In a previous study [1], several multiaxial
fatigue models were investigated and compared based on their ability to predict
the fatigue limit under multiaxial loading conditions. The widely used
historical models such as Findley [2] and
Dang Van [3] were compared to several
recently developed models. The methods were investigated for the purpose of
assessing their potential use in automotive design. In the current study, the
same multiaxial fatigue models were assessed based on their ability to perform
life prediction under high cycle multiaxial loading. The experimental data used
for the assessment of the seven different multiaxial models was taken from
literature. Five of the models, Findley, McDiarmid, Susmel-Lazzarin, MZSL, and
scaled normal stress were critical plane approaches. The other two models were
the LTJ approach and the prismatic hull method, both of which are based on the
von Mises criteria. The scaled normal stress approach was the only tensile
failure mode model investigated with all other models being shear failure mode.
Each stress-based model was used to predict the fatigue life and compare to the
experimental results obtained from literature. When selecting data from
literature, only high cycle multiaxial fatigue data was used. Experimental data
from steel, stainless steel, and aluminum materials were investigated. Most of
the materials exhibited shear failure mode, but some materials had mixed mode
cracking. The models were judged based on their ability to predict the
multiaxial fatigue life within factors of 3 and 5. The LTJ model had the best
overall agreement with the experimental data, with 82% of life predictions
within a factor of 3 and 96% within a factor of 5. This was due in part to its
material parameter, which is derived from multiaxial test data. The LTJ model
was one of two models that required multiaxial test data to generate a model
parameter. All other models relied on either monotonic or uniaxial fatigue data,
which is more readily available and much easier to generate. The Susmel-Lazzarin
approach had the second-best overall agreement with 60% and 79% of predictions
within factors of 3 and 5, respectively. The scaled normal stress approach,
prismatic hull approach, and MZLS approach had life predications that were just
slightly less accurate than the Susmel-Lazzarin method. The McDiarmid and
Findley models had the worst correlation with the experimental data investigated
in this study.