以基因演算法與平行運算進行二維翼型優化

Airfoil Shape Optimization Using Genetic Algorithms and Parallel CFD

潘大知(Dartzi Pan) ； 邱永裕(Yung-Yu Chiu) ； 趙茂吉(Maw-Jyi Chao)

《中國航空太空學會學刊》 34卷4期 (2002/12) Pp. 347-352

https://doi.org/10.6124/TAASRC.200212_34(4).07

摘要

本文之目的在應用平行運算及基因演算法(Genetic Algorithms, GA)於機翼翼型最佳化工作上。基因演算法是近年來興起而極受注意的最佳化方法，其原理是模仿生物基因在「物競天擇，適者生存」之達爾文原理作用下，藉由不斷的世代交替來去蕪存菁、汰弱存強，進而達到改善物種的目的。基因演算法之優點是適用性廣，可處理一般以梯度爲依據的方法(Gradient Based Methods)所不能處理之大型複雜系統最佳化問題，基因演算法的另一個優點是其本質上即是可平行處理的，因此可充分利用目前「電腦群組」(PC Cluster)之發展，進行以MPI (Message Passing Interface)爲工具之分散式平行運算(Distributed Paraller Computing)。本研究將整合(1)實數型基因演算法(Real-Valued GA)，(2)計算流體力學二維Navier-Stokes流場解子，(3)格點產生法，(4)平行處理環境，完整建立二維機翼外型優化之能力，並對民航用二維機翼LS-0417進行改良嘗試。

關鍵字

基因演算法 ； MPI平行運算 ； 二維機翼翼型最佳化

並列摘要

This paper applies a real-valued genetic algorithm (GA) and computational fluid dynamics (CFD) to the task of airfoil shape optimization. The airfoil profiles are modeled by a combination of Bezier curves, whose control points are used as real-valued genes. The fitness function for maximization is an appropriate combination of the lift and drag coefficients. The required lift and drag coefficients are obtained by parallel incompressible Navier-Stokes computations on a 8-node P-II PC cluster using MPI calls. A master node is responsible for executing the GA procedure that generates new chromosomes (airfoil shapes), distributing chromosomes evenly to computational nodes in the cluster and collecting the computed lift and drag coefficients for the GA procedure. On each individual computational node, a grid generator and a flow solver is called to calculate the lift and drag coefficient for each chromosome received. Optimization attempt on the performance of a general aviation airfoil LS0417 is tried as an example case.

並列關鍵字

Genetic Algorithms ； Airfoil Shape Optimization ； Parallel Computation