NOVEL ALGORITHM FOR GEOMETRICAL OPTIMIZATION OF FLOW CHANNELS

Autor/autori: FLORIN GABRIEL FLOREAN, IONUČš PORUMBEL


Abstract: The paper proposes a novel algorithm developed for the geometrical optimization of various flow channels. Mainly, such optimization aims at decreasing the total pressure losses in the channel, but the present algorithm may be used with trivial modification for any optimization criterion. The concept of the proposed algorithm is using an approach consisting in a combination of Factorial Design of Experiments (FDoE) and Artificial Neural Networks (ANN) to evaluate the effect of the flow channel geometry on the aerodynamic parameter defining the objective of the optimization. The training and testing sets for the development of the ANN will be obtained through simple CFD Reynolds Averaged Numerical Simulations (RANS) of the flow in geometries selected by means of the previously performed FFDoE. These numerical simulations will be carried out as "experiments" able to provide values for the aerodynamic objective. Finally, the aerodynamic objective of choice will be optimized by means of a Genetic Algorithm (GA). The GA results may be, in the end, validated through detailed, time accurate, Large Eddy Simulations (LES).

Keywords: aerodynamic optimization, artificial neural networks, design of experiments

 

DOWNLOAD PDF