频域响应函数估计的非参数辨识法

A nonparametric identification method for frequency response funtion estimation

针对振动工程领域中频响函数的辨识估计,本文深入研究频响函数估计的非参数辨识法。在离散傅里叶变换中,考虑初始和终端状态带来的暂态泄露项和观测噪声谱项对频响函数估计的影响。为得到准确的频响函数估计值,联合频响函数、初始-终端状态和脉冲响应系数待辨识参数矢量,将频响函数估计问题转化为一个线性最小二乘优化问题。针对此线性最小二乘优化问题的特殊形式,提出一种可分离的求解过程。最后用仿真算例验证本文辨识方法的有效性。

Here,a nonparametric identification method for frequency response function estimation was proposed.It could be applied in vibration engineering field.In discrete Fourier transformation,the effect of the initial and end conditions and an observed noise spectrum on frequency response function estimation was studied.To obtain the more accurate frequency response function estimation values,the frequency response function,the initial and end conditions and impulse response coefficients were taken as a parameter vector to be identified.The problem of frequency response function estimation was converted into a linear least square optimal problem.Then,a new separable algorithm was proposed to solve this linear least square optimal problem.Finally,the simulation example results verified the effectiveness of the proposed method.