Effect on Code Predictions by Changing the Code Version of Relap5 on SBLOCA for Test 9.1B in BETHSY Test Facility

摘要

This paper mainly deals with the effect on code predictions by changing the code version of Relap5 on SBLOCA for test 9.1B in BETHSY integral test facility. This test was carried out in December 1989 with the objective to contribute to the validation of the physical basis of emergency operating procedures and the verification of the computer code and this test was also considered for CSNI International Standard Problem 18. Accident analysis of this test was also carried out by University of Pisa with Relap 5/mod 2 as a part of ISP27 in 1991. A noticeable progress in the capabilities of system code has been observed in past fifteen years. In this analysis updated version of best estimate code Relap 5/mod 3.3 has been used. In this code both the options are available for critical flow at the junction, modified Henry Fauske model (with only one discharge coefficient and thermal non equilibrium constant), and original model (with sub-cooled, two-phase and superheated discharge coefficient). Post test analyses are carried out with both the option and effect of updated code version (with compared to mod 2) on result predictions have been documented in this paper. The aim of this analysis is also to enlarge the data base of Code with the capability of Internal Assessment of Uncertainty (CIAU) developed at University of Pisa. Therefore, this analysis is carried out with all the procedure lead by Uncertainty Methodology based on Accuracy Extrapolation (UMAE). In order to achieve a qualified test facility nodalization both 'steady state level' and 'on transient level' qualifications are demonstrated. It is concluded that overall qualitative and quantitative accuracy of code prediction (mod 3.3) are acceptable as per UMAE. However noticeable effect of change of version of code has been observed from mod 2 to mod 3.3. Discharge coefficient for the modified Henry Fauske model is very sensitive for the break modeling point of view. The thermal non equilibrium constant for break discharge modeling is not having much effect on the analysis results. Especially when the core is refilled, code under predicts the break flow and integral break flow consequently code over predicts primary mass inventory. It is found that with the same input deck all the significant events and phenomena for the mod 3.3 occurring about 600 s earlier with compare to mod 2 calculation. Up to refilling of core, time sequence of the entire significant events was after the experiments in mod 2 calculation whereas in mod 3.3 it is before the experimental value. With new version of code better prediction for clad temperature during dry out has been observed. In both version of code prediction of results are poor for the last 2500 s of transients this may be due to large error in the experimental data. It is concluded that overall prediction of mod. 3.3 is better than mod 2 and overall good prediction can be obtained by both the option of mod 3.3.