|
DARWIN is an advanced optimization code that utilizes a Genetic
Algorithm (GA). Many design problems are inherently discrete in
nature, since design variables are usually restricted to a discrete
set of values due to manufacturing considerations. Most traditional
optimizers are designed to work with continuous variables only,
and are therefore not well suited to these problems. GAs are one
of the few optimization algorithms that work directly with discrete
design variables. GA's are also excellent all-purpose discrete
optimization algorithms because they can handle non-linear and
noisy search spaces by using objective function information only.
Compared to traditional gradient-based optimizers, genetic optimizers
are more likely to find the overall best (globally optimal) design.
In addition to finding the overall best design, GAs are also capable
of finding many near-optimal designs as well, providing the user
with many options when selecting a final design configuration.
Thus, genetic algorithms have emerged as the optimization procedure
of choice for noisy and discrete design spaces.
In addition to its ability to efficiently handle discrete design
variables, DARWIN also has the ability to handle continuous design
variables. This allows the user to performed optimizations involving
a mixed set of discrete and continuous design variables.
It is important to realize that DARWIN performs design optimization
only and does not have the capability to perform structural analysis.
Appropriate structural analysis codes must be provided by the
user and linked to DARWIN. In order to simplify the process of
connecting external analysis procedures to the optimization code,
data transfer mechanisms and an analysis interface subroutine
template have been designed into DARWIN.
|
