时域模态参数识别的直接特征系统实现算法

本文提出了一种新的结构模态参数识别算法,称为直接特征系统实现算法(DERA)。直接采用时域多输入多输出数据构成线性时不变离散系统的最小实现,并利用系统矩阵的特征值和特征向量求取模态参数。输入输出数据的直接应用避免了利用脉冲响应函数作原始数据的ERA、PRCE等算法的一些局限。奇异值分解技术的引入使系统的定阶和实现更方便,并具有了更良好的数值稳定性。为验证算法,构造了一个五自由度的弹簧质量系统进行数值模拟。良好的数值结果证明了本算法的有效性,并表明本算法对密集模态仍有较高的识别精度。

A DIRECT EIGENSYSTEM REALIZATION ALGORITHM FOR MODAL PARAMETER IDENTIFICATION IN TIME DOMAIN

A new method called direct eigensystem realization algorithm (DERA) is developed for structural modal parameter identification. A direct procedure is presented for obtaining the minimum realization of a multiple-input multiple output time-invariant discrete system from observations of the input and output in time domain, and then modal parameters are obtained from the eigenvalues and eigenvectors of the system transition matrix. Because of the use of direct input-output observations, some weaknesses of the ERA and the PRCE methods which use impulse response functions as initial data have been avoided. By using the singular-value decomposition technique, a more convenient procedure is derived to determine the minimum order and realization of the system. An example is shown to illustrate the algorithm, in which simulation data from a spring, mass and damping system of five degrees of freedom are utilized. Effectiveness of the algorithm is proved by good numerical results, and high accuracy is achieved for close-spaced frequency modes.