Optimization and Computational Fluid Dynamics - Department of ...
Optimization and Computational Fluid Dynamics - Department of ...
Optimization and Computational Fluid Dynamics - Department of ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
238 Marco Manzan, Enrico Nobile, Stefano Pieri <strong>and</strong> Francesco Pinto<br />
2. Evolutionary Algorithms (EA). They are heuristic methods that use some<br />
mechanisms inspired by biological evolution: reproduction, mutation, recombination,<br />
natural selection <strong>and</strong> survival <strong>of</strong> the fittest. Their most important<br />
feature is the applicability to almost all types <strong>of</strong> problems, because<br />
they do not make any assumption about the system under study as<br />
classical techniques do. They can be used when relation f is a completely<br />
unknown function. In their multi-objective version, Multi-Objective Evolutionary<br />
Algorithms (MOEA), being part <strong>of</strong> the recently cited research area<br />
known as evolutionary multi-objective optimization (EMOO), are shown<br />
to be capable <strong>of</strong> dealing with truly multi-objective optimizations [7].<br />
Differential calculus is the first example <strong>of</strong> traditional programming techniques,<br />
but there is a widely developed literature [45] on the subject. Linear<br />
programming, quadratic programming <strong>and</strong> nonlinear programming are just<br />
some <strong>of</strong> the examples. To perform the optimization processes described in<br />
this chapter, evolutionary techniques have been used that are available in<br />
the modeFRONTIER c○ optimization program [14]. modeFRONTIER c○ ,used<br />
throughout, is a state-<strong>of</strong>-the-art optimization package that includes most instruments<br />
relevant to data analysis <strong>and</strong> single- <strong>and</strong> multi-objective optimizations.<br />
8.6.1 Design <strong>of</strong> Experiment<br />
Heuristic evolutionary techniques do not make any assumption on the relation<br />
between objectives <strong>and</strong> design variables, thus providing an analogy with<br />
experimental dataset analysis. A good initial sampling, which allows an initial<br />
guess on the relations between inputs <strong>and</strong> outputs, is <strong>of</strong> great relevance<br />
in reducing optimization effort <strong>and</strong> improving results [40].<br />
Design Of Experiments (DOE) is a methodology applicable to the design <strong>of</strong><br />
all information-gathering activities where variation <strong>of</strong> decisional parameters<br />
(design variables) is present. It is a technique aimed at gaining the most<br />
possible knowledge within a given dataset. The first statistician to consider<br />
a formal mathematical methodology for the design <strong>of</strong> experiments was Sir<br />
Ronald A. Fisher, in 1920.<br />
Before the advent <strong>of</strong> DOE methodology, the traditional approach was the<br />
so-called One Factor At a Time (OFAT). Each factor (design variable) which<br />
would influence the system used to be moved within its interval, while keeping<br />
the others constant. In as much as it would require a usually large number<br />
<strong>of</strong> evaluations, such a process could be quite time consuming. By contrast,<br />
Fischer’s approach was to consider all variables simultaneously, varying more<br />
than one at a time, so as to obtain the most relevant information with minimum<br />
effort. It is a powerful tool for designing <strong>and</strong> analyzing data: it eliminates<br />
redundant observations, thus reducing time <strong>and</strong> resources in experi-