D.H. Lammlein PhD Dissertation - Vanderbilt University

The results presented here show that aluminum alloy hemispheres can be reliably
joined by FSW. Additionally, the control techniques used in this experiment show that
system eccentricities can be accounted for via multiple methods. The relative height of
the material surface can be taken at various points around the circumference of the sphere
and a preplanned path can be taken by the tool to maintain the appropriate penetration
depth. The unique geometry of the conical tool can be used advantageously in torque
based feedback control of weld penetration depth, as measured weld toque proved to be a
strong and reliable indicator of weld depth. Additionally, it was shown that for cases
where the interior of the vessel is inaccessible and internal support is not possible, partial
penetration welds of similar proportional strength can be achieved without compromising
the material interior surface. A lack of material distortion at the joint base is required for
external compression applications.
The conical tool produced, 50% penetration welds in an unsupported
configuration which ranged from 20-26% parent material strength in a practical
comparison using their apparent stress. The cupped shoulder tool produced full
penetration welds with interior support which ranged from 54-68% of the parent material
strength comparing apparent stress. The maximum strength could be improved by testing
a wider range of parameters, as welds made with both setups were stronger at higher
travel and rotational settings. Additionally, a penetration ligament of 0.014” was used in
the full-penetration portion of the experiment. A more aggressive penetration ligament
(i.e. one closer to 0.006”) would likely increase weld strength by eliminating any joint
remnant at the weld root. Partial penetration welds continued to increase in strength with
depth outside the matrix depth of 0.10”, however with the introduction of some distortion
at the weld root. In this experiment it was assumed that this distortion was unacceptable
and the experiment was done at a depth where no distortion occurred. Future work could
explore the depth to which this relationship continues to hold should a suitable
application be identified.
Additionally, modifications to the tooling used in the experiment would likely
result in strength increases as only a single tool geometry was considered for each case.
A more complicated probe geometry and scrolled shoulder design could improve test
strengths for the supported welds. Adjustments to the inclusive angle of the conical tool
147