This paper presents a framework of seismic vulnerability assessment for bridges based on Bayesian Networks. This framework incorporates the fragility function, the capacity model and the demand model. The fragility function and the capacity model are constructed using HAZUS model. The demand on the bridge is expressed in term of PGA (Peak ground acceleration) which is calculated by an attenuation function. In order to apply the exact inference algorithm of continuous BNs into this framework, all the continuous variables in the framework are converted into Gaussian distribution and the continuous variables can not have discrete children variables. This framework is applied to ‘twin’ bridges that are correlated with the demand model and the capacity model. First, the reliabilities of the two bridges before the earthquake happens are calculated. Then, when an earthquake happens, the reliability of one bridge are updated based on the observation of the damage occurred on one bridge. Two similar bridges in province of Trentino are used to illustrate the procedures.