基于CFD气动力辨识模型的气动弹性数值计算

摘要

Unsteady aerodynamic reduced order model (ROM)based on computational fluid dynamics (CFD)can either improve computational efficiency or hold the same computational accuracy as CFD.However,the modeling of ROMs based on the identification technology is limited by the loading form and/or frequency spectrum of excitation signals.To overcome the shortcoming and improve the identification efficiency,the random white noise signal was taken as an excitation to model unsteady aerodynamic forces with one-mode-at-a-time loading form.A multiple-input multiple-output (MIMO)model was obtained by the linear superposition of identified single-input multiple-output (SIMO)autoregressive moving average (ARMA)models.With excitation signals of different frequencies and shape on the identified model,the simulation results by ROM and direct CFD computation were compared and it indicates that the same computation accuracy as by CFD can be acquired by ROM.Coupled with structural model,the ROM has been used to predict the flutter boundaries of the wing of AGARD445 .6 .The numerical simulations show that the flutter results predicted by ROM are in general agreement with those by direct computation of unsteady Navier-Stokes equations and wind-tunnel experiments.It verifies that ROM can provide a high efficient method for transonic aeroelastic numerical analysis.%基于系统辨识技术的非定常气动力降阶模型（Reduced-Order Models，ROMs），保留流体力学（Computa-tional Fluid Dynamics，CFD）计算精度的同时能提高计算效率，但对输入信号加载方式或/和信号频谱要求较高。因此采用基于随机白噪声的模态激励信号进行模型辨识。以一次激励一个模态加载方式在时域内建立多输入、输出的非定常气动力降阶模型。该模型由多个辨识所得单输入多输出自回归滑动平均（Autoregressive Moving Average，ARMA）模型线性叠加而成。用不同频率、形状激励信号对辨识模型与直接CFD计算结果比较结果表明，辨识模型与CFD计算精度一致。将该降阶模型转换为状态空间形式后与结构模型耦合用于Agard445．6机翼颤振边界预测。辨识模型计算结果与非定常NS方程计算结果及风洞试验结果比较证实，该方法能用于高效气动弹性分析。