识别传递函数参数的一种广义最小二乘法

线性动力学系统的频响特性的量测误差具有相关性,用最小二乘估计求得的传递函数系数有偏.为此,本文用广义最小二乘法导出识别传递函数参数的迭代算法,并用数值模拟法产生强噪声污染的频响函数数据,用它作识别计算.结果表明本文建议的算法比最小二乘法的精度有明显提高,计算量也不大.

A Generalized Least Squares Algorithm for Identifying Transfer Function Parameter

The errot of measured frequency response data of linear dynamic systems is coherent, so the transfer function parameters estimated by the least squares algorithm will produce error. For this reason, a generalized least squares technique is adopted in this paper to derive an iterative algorithm for identifying the transfer function parameters. Procedure of calculation is listed. Identification results of a vibration system indicate that the generalized least squares algorithm proposed in this paper is superior to the least squares algorithm both in estimation accuracy and computation time.