Comparison of Methods for Developing and Using Dynamic Reduced Models for Design Optimization

This thesis compares different methods used in developing and using dynamic reduced models for design optimization. Several methods have been used for forming functional approximation models with which the GA search is performed more efficiently. The informed operators approach for search speedup is discussed. Another hypothesis-instantiation approach in using reduced models has been tested and compared with the informed operator approach. Chapter one first gives some background knowledge. Chapter two describes three different methods for forming reduced models to speed up genetic-algorithm-based design optimization. Chapter three describes two methods for using reduced models to speed up genetic-algorithmbased design optimization. Chapter four gives the experimental results and discussion. Finally, chapter five serves as a summary of the whole thesis. Index words: GADO, Quickprop, RBF, Quadratic least squares, Genetic engineering, Informed operators, Reduced models Comparison of Methods for Developing and Using Dynamic Reduced Models for Design Optimization