Abstract: A model based on discrete fuzzy dynamic Bayesian network is derived for radiant threat rank assessment due to the uncertainty of the evidence.The continuous observation values are fuzzily classified by the membership functions，which are determined by fuzzy sets.The results are inputted into the model.The interface algorithm is modified in order to process evidences with multi-states.It is shown that the presented model can predict accurately changes of the situation in varying dynamical environment and has a good performance in terms of effective noise filtering from observations.