Global Optimization of a Turbine Design via Neural Networks and an Evolutionary Algorithm

This work discusses an effective approach to find the optimal solution for constrained engineering design problems. Specifically, the computational platform herein implemented exploits a neural network and a differential evolution algorithm, and it leverages on a parametric finite element modelling for the fully automation of the design process. The presented approach is applied to the design of the rear flange of a low-pressure turbine casing for an aircraft engine, whose shape is optimized in order to reduce the manufacturing cost while preserving the overall integrity through the fulfilment of stress-based constraints.