Torben Pullmann, Martina Schnellenbach-Held, Peer Lubasch
GPCore – a generic framework for genetic programming
Abstract: Complex engineering tasks are rarely unique in their solution. The question is not to determine a single solution – it is more to retrieve an optimal solution. Automated, software-based optimization technologies turned out to be a very promising and applicable solution for the class of non-deterministic optimization problems. In this paper a generic framework for the application of genetic programming will be introduced. The framework is based on the Backus-Naur representation, which can be seen as a meta-programming language. According to specific demands on engineering problems, various detail solutions have been evolved in this context. Finally, the application of the developed framework in load identification will be presented. An example of an evolutionary optimization within the analysis of measured data will be given. By means of two encapsulated optimization routines structural responses re-corded at instrumented bridges are analyzed to determine gross weights, velocities, axle loads as well as axle spacings of passing vehicles. Most significant aspects of the developed framework will be covered within the example: Problem decomposition, problem space definition based on the meta-language, genotype generation, examples of crucial cross-over and mutation operations, fitness evaluation and entire operation of the optimization process.
Keywords: genetic programming, framework, genetic algorithms, evolutionary algorithms, evolutionary optimiza-tion, evolutionary computation
Full text: content.pdf (317,055 bytes) (available to registered users only)