Fatigue of Aluminum Gas Metal Arc Welds in Electric Vehicle Battery Pack

Abstract

Aluminum extrusions, when strengthened with precipitation hardening, are an ideal material for lightweight structures. The gas metal arc welding (GMAW) process offers a cost effective and high-volume method of joining mating plates for large structural components. As the automotive industry is looking for lightweight structures to offset the increased weight of electric vehicle battery packs, it is crucial to understand the process limitations and resulting fatigue properties of aluminum GMA welds to ensure the structure outlasts the battery chemistry and warranty.

High volume manufactured aluminum welds are controlled with weld acceptance criteria, which are in turn predicted with statistical representation of randomly tested samples. This research investigated the aluminum GMAW process window that consistently produces welds within the industry partner acceptance criteria and resulting microstructure. This research performed component level testing of 2.3 mm AA-6061-T6 mating plates in the tee joint and lap configurations under quasi static and cyclic loading conditions. The quasi-static testing revealed the influence of porosity on the maximum load before rupture of lap shear joints, and the failure location of the joint and lap joint geometries. The cyclic test results showed crack initiation behavior through the thickness of the heat affected zone (HAZ) when observed by digital image correlation (DIC) during testing. Fracture surface analysis revealed crack initiation zones along existing defects and preexisting cracks attached to the root area of the weld for both tee joint and lap joint samples.

Structural stress methods are employed to correlate far field nodal force and moment derived stresses to the observed failure locations in thin sheet aluminum welds, excluding local effects. Load life data is translated to structural stress life data for 2.3 mm thick tested samples, as well as for additional 4mm and 8 mm thick samples provided by the industry partner. Stress life data is segregated based on a bending stress to total stress bending ratio into two distinct structural stress curves. Power law regression is used to calculate a line of best fit through each curve.

Random samples configurations are excluded from a separate regression, which is used to predict the life of the excluded samples within 3 folds (3x) from tested sample life. Mean stress corrections are used to further collapse test data into single membrane and bending curves but applied thickness corrections increased observed scatter amongst test data.