Determination of Erroneous Velocity Vectors by Co-operative Co-evolutionary Genetic Algorithms

Abstract

The effects of incorporating co-operative co-evolutionary strategy into a genetic algorithm (GA) for the identification of erroneous velocity vectors in particle image velocimetry (PIV) are studied. The search objective is to eliminate vectors that are dissimilar to their adjacent neighbors. A simulated cavity flow, which is modified to contain 20% erroneous vectors, is used as the case study. The co-operative co-evolutionary strategy is found to decisively improve the search effectiveness. When the effect of species size and arrangement are considered, the search rate improves with smaller species, reflecting the weak linkage between species due to the locality nature of the objective function. Best results are obtained with the 25-bit species under square arrangement. It is also observed that the current vector similarity calculation as the objective function needs further assessments for the erroneous vector detection of complex velocity flows with high error rates