[en] Fatigue is an important mode of failure in mechanical engineering and accounting for it as
soon as the early stage of design using topology optimization sounds primordial. Structures
undergoing high-cycle fatigue can be described by the stress-based approach and then a
stress-based topology optimization framework, which has received great interest since
almost 20 years because of the innovative designs that can be achieved to answer strength
requirements, can be used. Literature reports many good results for shape optimization
[Mrzyglod & Zielinsky(2006)] whereas in the eld of topology optimization several authors
have shown that considering fatigue in an optimization framework leads to more relevant
solutions where fluctuating loads are involved [Holmberg E.(2015), Collet et al(2016),
Sv ard(2015)].
The good behaviour of the implementation of an advanced fatigue criterion, i.e. the multiaxial
Dang Van criterion [Dang Van et al(1989)] is first investigated in the framework
of a density-based topology optimization problem. The choice of this fatigue criterion is
justifed by its good applicability in automotive or aeronautic industry as well as its relevancy
with respect to experimental results. We present the sensitivity analysis with stress
constraints and present some classical benchmarks to illustrate the behaviour of the optimized
solution. In a second time, the fatigue resistance is introduced in the well-known
microstructural design [Sigmund (2000)] also know as architectured material design which
are now considered in mechanical engineering because of their manufacturability thanks
to additive manufacturing processes. Ensuring the fatigue resistance of the cellular material
will by extension ensure the structural integrity of the overall structure itself. The
optimization is performed by using the MMA optimizer [Svanberg(1987)] whereas the singularity
phenomenon of the stress constraints is circumvented by using the qp-relaxation
[Bruggi(2008)].
Both types of optimization framework are evaluated in term of their numerical performances
and are compared to classical results generated by a regular stress-based topology
optimization. Finally, the results are 3D-printed to assess for their manufacturability.
