defect directly, the procedure essentially quantifies the quality of the weld, thereby determining whether the weld nugget has been correctly forged.