Optimization and Computational Fluid Dynamics - Department of ...
Optimization and Computational Fluid Dynamics - Department of ...
Optimization and Computational Fluid Dynamics - Department of ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
18 Gábor Janiga<br />
2.1 Introduction<br />
Designing optimal shapes or optimal configurations for practical engineering<br />
applications has been the subject <strong>of</strong> numerous publications during the<br />
last decade. Generic <strong>and</strong> robust search methods inside the design space, such<br />
as Evolutionary Algorithms (EAs), <strong>of</strong>fer several attractive features for such<br />
problems [38]. The basic idea associated with the EA approach is to search<br />
for optimal solutions using an analogy to the evolution theory. During the<br />
iteration (or “evolution” using EA terminology) procedure, the decision variables<br />
or genes are manipulated using various operators (selection, combination,<br />
crossover or mutation) to create new design populations, i.e., new sets<br />
<strong>of</strong> decision variables. For simpler optimization problems, more classical optimization<br />
methods, like the Simplex approach, are <strong>of</strong>ten better adapted to<br />
find the optimal solution within a small number <strong>of</strong> iterations [74].<br />
The main goal <strong>of</strong> this work is to achieve cost-efficient design optimization<br />
<strong>of</strong> problems involving complex flows with heat transfer or chemical reactions<br />
using <strong>Computational</strong> <strong>Fluid</strong> <strong>Dynamics</strong> (CFD) codes for practical configurations,<br />
while keeping reasonable overall computing times.<br />
Classical optimization techniques, like gradient-based methods are known<br />
for their lack <strong>of</strong> robustness <strong>and</strong> for their tendency to fall into local optima.<br />
Generic <strong>and</strong> robust search methods, such as EAs [16, 34], <strong>of</strong>fer several<br />
attractive features <strong>and</strong> have been used widely for design shape optimization<br />
[1, 48, 55, 62, 64]. They can, in particular, be used for multi-objective<br />
multi-parameter problems. They have been successfully tested in many practical<br />
cases, for example for design shape optimization in the aerospace<br />
[1, 24, 54, 55, 64] <strong>and</strong> automotive industry [61]. The use <strong>of</strong> a fully automatic<br />
EA coupled with CFD for a multi-objective problem still remains limited by<br />
the computing time <strong>and</strong> is up to now far from being a practical tool for all<br />
engineering applications.<br />
2.1.1 Purpose<br />
The purpose <strong>of</strong> this chapter is to illustrate possible methodologies for the<br />
fully automatic optimization <strong>of</strong> various engineering problems involving CFD.<br />
We are not interested here in developing a new algorithm for optimization.<br />
We solely wish to demonstrate that it is possible to reach an optimal configuration<br />
for a case involving coupled fluid flow, heat transfer <strong>and</strong> chemical<br />
reactions, investigated using CFD within a reasonable computing time, for<br />
configurations very close to practical ones.<br />
In Case A, we consider laminar flows because it corresponds to a realistic<br />
engineering problem, for example for low-power systems. A model configuration<br />
is chosen consisting <strong>of</strong> a cross-flow tube bank heat exchanger. The<br />
problem is to optimize the positions <strong>of</strong> the tubes so that the heat exchange is