Soundappan Ramanathan, "Knowledge Incorporation into Evolutionary Algorithms to Speed up Aerodynamic Design Optimization", 2009.

This work deals with making the search process with Evolution Strat- egy with Covariance Matrix Adaptation (CMA-ES) more efficient. Methods are proposed to initialise the optimisation with predefined knowledge instead using a random initialisation in the context of aero- dynamic design optimisation. In aerodynamics, the search for better shape is always in research. Conventional numerical optimisation pro- cesses results in huge amount of data sets as a result of each optimi- sation. But most often only the best designs are taken into considera- tion. Graening, (2) developed a framework to extract the meaningful information about the shape from those data sets. This approach extends by means of developing a common framework to incorporate the knowledge extracted from the designs, so that the search process can be initialised in order to achieve the new outperforming designs. This thesis is split into two parts: (i) Development of frame work for quantifying interaction effects between the design and the perfor- mance. (ii) Developing an optimal strategy to initialise the search process. Firstly, for quantifying interaction effects linear as well as non-linear interaction effects are analysed. Statistical techniques such as multiple regression and information theory are applied to quantify the parameter interaction effects. Second part deals with developing a common framework to initialise the optimisation by means of in- corporating the knowledge into the covariance matrix. The developed framework is applied to 2D Gas turbine blade optimisation for the validation of the knowledge incorporation technique.