An uncertainty model updating method, which is targeted at frequency response function and using an approximate likelihood function, is proposed under the Bayesian framework. Compared with modal parameters, the frequency response function contains more structural information which offers advantages in finite element model updating of structural dynamics. However, the current uncertainty model updating methods rarely work well with a target of the frequency response function. In light of this problem, the uncertainty model updating theory under the Bayesian framework is introduced. An approximate likelihood function suitable based on the approximate Bayesian computational method is presented for the uncertainty model updating with e tfrequency response function. The proposed uncertainty model updating method is verified on a vibration system with three-degree-of-freedom and an unsymmetrical H-shaped beam. The DREAM algorithm is introduced to identify the parameters. The results show that the upper and lower bounds of updated parameters as well as the frequency response function of the updated model vary little from the desired values. The uncertainties are well updated under noise, which verifies the effectiveness of the proposed method.