Domain Knowledge for Genetic Algorithms

Sushil J. Louis
Dept. of Computer Science
Mackay School of Mines
University of Nevada, Reno, NV 89557
sushil@cs.unr.edu - Fang Zhao
Dept. of Civil and Environmental Eng.
Florida International University
Miami, FL 33199
zhaof@fiu.edu

Abstract:

This paper describes the encoding of domain knowledge for use by a genetic algorithm. We use the domain of system configuration design problems; specifically, the structural design and optimization of trusses to ground our discussion and results. The approach applies evolutionary principles to the optimally directed configuration design of complex structures and incorporates engineering domain knowledge into a genetic algorithm to synthesize the topology, geometry, and component properties of the structure. Preliminary results indicate that genetic algorithms with domain knowledge can generate feasible and useful designs.

Keywords: Genetic Algorithms, Domain Knowledge, Truss Design

• Introduction
• Design and Search
• Genetic Algorithms and Truss Design
• Methodology
• REPRESENTATION
• Results
• Discussion and Conclusions
• References
• About this document ...