Gamma Dose Rate Analysis in case of loss of water event at the Jožef Stefan Institute TRIGA Mark II Research Reactor
Anže Jazbec Jožef Stefan Institute
Jamova cesta 39 1000 Ljubljana, Slovenia
anze.jazbec@ijs.si
Bor Kos, Luka Snoj Jožef Stefan Institute
Jamova cesta 39 1000 Ljubljana, Slovenia bor.kos@ijs.si, luka.snoj@ijs.si
ABSTRACT
Monte Carlo particle transport code MCNP was used to calculate dose rates around the biological shield of the Jožef Stefan Institute (JSI) TRIGA research reactor in case of loss of water event (LOWE). In order to analyse dose rate fields inside the reactor building in case of LOWE we modelled the reactor, biological shield and the reactor building including the control room to study. If LOWE occurs during reactor operation, dose rate at reactor platform exceeds values of 10 Sv/h and 1 mSv/h inside control room. 10 days later, dose rates are low enough, i.e. less than 20 mSv/h at the platform to start with corrective actions.
1 INTRODUCTION
Since 2007 the Monte Carlo particle transport code MCNP [1] has been used for calculation of reactor physical parameters as well as of neutron and gamma fields inside and in the vicinity of the Jožef Stefan Institute (JSI) TRIGA reactor core ([2] – [4]). In 2012 MCNP was used to characterise ex-core irradiation facilities ([5] and [6]).
In the period 2012-2016 a reactor periodic safety review was performed [7]. One of the findings was that some methods to determine dose rates in case of accidents, e.g. loss of water event (LOWE) are relatively old and it is suggested that analyses are repeated with up-to-date methods. Hence a decision was made to use best available methods and repeat safety analyses.
In the JSI TRIGA reactor, water acts as moderator, coolant and shield. It is important to note that in the event of water leakage from the reactor tank, cooling of the fuel by natural convection in air is sufficient to keep the fuel temperature below safety limit of 600 °C. The major consequence is loss of radiation shield. It was decided to use state of the art particle transport code MCNP for calculating gamma dose rates outside the reactor biological shield. Such calculations are practically impossible to perform with analog Monte Carlo method due to poor statistics so far away from the particle source. Hence variance reduction methods should be used. It was decided to use AutomateD VAriaNce reducTion Generator (ADVANTG) code [8]
to calculate variance reduction parameters for Monte Carlo particle transport. In addition, the radiation source should also be determined by using the best available code. It was decided to use the SERPENT code [9] to determine the radiation source term, since it can be used to perform burnup calculations.
2 COMPUTATIONAL TOOLS 2.1 Codes
MCNP, a General Monte Carlo N-Particle Transport Code, developed by the Los Alamos National Laboratory [1], is a general-purpose, continuous-energy, generalized-geometry, time- dependent coupled neutron/photon/electron Monte Carlo transport code. It simulates particle transport by stochastically sampling the individual particle events and tally them to estimate their average behaviour. The user-generated input file contains all the necessary information such as the geometry specification, the description of the materials with the selection of cross- section evaluations to be used, the location and characteristics of the source, and the tallies of interest. The ENDF/B-VII.1 library [10] was used in our calculations.
ADVANTG [8] is a computer code that performs an automatic generation of variance reduction parameters for neutron and photon transport problems defined in a MCNP input file.
The software was developed by the Oak Ridge National Laboratory to accelerate Monte Carlo simulations and to reduce required CPU times as well as user-time needed to generate variance reduction parameters. It generates space- and energy-dependent mesh-based weight windows, and biased source distribution using the three-dimensional discrete ordinates solution of the adjoint and forward transport equation. The method was validated on several fission and fusion benchmark experiments ([11] – [13]).
Serpent [9] is a VTT developed Monte Carlo code and is used to perform physical calculations of burnup. Operational history can be divided into multiple intervals with different normalizations, defined by power, power density, total flux, fission or source rate. Depletion steps are given in units of burnup or time. With the increased multi-core CPU’s capabilities, full 3-D burnup calculations of research reactors are possible in acceptable time.
3 COMPUTATIONAL MODEL 3.1 Geometry
A complete geometrical model of the JSI reactor building was designed in MCNP that includes detailed geometry of reactor core components, irradiation facilities, reactor hall, reactor platform, control room and reactor basement including spent fuel pool [14]. The model was firstly generated in CAD format (Figure 1) and later transformed into MCNP version including all required data on material composition and densities. The transformation was done using a program written in Grasshopper that exports basic geometrical shapes from CAD to MCNP format [15]. The origin of x and y axes is set to the centre of the core. The origin for z axis is set on the ground level of the reactor hall.
3.2 Source term
There are two sources of gamma ray in a shutdown nuclear reactor, i.e. fission products in irradiated fuel and activation product in the fuel and structural components in the fuel and the core
Major radiations source from irradiated fuel are gamma rays while neutrons are negligible. Hence only photons - gamma rays were simulated if there is no coolant surrounding fuel elements the reactor core is deeply subcritical [16]. If there are any neutrons emitted from the fuel, their contribution to the dose rates is considered negligible [4].
The first assumption made is that delayed gamma rays are emitted from fuel region only.
For the first order of approximation, only gamma rays emitted from fuel region were taken into account. The next assumption is that delayed gamma intensity from the fuel elements is proportional to the fission rate during reactor operation. Eigenvalue mode calculation in MCNP was used to calculate total fission rate profile in each fuel element. In order to make distribution as accurate as possible, each fuel element was divided in vertical direction into 100 discs. Radial distribution was taken into account as well. Spatial distribution of Modelled delayed gamma source was the same as calculated normalized spatial distribution of neutron flux.
Figure 1: CAD model of the reactor building.
Delayed gamma spectrum changes through the time after reactor shutdown, but not significantly which can be seen on Figure 2, where gamma ray spectra of irradiated fuel 1h and 20 h after shutdown is presented. Spectra was obtained using R2S method [4]. A Rigorous two- step method (R2S) utilizes Monte Carlo particle transport code MCNP 6.1 for particle transport part of the problem, and FISPACT-II [17] for neutron activation and delayed-ray generation, and custom Python scripts, joining the two codes. EAF 2010 cross section library [18] was used. For the following analysis, gamma spectra of irradiated fuel 1 hour after shutdown was taken. According to the small difference in spectra between 1 h and 20 h after shutdown, such assumption does not affect the final result (dose rate) strongly and can be neglected.
Figure 2: Gamma ray spectra of irradiated fuel 1h and 20 h after shutdown. Gamma ray spectrum integral is normalised to 1.
The calculated photon fluxes in MCNP are normalized per source particle, hence there was a need to calculate absolute activity of each fuel element. To obtain activity at certain point of reactor operation, the whole history of JSI TRIGA reactor was studied [19] using reactor operational log books from year 1966 until present. The calculations were performed using the SERPENT Monte Carlo neutron transport code [9], which is already established as a viable
burnup code, due to its built-in burnup routine. 300 different isotopes (fission products) were taken into account and transported from the previous fuel cycle (core configuration) into the next one. ENDF/B-VII.0 [10] nuclear library was used. Calculated activities for all fuel elements after shutdown for several time periods are presented In Table 1. For the following analysis, conservative operational parameters were taken – reactor core containing 60 fuel elements was in operation for 2 months at full power. Therefore, all calculated dose rates are overestimated by about one order of magnitude since in reality, reactor is in operation by maximum of 20 hours without interruptions.
Table 1: Core activity after reactor shutdown for several time periods Time after shutdown Fuel activity [Bq]
1 second 4.37E+16
1 day 7.29E+15
10 days 2.82E+15
100 days 5.59E+14
4 METHOD VALIDATION
To confirm all the assumptions made provide a satisfactory result, a simple experiment was done. One of the fuel elements was packed inside a transport cask (1330 kg of lead, 17 cm thick walls). The cask was placed onto the floor of the reactor hall and dose rates around it were measured. The same configuration was also modelled in MCNP and dose rates were calculated at the same locations. For that calculation, realistic operational history was taken in order to recreate dose rates around transport cask. The difference was within the order of magnitude for vertical direction and even smaller (within 25 %) in radial direction. We believe that such agreement is sufficient for performing the LOWE analysis so the method can be applied to all fuel elements loaded into the core especially since accidental events are analysed in conservative way, which was already ensured by overestimating maximum operational time of the reactor.
5 DOSE RATES AT REACTOR PLATFORM
After LOWE, the highest dose rates are expected at reactor platform. Therefore, the first set of calculations was to estimate dose rates there. Dose rates were calculated at 4 different locations presented in Figure 3 – left. Furthermore, we were also interested in dose rates inside the reactor tank and how the dose rate drops in dependence of the vertical distance. Therefore, dose rates were calculated also above the core at several locations (Figure 3 – right).
Figure 3: Left: red spheres at reactor platform, where dose rates were calculated. Right: red spheres were dose rates were calculated vertically above the core.
Calculated dose rates are presented in Table 2. One can see that if LOWE occurs during reactor operation, reactor platform has to be evacuated immediately. The dose rate above reactor tank exceeds 10 Sv/h. Dose rate at reactor platform (average over points 1 to 4) is a lot lower, but still too high for staff to enter the area. Annual allowed dose for personnel would be achieved in about 5 minutes. After approximately 10 days, dose rates drop low enough to start with corrective actions like adding shield on top of the pool/core. At that time, operators would receive yearly allowed dose in about 80 minutes.
Table 2: Dose rates at reactor platform at 10 preselected locations for different time periods after reactor shutdown.
No. Location (x, y, z)
Dose rate after shutdown – H*(10) [mSv /h] 1 σ statistical uncertainty [%]
1s 1d 10d 100d
1 200, 0, 750 235 39.2 15.2 3.00 0.5
2 0, 200, 750 239 39.9 15.4 3.06 0.35
3 -200, 0, 750 235 39.1 15.1 3.00 0.33
4 0, -200, 750 230 38.4 14.8 2.94 0.34
5 0, 0, 200 209,000 34,900 13,500 2,660 2.06
6 0, 0, 300 83,600 13,900 5,400 1,070 3.4
7 0, 0, 500 32,600 5,440 2,100 417 0.87
8 0, 0, 700 13,500 2,250 872 173 0.92
9 0, 0, 1000 4,820 804 311 61.6 0.83
10 0, 0, 1500 1 670 278 108 21.3 1.23
From the operational experience it is known that “gamma beam” from the core during reactor operation is highly collimated. We were interested if it is the same during LOWE and what would be gamma backscatter at the platform. This can be calculated in MCNP by flagging the particles that contribute to the dose and travel through the walls and roof of the reactor building. The contribution of the backscattered gamma and total gamma is presented in Table 3. Backscattered radiation compared to the total one can be neglected for the reactor platform, since it represents only few percent of the dose rate. On the other hand, at regions that are shielded for direct radiation from the core, backscattered gamma radiation prevails.
Table 3: Dose rates from scatter gamma rays at reactor platform (average over points 1 to 4) several times after reactor shutdown. Dose rate can be compared to the total one.
Dose rate Gamma dose rate – H*(10) [mSv/h] 1 σ statistical uncertainty [%]
1s 1d 10d 100d
Total 235 39.2 15.1 3.00 0.38
Backscattered 6.5 1.1 0.42 0.083 0.35
The great advantage of calculations over measurements is that we can obtain result for a larger region at the same time while measurements can only be taken at specific points separately. An example of calculated data for larger area is presented on Figure 4. It is seen that the gamma stream from the uncovered core is highly collimated. That explains lower dose rates on the reactor platform away from the pool – these are two orders of magnitude lower than dose rates just above the pool.
6 DOSE RATES INSIDE CONTROL ROOM
The next interesting region for dose rates predictions during LOWE is the control room.
From the control room, operators can control the reactor and even more important during an accidental scenario, read various parameters like dose rates in the reactor hall, temperatures of the instrumented fuel elements, level of the coolant etc. Furthermore, the control room can serve
as headquarters during accidental scenarios. Dose rates were calculated at 10 different locations shown in Figure 5.
Figure 4: Calculated dose rate at reactor platform normalized per source particle (Sv/hBq).
Calculated results of dose rates for selected locations are presented in Table 4. If LOWE happens during operation, dose rates in the control room exceeds 1 mSv/h and therefore the control room has to be evacuated. If LOWE occurs one day after reactor operation, dose rates near the instrumentation panel are low enough to read important data. LOWE 10 days after reactor operation results in sage dose rates for the operating staff in the control room. Analogous to the reactor platform, dose rates in the control room were also calculated across a larger area.
The results are presented in Figure 6. On the horizontal cross section of the dose rate field, one can see decrease of dose rate with distance from the reactor core. Major contribution to the radiation does originate from back-scattered gamma radiation from the reactor hall ceiling.
(Figure 6 – right).
Table 4: Dose rates at reactor control room at 10 preselected locations for different time periods after reactor shutdown.
No. Location (x, y, z)
Dose rate after shutdown – H*(10) [µSv /h] 1 σ statistical uncertainty [%]
1s 1d 10d 100d
1 -150, -1050, 450 2,930 489 189 37.5 1.37
2 -150, -1250, 450 1,710 286 111 21.9 1.75
3 -350, -1250, 450 1,390 232 89.9 17.8 1.67
4 50, -1250, 450 649 108 41.9 8.3 2.26
5 -150, -1600, 450 249 41.5 16.1 3.2 1.98
6 -150, -1050, 550 3,200 534 206 40.9 1.26
7 -150, -1250, 550 744 124 48.0 9.5 3.14
8 -350, -1250, 550 592 98.8 38.2 7.6 2.37
9 50, -1250, 550 330 55.1 21.3 4.2 2.79
10 -150, -1600, 550 170 28.4 11.0 2.2 2.78
7 EVALUATION OF UNCERTAINTIES.
In Tables 2-4 we present 1 σ relative statistical uncertainty of calculated physical quantities. They do not include uncertainties due to the inaccurate modelling, unknown material composition like content of the water inside concrete and fuel elements activity and the uncertainty due to nuclear data. According to the dose rate simulations around the fuel transport cask, where there was an irradiated fuel element inside (validation experiment), we can estimate
gained results are accurate down to an order of magnitude. Since the whole analysis is done in conservative manner (overestimated fuel activity), we can conclude that dose rates in case of actual LOWE would not exceed calculated results.
Figure 5: Spheres in control room, where dose rates were calculated. Dose rates were also calculated 1 m above each sphere (not presented to obtain clearer view).
Figure 6: Calculated dose rate inside control room normalized per source particle (Sv/h·Bq).
Left: horizontal cross section, right: vertical cross section.
8 CONCLUSION
In this paper dose rate calculations were presented in case of LOWE at the JSI TRIGA reactor. For the first time the analysis was performed by using a Monte Carlo particle transport method.
Dose rate analyses in case of LOWE at JSI TRIGA reactor were made for the reactor platform and reactor control room. Dose rates depend strongly on the time since reactor shutdown, i.e., fuel cooling time. If LOWE occurs during operation, reactor goes subcritical due to missing moderator. Due to high dose rates on the order of 10 Sv/h there is a need for evacuating the reactor building. If LOWE occurs 10 days since the last reactor operation, the dose rate at the reactor platform is about 15 mSv/h, therefore, corrective actions can be done immediately. These predictions are conservative, i.e the dose rates are overestimated. Usually reactor operates several hours per day which would result in about an order of magnitude lower dose rates.
9 ACKNOWLEDGEMENT
The author would like to thank Anže Pungerčič and Klemen Ambrožič who provided fuel activities for selected cool down times and delayed gamma spectra.
REFERENCES
[1] X-5 Monte Carlo Team, MCNP - A General Monte Carlo N-Particle Transport Code, Version 5, Volume I & II, 2003.
[2] L. Snoj, et al., “Analysis of neutron flux distribution for the validation of computational methods for the optimization of RR utilization”, Appl. Radiat. Iso., Vol. 69, pp. 136-141, 2011.
[3] L. Snoj, et al., “Computational analysis of irradiation facilities at the JSI TRIGA reactor”, Appl. Radiat. Iso., Vol. 70, pp. 483-488, 2011.
[4] K. Ambrožič, L. Snoj, “Delayed Gamma Ray Modelling Around Irradiated JSI TRIGA Fuel Element by R2S Method «, NENE 2016, Portorož, 2016.
[5] V. Radulović, et al., “Characterization of the Ex-Core Irradiation Facilities of the JSI TRIGA Mark II Reactor”, NENE, Ljubljana, 2012.
[6] A. Kolšek, V. Radulovič, L. Snoj, “Using TRIGA Mark-II Research Reactor for Irradiation with Thermal Neutrons”, NENE conference, Bled, 2013.
[7] A. Jazbec, et al., “Periodic Safety Review of JSI TRIGA Mark II and Inspection of the Reactor Vessel”, IGORR Conference, Daejeon, 2013
[8] S.W. Mosher, et. al., “ADVANTG – An Automated Variance Reduction Parameter Generator”, ORNL/TM-2013/416, ORNL, 2013.
[9] J. Leppänen, et al., “The Serpent Monte Carlo code: Status, development and applications in 2013”, Ann. Nucl. Energy, 82 (2015) 142-150.
[10] M. B. Chadwick, et. al., “ENDF/B-VII.0: Next Generation Evaluated Nuclear Data Library for Nuclear Science and Technology”, Nuclear Data Sheets, Special Issue on Evaluated Nuclear Data File ENDF/B-VII.0, 2006.
[11] M. W. Shaver et. al., “Evaluation and Testing of the ADVANTG Code on SNM Detection”, PNNL-22793, Prepared for U.S. DOE under Contract DE-AC05-76RL01830, 2013.
[12] Kotnik, D., et. al., “Validation end evaluation of the ADVANTG hybrid code on the ICSBEP labyrinth benchmark experiment”, Ann. Of Nucl. Energy, Vol. 114, pp. 464-481, 2018.
[13] Kotnik, D., et. al., “Validation end evaluation of the ADVANTG code on the ICSBEP skyshine benchmark experiment”, Ann. of Nucl. Energy, Vol. 125, pp. 249-260, 2019.
[14] A. Jazbec, L. Snoj, 3D Model of Jožef Stefan Institute TRIGA Mark II Reactor, NENE 2016, Portorož 2016.
[15] B. Kos, L. Snoj, On using Grasshopper add-on for CAD to MCNP conversion, PHYSOR 2016, Idaho, 2016.
[16] General atomic, “250 kW TRGA Mark II Reactor Mechanical Maintenance and Operating Manual for the Jožef Stefan Nuclear Institute Ljubljana”, Yugoslavia, GA-6535, 1965 [17] J. C. Sublet, et. al., “The FISPACT-II User Manual” UK Atomic Energy Authority, Culham
Science Centre, Oxford, 2015.
[18] J. C. Sublet, et al., “The European Activation File: EAF-2010 neutron-induced cross section library”, CCFE-R (10) 05, 2010.
[19] A. Pungerčič, et al., “Analiza zgorelosti gorivnih elementov reaktorja TRIGA”, Master Thesis, Ljubljana, 2018.
