Abstract: Dam break risks exist in all stages of the tailings pond operation period, but traditional risk predictions cannot reflect the real-time dynamic changes of tailings pond dam-break risks.To solve this problem, this paper builds an evaluation index system including dynamic indicators and static indicators is constructed, and index weights and time weights.A dynamic weighted Bayesian network tailings reservoir dam break risk assessment model with embedded time weights and indicator weights is established.Then improves the method of obtaining conditional probability in the case of insufficient parameter learning data through interval number goodness sorting.Finally, the dynamic Bayesian network of tailing pond dam break risk under different rainfall durations and different rainfall effects is carried out probabilistic inference.The results show that when the rainfall for 3 days is class A (high risk) and class B (higher risk), the risk probability of dam break of tailings pond caused by rainfall increases from 19.0% and 9.1% to 34.9% and 20.4% respectively.Research shows that the introduction of the timeweighted evaluation model reflects the dynamic development process of the research object, dynamic assessment is more in line with objective facts.