GA-Suite

Overview

John Ross has developed a unique software program that can automate design searches involving mixed mode EM, scattering and circuit simulations. GA-Suite couples a genetic algorithm (GA) to multiple analysis modules such as the Numerical Electromagnetic Code (NEC), Ohio State Basic Scattering Code, EMSS FEKO and Berkley Spice. This combination of an advanced search method and multiple analysis codes means that its now possible to automate the design process for coupled problems that require mixed mode simulations. GA-SUITE is still in the research phase but has shown great promise in our internal testing. The submodules known as GA-NEC, GA-FEKO, GA-SPICE, etc. are more mature and have been used extensively and successfully by us and by various universities working in design optimization and self-adaptive reconfigurable antennas.

Features

GA-Suite and the submodules currently run under Windows NT/2000/XP. Windows Vista and Windows 7 is not supported. Each submodule consists of several major components as described briefly below:

Template Editors

Establishes genetic encoding template which defines what problem parameters are encoded and how they are encoded.
Encoding can be performed on virtually any thing that is available in the analysis program's standard input file
Encoding options include: none, linear, decade, octave, list (discrete values) and symbolic link

Fitness Editors

Defines the optimization variables or expressions, the type of optimization, analysis type and weighting used in computing fitness functions.
Optimization types include: minimization, maximization, matching on a relative basis, matching on an absolute basis, constraint less than, constraint greater than, constraint inside a range, constraint outside a range.
Fitness can be defined using a combination of different analysis types for each analysis program. In the case of NEC for example it can include input impedance, transmit patterns, receiving patterns, near-zone fields, and RCS. The options are different for Spice and the other analysis modules.

GA-Control

Specifies execuatble analysis programs. Currently supports NEC 2, NEC 4, OSU BSC 4.2, Berkley Spice 3.5, FEKO
Configuration menu allows setting population size, probability of crossovers, probability of mutation, convergence criteria, encoded circuit and fitness parameters
Runtime display of latest generation and population statistics as well as total elapsed time and elapsed time for current generation.
Automatically sorts individuals according to fitness and provides one click linkage to associated GUI for viewing of output files
Generations can be run automatically or stepped manually
Archive of results for each generation.

Screen Shots

GA-Control showing chromosomes, template editor, parameter editor and fitness specifier.

Examples

Additional References:

E.J. Rothwell, C.M. Coleman, J.E. Ross and L.L. Nagy, "Self-Structuring Antennas," IEEE Antennas and Propagation Magazine, June 2002.
Kenneth V. Noren and John E. Ross, "Analog Circuit Design Using Genetic Algorithms," Second Online Symposium for Electronics Engineers, Summer 2001.

Licensing and Pricing

Now that Windows XP and the Visual Stuido 6 are both EOL, we are no longer able to license and support this code. We continue to use it internally and on consulting projects however. Please contact Dr. Ross for additional details on how we can help solve your problems using this extraordinary tool.

Home

Last updated July 23, 2014.