An optimization method based on design of experiments and its application to centrifugal pumps

Many existing optimization methods have inherent limitations which cannot be overcome. For traditional methods, the global optimization values are often outside the searching domain; for recent methods, such as genetic algorithm and simulated annealing, their computation efficiency is very low. In an attempt to find a method that does not have such inherent limitations, we present an optimization method based on Design of Experiments (DOE) and approximation model method. The first step of this optimization method is to set up a DOE-guided database. Once the database is set up, approximation model fitting technique is used to build an approximation model for independent variables and dependent variables. Finally, numerical optimization method is used to obtain the optimal solution for this optimization model. This new optimization method is applied to optimizing centrifugal impellers. Results show that this method is very efficient and accurate.