基于CFD系统辨识的气弹分析及GPU并行算法初探
黄灿，赵永辉
摘要:通过求解Euler方程获得运动翼段的非定常流场,并用CUDA语言对流场求解器进行GPU并行计算.使用ARMA（auto-regressive-moving-average）模型对非定常气动力进行辨识,由系统辨识模型得到的结果与全阶CFD计算结果十分吻合.基于降阶气动模型与结构的耦合,计算了具有S型颤振边界的气动弹性标准算例-Isogai Wing的跨音速颤振.本文给出的方法可以在保证气动弹性计算精度的前提下大幅提高计算效率.
关键词：EULER方程；ARMA模型；系统辨识；气动弹性；并行计算
Identification of cfd-based aeroealstic analysis and gpu parallel computing
Can Huang, Yonghui Zhao
Abstract：The unsteady flow field around a moving airfoil section was obtained by solving the Euler equation. The CUDA programming language was used to provide a parallel computing capacity on GPU. The ARMA（autoregressive-moving-average） model was used to identify the unsteady aerodynamic forces, and simulation results using the identified aerodynamic model of the system identification agree well with those using the full ordered CFD model. Furthermore, the transonic flutter of the Isogai Wing with an Stype flutter boundary, which is the standard example in aero-elasticity, was calculated. The approach presented can greatly improve the efficiency of computing without loss of accuracy.
Keyword：Euler equation; ARMA model; system identification; aero-elasticity; parallel computing
原文链接：http://www.cnki.com.cn/Article/CJFDTotal-DLXK201502002.htm
