Abstract: Effect of Human factors on system safety is increasingly serious,which is often ignored in traditional probabilistic safety assessment methods however.A new probabilistic safety assessment model based on object-oriented Bayesian networks is proposed in this paper.Human factors are integrated into the existed event sequence diagrams.Then the classes of the object-oriented Bayesian networks are constructed which are converted to latent Bayesian networks for inference.Finally,the inference results are integrated into event sequence diagrams for probabilistic safety assessment.The new method is applied to the accident of loss of coolant in a nuclear power plant.The results show that the model is not only applicable to real-time situation assessment,but also applicable to situation assessment based certain amount of information.The modeling complexity is kept down and the new method is appropriate to large complex systems due to the thoughts of object-oriented.