A Real-coded Genetic Algorithm Applied to Optimum Design of a Low Solidity Vaned Diffuser for Diffuser Pinup

A numerical procedure for hydrodynamic redesign of the conventional vaned diffuser into the low solidity vaned diffuser by means of a real-coded genetic algorithm with Boltzmann, Tournament and Roulette Wheel selection is presented. In the first part, an investigation on the relative efficiency of the different real-coded genetic algorithm is carried out on a typical mathematical test function. The real-coded genetic algorithm with Boltzmann selection shows the best optimization performance compared to the Tournament and Roulette Wheel selection. In the second part, an approach to redesign the vaned diffuser profile is introduced. Goal of the optimum design is to search the highest static pressure recovery coefficient and low solidity vaned diffuser. The result of the low solidity vaned diffuser optimum design confirms that the efficiency and optimization performance of the real-coded Boltzmann selection genetic algorithm outperforms the other selection methods. A comparison between the designed low solidity vaned diffuser and original vaned diffuser shows that the diffuser pump with the redesigned low solidity vaned diffuser has the higher static pressure recovery and improved total hydrodynamic performance. In addition, the smaller outlet diameter of designed vaned diffuser tends to a more compact size of diffuser pump compared to the original diffuser pump. The obtained results also demonstrate the real-coded Boltzmann selection genetic algorithm is a promising optimization algorithm for centrifugal pumps design.