识别传递函数参数的一种广义最小二乘法

A Generalized Least Squares Algorithm for Identifying Transfer Function Parameter

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.

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.