1986 年 52 巻 484 号 p. 3154-3160
In the Bayesian parameter estimation method, the uncertainties of both experimental responses and theoretical model parameters are taken account of to identify the finite element model from experimental modal test results such as the natural frequencies and mode shapes of a structure. This paper presents a modified theory of the Bayesian parameter estimation method. In this theory, the Bayesian parameter estimation method is extended by normalizing the error function, and a constant that weights the parameter valuesto the test data can be reasonably determined; that is, a parameter identification that weights the experiment, or that weights the theory can be carried out quantitatively and easily. And constants that denote the various uncertainties of each parameter and response reflect the identified values. The effects of relative parameter and measurement uncertaities on the identified values are illustrated by two five-degrees-of-freedom models.