(A) Study on DNBR evaluation method by grouping of axial power distribution

Abstract

DNBR (Departure from Nucleate Boiling Ratio) evaluation method has been selected to quantify fuel damage during operation of the NPP (Nuclear Power Plant). Various initial conditions such as core inlet temperature, pressure of the reactor coolant system, coolant flow rate, APD (Axial Power Distribution), etc. are used for DNBR evaluation method of Non-LOCA (Non Loss Of Coolant Accident). In case of core inlet temperature, pressure of the reactor coolant system, coolant flow rate, limiting direction can be predicted in the range of LCO (Limiting Conditions for Operation). But APD has inherent limitations in finding limiting direction. The current DNBR evaluation method for Non-LOCA safety analysis has the problem of performing repetitive and inefficient sensitivity study using thousands of APDs generated by nuclear design code under various conditions.

Therefore, this paper proposes new DNBR evaluation method to determine the limiting APD for Non-LOCA by grouping of APDs. The grouping is performed by using softmax regression to analyze the features of thousands of APDs and 3670 APDs are organized into 25 groups through the shape classification model. More than 2 million DNBR based on APDs, initial conditions, accident conditions and ROPM have been calculated by using THALES code to select representative 36 APDs for 25 groups. In particular, the accident conditions are determined from AOO accident results described in the current FSAR (Final Safety Analysis Report).

The approach on the new evaluation method based on tendency analysis of DNBR calculation results is presented with a limiting shape that exceeds the LHGR (Linear Heat Generation Rate) through comparison of shapes within the LHGR. The two derived limiting APDs classified by 80% RCS flow rate are expected to cover Non-LOCA and to be applied to APR1400. In addition, if the thermal-hydraulic conditions change, sensitivity analysis of representative 36 APDs should be performed to derive limiting APD. The quantitative evaluation shows that the new DNBR analysis save the current DNBR analysis time by more than 2 to 3 times.