تجزیه و تحلیل حساسیت به RELAP5 توسعه یافته برای یک راکتور تحقیقاتی TRIGA معمولی

The main aim of this work is to identify how much the code results are affected by the code user in the choice of, for example, the number of thermal hydraulic channels in a nuclear reactor nodalization. To perform this, two essential modifications were made on a previously validated nodalization for analysis of steady-state and forced recirculation off transient in the IPR-R1 TRIGA research reactor. Experimental data were taken as reference to compare the behavior of the reactor for two different types of modeling. The results highlight the necessity of sensitivity analysis to obtain the ideal modeling to simulate a specific system.

The user of a thermal hydraulic system code has a very large number of available basic elements (single volumes, pipes, branches, junctions, heat structures, pumps, etc.) to develop a detailed reactor nodalization. The model can reproduce a specific part or the whole system to be simulated. However, as there is not a fixed rule to perform the nodalization, a large responsibility is passed to the user of the code in order to develop an adequate model scheme which makes best use of the various modules and the prediction capabilities of the specific code (Petruzzi and D’Auria, 2008 and D’Auria and Galassi, 1998). The influence of the user on calculations for thermal hydraulic codes is clearly evident in the relatively wide variation in results from different organizations and code users participating in international standard problem (ISP) exercises. Although some of the user-to-user variation is due in part to the use of different computer codes, a substantial variation is also observed when different users apply the same codes (Adorni, 2007). Particularly, in the RELAP5 code, a physical system consisting of flow paths, volumes, areas, etc., is simulated by building a network of volumes connected with junctions. Therefore, the transformation of the physical system to a system of volumes and junctions is an approximate process (US NRC, 2001). In spite of the substantial progress over the past two decades in the development of more accurate and more user tolerant computer codes for accident analysis, the user can still have a significant effect on the quality of the analyses. Sensitivity analysis including systematic variations in code input variables or modeling parameters, must be used to help identify the relevant parameters necessary for an accident analysis by ranking the influence of accident phenomena or to bound the overall results of the analysis. Results of experiments can also be used to identify important parameters. The main aim of this work is to identify how much the code results are affected by code user choices. To perform this, two modifications were made on a previous validated nodalization for analysis of steady-state and forced recirculation off transient in the IPR-R1 TRIGA research reactor (Reis et al., 2010). The modifications include: (1) variation in the number of the thermal hydraulic (TH) channels in the core (from 13 in the original nodalization to 91 in the modified nodalization) and (2) insertion of cross-flow model in several core channels of the new nodalization. The original nodalization has been identified with the name 13-THC, as a reference to the 13 TH channels present in the core of the original modeling and the new one, the modified nodalization, is referenced as 91-THC. The comparison between the results obtained with both nodalizations is presented here. The code used in this study is the RELAP5 Mod3.3, the same code used in Reis et al. (2010). Experimental data from Veloso (2004) were taken as reference to compare the behavior of the reactor for different types of model. The results demonstrate the necessity for sensitivity analysis to obtain the ideal modeling to simulate a system. The RELAP5 system code was firstly developed to simulate transient scenarios in power reactors such as PWR and BWR. However, several works have been performed to investigate the applicability of the code to research reactors operating conditions (Antariksawan et al., 2005, Khedr et al., 2005 and Marcum et al., 2010).

Sensitivity analysis was presented in this work being one of the steps for the validation process of the IPR-R1 modeling using the RELAP5 code. Two essential modifications were made on the previous validated nodalization (Reis et al., 2010) including variation in the number of the thermal hydraulic channels in the core (from 13 in the original nodalization to 91 in the modified nodalization) and the insertion of cross-flow model in several of the 91 core channels in the new nodalization. The comparison between the new and the preceding results has presented good results, that is, it was verified few variation of the observed parameters for both models. In spite of the few differences, the new nodalization with 91 TH channels presented results slightly better than the 13-THC one. This is due to the presence of the cross-flow model in the core interconnecting the channels and improving the coolant flow between them and consequently providing a more realistic temperature distribution in the core. However, the cross-flow model caused flow rate oscillation probably due the axial and radial water moving. Even if being of small amplitude, the oscillations can increase in time generating instability in the core mainly in the power reactors as the boiling water reactor, for example. Increasing of the number of TH channels has a serious consequence for the time processing calculation that was increased in approximately 16 times respect to the original nodalization applying the same time step. The simulation using the new core configuration can take up to 4 days to complete 10,000 s of calculation. In the transient analysis, both RELAP5 models reproduced in perfect accordance the increasing of the coolant temperature presenting the same rate of that obtained experimentally. Therefore, it is possible to conclude that both models represent the IPR-R1 behavior in a good way confirming the applicability of the RELAP5 for the natural circulation of the TRIGA. However, for fast and general results, it is more appropriate to use the 13-THC model. If it is necessary more detailed analysis of a specific part of the core, the 91-THC model is preferable. In this way, the user of the code must develop and to investigate the model scheme which make best use of the various modules and the prediction capabilities of the code considering also the type of required investigation.