The authors present a method for radial flow impeller design which pays attention to the set-up of optimal design strategies, the latter based upon the employment of a genetic algorithm which controls a Quasi-3D flow simulation of a radial flow impeller, with integral boundary layer calculation.
The design procedure develops through a number of steps, which are characterized by a progressive reduction in the search interval of each independent variable. In addition, a more accurate definition of the blade geometry and a more reliable flow analysis is obtained through a mesh refinement in the flow model, when the whole optimization process is close to convergence.
The results demonstrate the possibility of achieving an optimal choice of the impeller geometrical parameters by approaching, in a first phase, the design data, and then by proceeding with a number of refinement steps, as outlined above. In this way. the accuracy of the whole procedure is strongly increased, so that the typical effectiveness of the genetic algorithms in the preliminary selection steps is combined with a detailed investigation of the impeller blade features in the final steps of the optimized design. The effectiveness becomes, therefore, comparable with that of classical gradient based methods of optimization. At the same time, the genetic algorithm is capable of handling discontinuities in the objective function, thus overcoming the typical limits of numerical methods.
In order to emphasize the effectiveness of the proposed procedure, the results refer to a wide variety of design data. Both the impeller configurations and the flow distributions found demonstrate that the final solution is in agreement with the design target and constraints.