Optimization and Computational Fluid Dynamics - Department of ...

258 Marco Manzan, Enrico Nobile, Stefano Pieri and Francesco Pinto
10
8
f/f
0 6
4
2
Linear−piecewise wall profile
NURBS wall profile
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
Nu/Nu
0
Fig. 8.20 Pareto front comparison between linear piecewise and NURBS-based channels
and longitudinal steady vortices on the heat transfer rate is again clear. For
illustrative purposes, the secondary flow pattern for two channels obtained
with a 20 ◦ and 40 ◦ extrusion respectively, are depicted in Fig. 8.22.
Finally, a MOGA optimization has been started on extruded linear piecewise
channels for a short number of generations. The Pareto front of this
optimization is compared in Fig. 8.23 with the 2D linear piecewise and 2D
NURBS fronts. Though the 3D linear piecewise front is rather sparse because
of a small number of individuals processed, the heat transfer augmentation
due to secondary motions is clearly visible.
8.8.5 CC Module
Again with MOGA-II, an original design was used for the DOE when optimization
was first attempted. The design had sinusoidal wall profile (design
0), characterized by a corrugation angle θ =60 ◦ , and 35 designs have
been chosen with Sobol method. The value of the angle θ has been chosen,
following Stasiek et al. [52], as a good compromise in terms of heat transfer
rate and friction factor. The six design variables were, with reference to
Fig. 8.6, P1, P2, P3, P4, W and θ. The three objectives to be minimized
were the pressure gradient β, the temperature difference between the two
fluids ∆T and the heat transfer surface area. The first objective has been
selected to reduce the pressure drop across the regenerator. The second one
drives the heat transfer performance of the regenerator. For instance, based
on inspection of Eq. (8.35), a lower ∆T leads to a higher mean Nusselt number.
The last objective has been selected to reduce the overall cost of the