中文摘要：

由正交Walsh函数构造Walsh-单信号-复合-输入，对其作用下的计算流体力学响应采用单信号-复合-输入／特征系统实现算法SCI／ERA（Single-Composite-Input／Eigensystem Realization Algorithm）辨识得到离散时间非定常气动力状态空间降阶模型。通过对Isogai机翼剖面气动弹性算例的计算证明该方法具有和非定常计算流体力学方法相当的精度同时模型维数降低2个数量级；在模型构造时间上，SCI／ERA方法比脉冲／ERA方法计算效率提高24％，同时内存占用减小34％；由理论分析可知当耦合结构模态数目增加时，SCI／ERA方法所需的计算开销增幅远小于脉冲／ERA方法；采用频域平衡特征正交分解BPOD（Balanced Proper Orthogonal Decomposition）方法可以准确地从降阶模型中提取出一个低频二次降阶模型，同时保持与原模型相当的精度。二次降阶后模型维数进一步减小88％。

英文摘要：

The Walsh-single-composite-input signals based on orthogonal Walsh function were simultaneously inputted into the CFD（Computational Fluid Dynamics） model,using SCI/ERA（Single-Composite- Input/Eigensystem Realization Algorithm） to identify a discrete-time-domain,unsteady aero-dynamic reduced-order-model（ROM） in state space formulation from CFD responses directly. The Isogai wing stand aeroelastic example was computed, and the results demonstrate that the SCI/ERA method decreases model＇s dimension in 2 orders while still maintains the same accuracy of unsteady CFD codes. Compared to Pulse/ERA method,it improves the computational efficiency by 24% and saves memory space by a4 %. Besides, the SCI/ERA method needs much less additional computational resources than Pulse/ERA method while increasing the coupled structure modes. The frequency-domain BPOD（Balanced Proper Orthogonal Decomposition） method can extract a low frequency second-reduced-order-model（SROM） from ROM while remains the same accuracy. The dimension of ROM further decreases by 88% after second order reduction.