In view of the problems of poor data fusion and strong subjectivity in dam risk assessment methods,this paper proposes a new idea of multi-source information fusion for dam risk assessment by sorting out dam risk failure modes and dam break paths,and using critical paths and critical events.Based on the neural network theory,the dam risk assessment model is constructed.The case analysis shows that the model has high accuracy and reliability,which effectively improves the shortcomings of the traditional risk mechanism models with experts' subjective evaluations,providing a new method for reservoir risk assessment.