AN INVERSE DESIGN METHOD OF DIFFUSER BLADES BY GENETIC ALGORITHMS

A genetic algorithm incorporating a neural network technique is propos ed to search for a turbomachinery diffuser blade profile that produces a given velocity distribution on its surface. Such a new inverse desi gn method works through minimizing the error between the surface veloc ity distribution of candidate blades and the target velocity distribut ion. For ease of employing the genetic algorithm, the blade profiles t o be searched are parameterized by Bezier curves. To fix the surface v elocity distribution of a candidate blade, a special type of back prop agation (BP) neural network is implemented. The proposed approach is i llustrated by a diffuser having two-dimensional blades with constant h eight and thickness. The simulations show that the new method is not o nly feasible but also reliable and efficient.