|
Abstract: Interactive physical programming is based on physical
programming, a new effective and computationally efficient approach for
multidisciplinary design optimization．It
takes into account the designer’s or the decision maker’s (DM’s)
preferences during the optimization process, and allows for design
exploration at a given Pareto design．The
approximate model of Pareto surface at a given Pareto design is
developed using neural networks for design exploration．At
the given Pareto design, a group of Pareto designs are generated randomly．They’re
brought to the decision maker using a Pareto visualization tool, and are
evaluated with both qualitative and quantitative analysis．A
map from Pareto designs to their corresponding evaluation value is
established using a neural networks model, it illustrates the decision
maker’s local preference at the given Pareto design．Then
genetic algorithms is used in optimization to find the Pareto design
which mostly accords with the decision maker’s local preference．The
obtained Pareto design is used as the aspiration point in compromise
programming, and the final design solution can be obtained．
Key words: Physical programming Interactive physical programming Neural
networks Multiobjective optimization Genetic algorithms Interactive
design Compromise programming
CLC No: H122
国家自然科学基金(59685003)、跨世纪杰出青年学科带头人基金和中国博士后科学基金资助项目.
Received 20010702, received
in revised form 20011120
|