Abstract: When time-domain TPA methods used for tackling vibrations and noise problems in complex systems during transient conditions, which failed to overcome the ill-conditioned problems of the frequency response functions matrix near natural frequencies, and showed low accuracy in extracting the required time-domain information from the existing frequency-domain data. Therefore, a new time-domain TPA method was developed based on AKF. This method started with estimating time-domain operational loads using AKF supplemented by GA, and then identified the impulse response functions using LS algorithm. Finally, time-domain contributions for each transfer path was computed by linearly convolving time-domain operational loads with respective impulse response functions. Case study demonstrates that the load-identification errors of AKF applied in the proposed method are smaller than that of traditional deconvolution filters. Additionally, the errors of impulse response functions identified by LS algorithm are smaller than those obtained by direct inverse fast Fourier transform or creating finite impulse response filters from frequency response functions. Furthermore, the proposed method achieves small errors even in complex structures.

Key words: time-domain transfer path analysis（TPA） method, augmented Kalman filter(AKF), genetic algorithm（GA）, least square（LS） algorithm

摘要： 时域传递路径分析方法用于解决瞬态工况下复杂系统振动噪声问题不仅未能解决自然频率附近频响函数矩阵的病态问题，而且利用现有频域信息转换提取得到的所需时域信息精度较低，因此提出一种基于增广卡尔曼滤波器的时域传递路径分析方法。该方法采用增广卡尔曼滤波器辅以遗传算法估计时域工况载荷，通过最小二乘算法辨识单位脉冲响应函数，将时域工况载荷和对应的单位脉冲响应函数进行线性卷积以计算各传递路径的时域贡献量。算例表明，所提方法采用的增广卡尔曼滤波器载荷识别误差小于传统方法的去卷积滤波器所识别载荷的误差，最小二乘算法辨识的单位脉冲响应函数误差小于对频响函数直接进行快速逆傅里叶变换或者构造有限单位脉冲响应滤波器的误差，且所提方法在复杂结构上也同样具有较小的误差。

关键词: 时域传递路径分析方法, 增广卡尔曼滤波器, 遗传算法, 最小二乘算法