Reconstruction of Core Overheating Damage Fraction Based on Neural Network Method

Core damage assessment is of great importance to the emergency response of nuclear power plants. In this paper, the neural network method is introduced into the core damage assessment process. The hydrogen concentration, together with the temperature and pressure in the containment, are taken as the input parameters of the model. With the simulated result of MAAP codes as the sample data, a neural network model is developed to reconstruct the core overheating damage fraction. According to the calculation of the neural network model, the deviations of the reconstructed results are quite small compared with the simulation results, and one of the typical errors is 1.76%. It can be concluded that the model based on neural network method satisfies the analysis accuracy requirements and can be used as a diverse analytical tool in the core damage assessment of nuclear power plant.