A Simulation-Based Learning Method for Automatic Dynamic Event Tree Model Generation and Analysis

In the paper, a preliminary algorithm tool for automatic dynamic event tree (DET) model generation and analysis is proposed to support dynamic probabilistic risk assessment in nuclear power plants. The DET model is dynamically tailored and adapted to the situation-specific accident scenario development by removing the irrelevant event sequences in accordance with the guided simulated-live events inputs from dynamic dynamics simulator. The algorithm is demonstrated through a small-break LOCA in nuclear power plants. The demonstration results show that the DET algorithm is capable of process-dependent phenomenological events and time-dependent stochastic events modeling. The dynamic risk quantification can also be easily achieved given that the failure probabilities of dynamic events are available.