This paper presents a new multi-reference modal parameter identification technique based on a weighted least squares principle. This principle is derived from the maximum likelihood method which had been developed in the field of statistics of inference. The proposed method performs the least squares estimation by using frequency response functions (FRFs) as input data and the reciprocal of variance of the FRF as a weighting function. The procedure finding the modal parameters that minimize the weighted squared errors becomes the non-linear least squares problem, which can be solved by the Gauss-Newton method. According to the method, it needs an iterative calculation. An idea is introduced into the identification algorithm, which drastically reduces the computer processing time and computer memory requirement. Validity of the proposed method is examined through the application to the experimentally acquired FRFs. Compared with former curve-fitting methods such as Poly-reference technique, it is proved that the reliability of the identified modal parameters is greatly improved.
展开▼
