Improving the performance of cluster oriented genetic algorithms (COGAs)

This paper presents an empirical investigation of the integration of the adaptive filter (developed for use with COGAs) with a series of evolutionary search algorithms. The relative merits of each technique are compared with variable mutation COGA (vmCOGA) (Parmee, 1996) using a suite of two-dimensional test functions and a series of performance measures. The generic capabilities of COGA are demonstrated by illustrating its ability to rapidly decompose multidimensional search spaces described by continuous and discrete parameter, real-world design models into regions of high performance. These models relate to conceptual airframe design and compressor blade cooling within a gas turbine engine. This investigation further supports the use of the COGA strategy as a conceptual, engineering design support tool