振动系统时变参数识别分析

本文利用变物理参数微分方程模型,讨论了刚度与阻尼随时间变化时,物理参数与模态参数的识别问题,比较了三种在线识别技术.数字模拟结果表明,采用变物理参数微分方程模型可比采用差分方程模型极大的提高最小二乘法识别阻尼比的精度.当加入系统的噪音较大时,有些改进识别方法失效,但最小二乘法仍能给出较好的估计.

Identification Analysis of the Time Varying Parameters of Vibration Systems

The identification of physical and modal parameters of damping and stiffness varying systems was examined by utilizing the differential equation model with varying physical parameters. Compared with the difference equation model, the differential equation model with varying physical parameters can considerably improve the damping identification accuracy by the Least Squares technique. On-line versions of three time domain system identification algorithms- Least Squares, Double Least Squares and Instrumental Variables, were implemented in order to determine the accuracy with which changes in the physical and modal parameters can be followed well. Data generated from various simulated time varying parameter systems, corrupted with both input and measurement noise, were used to examine the available algorithms. Comparisons were made among results obtained from the three different identification techniques. It was found-that when the noise level was increased, the best estimates on damping ratios were often obtained by the LS technique but not by the DLS and the IV techniques. When the IV approach failed, the LS approach can still give acceptable estimates.