Simulating safety for high temperature reactors
Jan-Patrice Simoneau, Julien Champigny, Framatome ANP, France.
Brian Mays, Lewis Lommers, Framatome ANP Inc., USA.

Introduction

When presented with a problem that combines 1000 ºC plus temperatures; convection, conduction, and radiation; time scales of over one hundred hours; and length scales that range from two millimeters to twenty meters, most CFD engineers and CFD codes would (quite understandably) admit defeat without so much as applying a boundary condition in anger.

But having been presented with just such a scenario, the engineers at AREVA didn’t raise the white flag, they took up their STAR-CD manuals and got to work.

The result: they have successfully developed a methodology for using STAR-CD to simulate the cooldown of a High Temperature Reactor (HTR), thereby obtaining a level of understanding which would otherwise be unobtainable through experimental work alone.

Recent years have seen a resurgence in interest in HTRs. This is driven by the possibility of using nuclear energy for the production of process heat (to be used for hydrogen production, for example) as well as the inherent safety characteristics of HTRs.

The most efficient means of hydrogen production, such as hightemperature electrolysis or thermo-chemical water splitting, require very high temperatures. As a result, HTRs have been adapted to increase their output temperatures from approximately 850ºC to temperatures approaching 1000ºC.

An important safety characteristic of the HTR design is that, if the normal heat sink becomes unavailable, the heat generated by the fuel is passively removed to cooling panels along the cavity walls by conduction and radiation. Consequently, the fuel particle temperatures remain at acceptable levels even when the normal heat removal method fails. Two variations of this “failure” scenario were investigated for an adapted HTR design. Figure 1, shows a schematic of the reactor.

The test case

Two hypothetical scenarios are under investigation here: in the first, the helium coolant is lost and the primary loop is in a depressurized state (called Depressurized Conduction Cooldown or DCC); in the other, the helium coolant is retained and the primary loop remains pressurized (called Pressurized Conduction Cooldown or PCC). The key difference for the analysis is that, during PCC, natural convection effects within the primary loop significantly alter the temperature profiles in the core.

Both STAR-CD simulations start at the point when the reactor is operating at full power (600 MW) and all heat sinks are lost except for the Reactor Cavity Cooling System (RCCS, a set of vertical panels along the concrete walls of the cavity containing the reactor vessel, which remove heat from the cavity to the outside). At this point the protection system automatically drops control rods into the core to shut down the reactor. The simulations end one hundred hours (or over four days) of simulated time later, when the energy removed from the reactor exceeds the heat produced by the decay of radionuclides in the core, thereby causing the total energy in the system to decrease.

The main barrier to simulating such long time scales is that time steps in transient calculations are limited by the requirement that the Courant number-a measure of the number of cells that information travels across in one time step-remains below fifty. Here, this corresponds to a maximum time step size of 0.1 seconds. In order to maintain calculation times at feasible magnitudes, a novel method was used. The calculation begins by solving both the momentum and temperature fields for the initial one thousand seconds. Then it alternates between solving both momentum and temperature fields using small time steps and solving the temperature fields alone, with the momentum field frozen, using large time steps.

Running one calculation, in which both the momentum and temperature fields were solved at every time step for ten hours of simulated time, and then comparing the results of this calculation with the results of the method described above validated this approach. Very little difference was found.

The computational domain is a 30º section of the reactor vessel (shown in figure 2). The various materials in the reactor are either modeled explicitly in STAR-CD or homogeneously: i.e., the “mean” physical properties of a particular heterogeneous component are used (an example being the fuel elements, which are composed of fuel compacts and a graphite web containing holes for the compacts and the coolant).

DCC Results

Figure 3 shows the time history of temperatures in the reactor during a DCC transient. With coolant no longer flowing downward through the core, the hottest region moves upward from the bottom of the core to the center. Meanwhile, the upper and lower parts of the reactor vessel interior—which are at the coolant’s inlet temperature during normal operation—become cooler. The core experiences its highest temperature of 1450 ºC at approximately eighty hours after shutdown. Initially during the transient, the temperature of the pressure vessel decreases as it is cooled by the RCCS. After ten hours, however, some of the heat from the core has passed through the outer graphite blocks and produces a rise in the peak vessel temperature, which reaches a maximum value (of nearly 480ºC) after one hundred hours of simulated time.

PCC Results

Figures 4 and 5 show the temperature profiles in the fluid and solid regions against time for a PCC transient. The natural convection currents develop quickly, with fluid circulating up through the core channels and reflector gaps. The STAR-CD simulation enabled detailed insight into the flow patterns and how they evolve over the cooldown period. A significant change in the flow patterns was observed after approximately ten hours, with the flow currents effectively reversing direction.

An important finding was that, although the peak temperatures are not much higher during PCC than under normal operation, the natural convection currents cause the hot spot to move up the reactor core, exposing the top of the reactor core to high temperatures. Although the DCC case exhibits higher peak fuel temperatures than the PCC case, the thermal stresses on the upper part of the reactor are much greater for the PCC.

Finally, figure 3 shows that, once the initial ten hours of the transient have passed and an offset is established, the behavior in the DCC and PCC scenarios is very similar.

Conclusion

A key characteristic of HTRs is the inherent safety of their design, due to passive heat removal. A methodology has been developed and actuated to simulate the reactor cooldown over a duration of one hundred hours. Initial calculations have been carried out to compare two scenarios, DCC and PCC. The simulation yielded detailed information about the flow patterns in the vessel and the transient thermal behavior. Further details can be found in “3D Simulation of the Coupled Convective, Conductive and Radiative Heat Transfer during Decay Heat Removal in a HTR” J. Simoneau, J. Champigny, B. Mays, and L. Lommers, Proceedings of the 11th International Topical Meeting on Nuclear Reactor Thermal-Hydraulics (NURETH-11), Avignon, France, October 2005. Figures 01: Section through reactor 02: Computational domain – solid part 03: Core temperature against time: PCC and DCC 04: Pressurized Conduction Cooldown – Fluid Part 05: Pressurized Conduction Cooldown – Solid Part

Further details can be found in “3D Simulation of the Coupled Convective, Conductive and
Radiative Heat Transfer during Decay Heat Removal in a HTR” J. Simoneau, J. Champigny, B.
Mays, and L. Lommers, Proceedings of the 11th International Topical Meeting on Nuclear
Reactor Thermal-Hydraulics (NURETH-11), Avignon, France, October 2005.
Figures
01: Section through reactor
02: Computational domain – solid part
03: Core temperature against time: PCC and DCC
04: Pressurized Conduction Cooldown – Fluid Part
05: Pressurized Conduction Cooldown – Solid Part