Using Evolutionary Methods with a Heterogeneous Genotype Representation for Design Optimization of a Tubular Steel Trellis Motorbike-Frame U. M. Fasel, O. K¨ onig ∗ , M. Wintermantel and P. Ermanni Centre of Structure Technologies, Institute of Mechanical Systems, Swiss Federal Institute of Technology, Zurich, Switzerland Abstract We present an optimization approach for a tubular steel trellis motorbike-frame based on evolutionary methods. The optimization objective is to minimize the mass of the frame subject to a given torsional stiffness and a strength constraint occurring during braking of the motorbike. Further we take into account different constraints on available tube sizes, what leads to continuous, discrete or even mixed optimiza- tion variables. For stiffness and strength calculations we use our own Finite Element code FELyX (Finite Element Library experiment). As optimization engine we use Evolutionary Algorithms within the framework of the Evolving Objects (EO) li- brary. There we implemented a new heterogeneous genotype that allows to combine different parameter types, represented by according genes into a single genotype. Further we discuss the impact of smart parameterization schemes onto the con- vergence of such methods. Finally we show how they can be linked naturally and efficiently to modern Evolutionary Methods that distinguish between phenotype and genotype representation. Key words: Evolutionary Algorithms, heterogeneous genotype, mixed parameter optimization, discrete parameter optimization, motorcycle main frame ∗ Corresponding author. Address: Leonhardstr. 27, LEO C4, CH-8092 Zurich Email addresses: okoenig@imes.mavt.ethz.ch (O. K¨ onig), wintermantel@imes.mavt.ethz.ch (M. Wintermantel), ermanni@imes.mavt.ethz.ch (P. Ermanni). Preprint submitted to Elsevier Science 7 October 2003