Page 30 - MSAM-2-1

Materials Science in Additive Manufacturing Fast fiber orientation optimization

Table 1. Comparison of the two optimization methods
Comparison criteria Direct optimization (NLPQL) Fast layering optimization
Computation time > 5 h 16 s
Resulting layering −30°/6°/0°/−9°/20°/ −2°/−28°/−20°/−15°/−11°/
−2°/−3°/14°/−4°/−33° −6°/6°/12°/−2°/−2°
Resulting maximal displacement 4.9554 × 10 m 5.4166 × 10 m (+9%)
−5
−5
NLPQL: Non-linear Programming Quadratic Line Search

Figure 10. Force-displacement curves of the six tested parts.

Figure 11. Convergence of the direct optimization model.

found. Then, this sequence was repeated 6 times to fill the
geometry of the wrench.
A simulation on a part with the directly optimized
angle sequence showed that the stiffness of the wrench Figure 12. Displacement fields of the two compared wrenches: direct
is only 9% higher than the wrench obtained with our optimized (top) and quick optimized (bottom).
quick method, with a much higher computation time
(Figure 12 and Table 1). If finding the best solution is Hence, our quick method is a good compromise between
requested, the NLPQL method is suitable but, if the goal performance and design time, as it led to a part that is
is to quickly find an improved solution, the approach only 9% less stiff than the optimal part within a minimal
proposed in this paper is a good compromise between computation time.
computation time and the search for the best solution. It 4. Conclusions
is also important to note that this comparison was possible
because the case was close to a 2D case. A more complex A method to quickly optimize the fiber’s orientations
case would be harder to optimize with a gradient method of a MEX-manufactured continuous fiber-reinforced
due to the number of design parameters it would require. composite was implemented with the finite element
Ansys Workbench software, which was used for this study, method in the Ansys Mechanical environment of
limits, for example, the number of parameters to 20 for programming. The use of stack-based model helped to
automatically computed Design of Experiments. reduce the numerical simulation time, which made the

Volume 2 Issue 1 (2023) 7 https://doi.org/10.36922/msam.49