Functional clustering methods for resistance spot welding process data in the automotive industry
share
Quality assessment of resistance spot welding (RSW) joints of metal sheets in the automotive industry is typically based on costly and lengthy off-line tests that are unfeasible on the full production, especially on large scale. However, the massive industrial digitalization triggered by the industry 4.0 framework makes available, for every produced joint, on-line RSW process parameters, such as, in particular, the so-called dynamic resistance curve (DRC), which is recognized as the full technological signature of the spot welds. Motivated by this context, the present paper means to show the potentiality and the practical applicability to clustering methods of the functional data approach that avoids the need for arbitrary and often controversial feature extraction to find out homogeneous groups of DRCs, which likely pertain to spot welds sharing common mechanical and metallurgical properties. We intend is to provide an essential hands-on overview of the most promising functional clustering methods, and to apply the latter to the DRCs collected from the RSW process at hand, even if they could go far beyond the specific application hereby investigated. The methods analyzed are demonstrated to possibly support practitioners along the identification of the mapping relationship between process parameters and the final quality of RSW joints as well as, more specifically, along the priority assignment for off-line testing of welded spots and the welding tool wear analysis. The analysis code, that has been developed through the software environment R, and the DRC data set are made openly available online at https://github.com/unina-sfere/funclustRSW/
READ FULL TEXT