摘 要：试飞紊流激励环境下，传统频域空间域分解法中奇异值曲线易受紊流干扰，导致伪峰较多无法准确分辨，同时阻尼对于带宽的选择具有高敏感性，人为预设固定带宽无法准确捕捉信号特征，阻尼识别精度受到限制。为了解决上述问题，本文提出基于多元自回归模型改进频域空间域分解法，采用自回归模型谱估计代替原有周期图谱估计，从而获得更平滑的功率谱以及奇异值曲线，有效解决峰值掩盖的问题；同时，采用拟合优度窄带选取法，根据各阶模态局部特征自适应选取窄带，有效提升了阻尼参数的识别精度。通过机翼仿真算例进行验证，结果表明，在20%噪声的情况下，本方法相较于传统频域空间域分解方法，对伪峰进行有效抑制，阻尼识别精度提升了32%，验证了本方法的有效性；飞行试验数据分析表明，该方法显著改善谱线的平滑性，紊流干扰导致的毛刺被有效滤除，且阻尼比识别差异 均小于10%，说明了本方法具有良好的工程适用性与较高的工程应用价值。

Abstract： Under the turbulent excitation environment of flight test, the singular value curve in the traditional frequency-domain spatial-domain decomposition method is easily disturbed by turbulence, resulting in numerous pseudo-peaks that cannot be accurately distinguished. Meanwhile, damping is highly sensitive to bandwidth selection, and the artificially preset fixed bandwidth fails to accurately capture signal characteristics, restricting the accuracy of damping identification. To address the above problems, this paper proposes an improved frequency-domain spatial-domain decomposition method based on the multivariate autoregressive model. The autoregressive spectral estimation is adopted to replace the original periodogram spectral estimation, thereby obtaining smoother power spectra and singular value curves and effectively solving the problem of peak masking. In addition, a goodness-of-fit-based narrowband selection method is utilized to adaptively select narrowbands according to the local characteristics of each mode, which significantly improves the identification accuracy of damping parameters. The proposed method is verified by wing simulation data and flight test data. The simulation results show that under the condition of 20% added noise, the damping identification accuracy of the proposed method is improved by 32% compared with the traditional frequency-domain spatial-domain decomposition method, verifying its effectiveness. Flight test validation indicates that the identification errors of damping ratio obtained by this method are all less than 10%, demonstrating its excellent engineering applicability and practical engineering value.