Development of an improved reflood model for RELAP5 and SPACE

Abstract

During the reflood phase of a Large Break Loss-Of-Coolant Accident (LBLOCA), almost all kinds of two-phase configuration expected in a vertical rod bundle geometry from single phase vapor/liquid flow to inverted flow regimes may exist in the core, sometimes all at once. Thus it has been regarded as extremely difficult to predict all the complicated reflood phenomena with a high accuracy. RELAP5 is a computer code developed by the United States Nuclear Regulatory Commission (USNRC) and it has been regarded as one of the best estimate codes having enough features to predict realistically the two-phase flows in Pressurized Water Reactors (PWRs). Due to its many strengths such as fast running speed, a good numerical stability, and/or relatively good predictive capability for various two-phase flows, the code has been applied to licensing LBLOCA analyses for all types of PWRs in this country. However, its reflood model, the models and correlations used to treat the complicated two-phase phenomena in the core during reflood, has not been assessed in detail since its offficial release in 2001. In addition, it was revealed from our assessment against reflood separate effect tests that the code or its reflood model can predict relatively well the peak clad temperatures (PCTs) but it can not do the PCT times, rod quench times, or the transient behavior of vapor temperatures. Therefore in this study, the models and correlations having significant effects on the predictions for reflood phenomena were investigated intensively by reviewing not only code manuals but also code source files and comparing them to the models and correlations of other best estimate codes such as COBRA-TF, TRAC-M, or TRACE. As a result, the following potential problems in the current reflood model were identified. First, the code uses the vapor temperature when it determines whether the flow regime of a hydraulic cell is a pre-Critical Heat Flux (pre-CHF) flow or a post-dryout flow. Basically thi...