Accidents In Light Water Reactors Research Articles

Thermophysical properties of UO2-ZrO2 melt measured by aerodynamic levitation

This study addresses the measurement of thermophysical properties in molten core materials (corium), which is crucial for modeling severe accidents in light water reactors. Due to the measurement challenges at temperatures above 3000 K, there is a significant lack of experimental data for corium. Focusing on a UO2-ZrO2 melt (atomic ratio U/Zr=0.456), density and viscosity measurements were conducted using aerodynamic levitation, a non-contact method chosen to avoid interactions between the samples and container walls. Samples were prepared using the pressing sintering method and levitated by argon gas above a conical converging-diverging nozzle. Density was measured using the cooling traces method under the axisymmetric ellipsoid hypothesis, while viscosity was measured by inducing damped oscillations which occurred when the levitated droplet was forced to oscillate around its resonant frequency through acoustic excitation, followed by the cessation of the acoustic excitation. The provided data and derived equations for the density and viscosity of the UO2-ZrO2 melt contribute to enriching the thermophysical property database for materials in nuclear reactors.

Oxidation of molten zirconium-containing droplet in water

During a severe accident in light water reactors, the molten reactor core (corium) falls into a water pool in the form of a jet. Complex interactions may occur between the melt and coolant known as molten fuel–coolant interactions (FCI), including energetic coolant evaporation and metallic melt (e.g., Zr and Fe) oxidation. This may further lead to steam and hydrogen explosions, which are both substantial safety risks for nuclear power plants.The heat of reaction and hydrogen production during oxidation can influence the progress and severity of the accidents. For example, the reaction heat may prolong the liquid state of corium, potentially leading to high-intensity explosions, whereas the generated hydrogen can create a combustible atmosphere, increasing the risk of hydrogen explosion. Therefore, this study evaluates the hydrogen production and oxidation degree of molten metallic droplets falling into a water pool to improve the FCI models for the risk evaluation of severe accident safety. The MISTEE-OX facility at KTH, which has been primarily built to study steam explosions is modified to investigate oxidation during FCI and provide experimental data on the oxidation behaviour of metallic droplets (Zr/Fe) quenched in a subcooled water pool. The dynamics of the falling droplets and generated bubbles are recorded using a high-speed camera, and the total volume of the bubbles is measured using a graduated cylinder. This study presents preliminary experimental results of the oxidation between Zr/Fe droplets and water, as well as recent improvements in measurement methods and facility upgrades. Our research findings are useful to enhance the knowledge of the oxidation process in FCI phenomena and validate the related mechanistic models in FCI codes.

STAR-CCM+ simulation of debris bed formation in the unheated DEFCON-S experiment

In a severe accident of light-water reactor, molten corium can be formed and discharged into the reactor cavity. If the resulting debris bed is not adequately cooled, it could threaten the containment integrity by causing a molten core-concrete interaction. This study analyzes a non-heating debris bed formation experiment of DEFCON-S using the STAR-CCM+ code. Relevant CFD models are established to reflect the interaction between particles and fluids to describe the particle behavior and flow of fluids. The shape and size of debris bed are predicted through the CFD calculation. The rolling resistance becomes an important parameter in this simulation. The prediction with the default values of STAR-CCM+ forms a spread shape of particle bed compared to the experimental result. The shape of debris bed obtained with modified rolling resistance coefficients gives consistency with the DEFCON-S experiment, having a deviation in diameter of 4.6 % and a deviation in height of 3.8 %.

Mechanical degradation of duplex SiC-fiber reinforced SiC matrix composite tubes under a controlled high-temperature steam environment

This paper describes results from the mechanical evaluation of unirradiated SiC fiber–reinforced SiC matrix composite tubes under a controlled high-temperature steam environment. The experiments were aimed at identifying key material degradation behavior under environments relevant to loss-of-coolant accidents of light water reactors. Mechanical tests of the SiC composite tubes at 1000 °C under steam and inert environments were conducted using a unique test capability. The material tested was a duplex tube with a thick monolithic SiC layer on the outer surface. The tubes were subjected to preloading at ∼100 MPa in tension before exposure to high-temperature steam with up to 75 % of the preload at a constant displacement. In the presence of matrix cracks, the steam exposure caused embrittlement of the SiC composite tubes and failure at a stress level below the pretest stress. The material degradation was explained by a fiber oxidation model, which can be applied to various SiC cladding concepts. The embrittlement could be a limiting factor for using SiC cladding subjected to loss-of-coolant accident conditions.

Estimating Allowable Operator Action Time for Mitigating Multiple‐Failure Accidents in Korean Advanced Power Reactor

In the case of multiple‐failure accidents in light water reactors, it is imperative to demonstrate that a significant fuel damage can be prevented via prearranged mitigation measures performed by operators. However, the time elapsed from the initiation of the event to the implementation of mitigation actions significantly influences the progress and consequences of the accident. Therefore, assuming a specific time for operator mitigation actions does not provide sufficient insights for safety assessments, thereby calling for further investigations into the effect of the time delay from the prescribed value. In this study, the allowable operator action time for multiple‐failure accidents of the Advanced Power Reactor 1400 was estimated by means of the sensitivity analysis using multi‐dimensional analysis for reactor safety‐KINS standard (MARS‐KS), a best‐estimate thermal–hydraulic analysis code used by the Korean nuclear regulatory authority. Three multiple‐failure accidents were investigated, including small‐break loss‐of‐coolant accident with the loss of safety injection (SI), intermediate‐break loss‐of‐coolant accident with the loss of SI, and total loss of feedwater accident. The upper limit of the mitigation action time, which prevents core damage, was estimated based on the requirement for peak cladding temperature. This was in turn used to derive the tolerable delay from the assumed action time in the Korean accident management program for each multiple‐failure accident.

Numerical simulation of melt penetration in debris beds using MPS method

During a severe accident of light water reactors (LWR), melt penetration in a debris bed (also called infiltration) may occur due to the failure of a crust which accommodates a molten pool over the debris bed, or due to earlier re-melting of metallic components (i.e., Zr and Fe) which has lower melting points than oxidic ones (i.e., UO2 and ZrO2) of corium debris. Although understanding and modeling of melt infiltration plays a significant role in the prediction of severe accident progression, little work has been done in this respect due to its complexity. To fill this knowledge gap, the present study develops a Moving Particle Semi-implicit (MPS) code with various models representing key physics in melt infiltration. REMCOD experiment, recently carried out at KTH, has been simulated to validate the updated MPS code. The comparative results show that the melt infiltration in the tests can be reproduced by the MPS code. For a debris bed of cylindrical particles, the melt infiltration process in the experiment can be predicted by using Sauter mean diameter of the particles in the simulation. The code is also applied to investigate the influences of important parameters on the melt infiltration, and it is found that contact angle has little impact on melt penetration process in the case of debris beds with large pore sizes.

Modeling of quenching temperature with the effects of axial conduction resulting from quench front propagation

Efficient quenching of fuel rods in a post-critical heat flux (post-CHF) film boiling environment is of utmost importance to maintain the nuclear fuel integrity. In the event of a loss of coolant accident (LOCA) in light water reactors, the ability to quench the fuel element surface is crucial to not compromise the core coolability. Many studies over the past several decades have indicated a dependence of the minimum film boiling temperature (Tmin) on system factors such as pressure and subcooling as well as local factors like surface properties. However, recent quenching analysis using high-resolution optical fiber sensors have indicated a substantial axial heat conduction present at the quenching front, which has the potential to change the local quenching temperature by a significant margin. In this paper, this axial conduction effect is investigated and applied to a previous Tmin model. A new semi-empirical model which considers distance from initial rewetting point as well as rod diameter is proposed using the optical fiber quenching data. This correlation assumes an initial rewetting temperature which initiates the quenching front propagation, and predicts the temperature drop locally due to ongoing film boiling beyond the quench front as well as heat conduction across the cladding from the significant axial temperature gradient. By doing so, this new correlation provides a significant improvement on earlier Tmin models by accounting for this temperature variation which has previously not been considered.

Effect of molybdenum release on UO2/MOX fuel oxidation under severe light water reactor accident conditions

To investigate the Mo release behavior and its influence on the oxidation of UO2/mixed oxide (MOX) fuel, the oxidation and evaporation (O&E) behaviors of Mo powders and their influencing mechanisms with respect to oxygen partial pressure around powders (pO2s) were researched by using a thermogravimetric analysis technique under Ar-0.1 % O2, Ar-1.0 % O2, and Ar-1.5 % O2 atmospheres in the temperature range from 1473 to 1673 K. The study results reveal that Mo O&E could quickly consume oxygen, substantially decreasing pO2s. This decrease in pO2s facilitates the attainment of an oxygen mass balance by enhancing the diffusion rate of O2(g) towards the powder surfaces and decelerating the O&E rates of the powders. Subsequently, the oxygen consumption rate with respect to Mo release (FO-con) and the potential oxidation rate of UO2 nuclear fuels (FO-ads) under the typical accident conditions created in VERCORS were estimated using the latest analytical model of Mo and the existing fuel oxidation model, respectively. The results demonstrate that, as accident proceeds, FO-con has the potential to surpass FO-ads, thereby retarding the oxidation of nuclear fuel.

Microstructure dependent burst behavior of oxide dispersion–strengthened FeCrAl cladding

Oxide dispersion–strengthened (ODS) FeCrAl alloys are an advanced cladding material that combines the high-temperature steam oxidation resistance of conventional wrought FeCrAl with the irradiation resistance and high-temperature strength of nanostructured ferritic alloys. In the present study, the effect of microstructure on the performance of the ODS FeCrAl alloy system in light-water reactor–accident scenarios is evaluated using a combination of experimentation and modeling. ODS FeCrAl alloy with nominal composition Fe-12Cr-6Al-0.3Zr-0.3Y2O3 was produced into a thin-walled tube via high-precision tube rolling. A portion of the as-received (AR) tube was subjected to a full recrystallization heat treatment. Both variants were subjected to burst testing at Oak Ridge National Laboratory’s Severe Accident Test Station. The recrystallized tube exhibited burst temperature margins 200 °C higher than that of C26M, whereas the AR tube only achieved burst properties comparable to the wrought FeCrAl alloy. The burst experiments were simulated in the BISON finite element code through the development of new viscoplastic material databases for each heat treatment condition, informed by high temperature tensile tests of tube specimens and existing literature data on creep mechanisms. Finally, an ODS FeCrAl burst criterion was obtained using simulated burst experiments across the entire range of test conditions.

An experimental study on quenching of metallic spheres in seawater

Motivated by quenching of melt droplets and debris particles in seawater during a hypothetical severe accident of light water reactors (LWRs), the quenching process of stainless-steel spheres in a seawater pool was investigated in the present study. The polished spheres were pre-heated up to 1000℃ in an induction furnace with inserted atmosphere, and then immersed into the subcooled water pool in a chamber made of transparent quartz. A thermocouple was embedded in the center of the sphere to measure the history of the sphere’s temperature, while a high-speed camera was employed to record the quenching process and vapor film dynamics. Quantitative data, e.g. film thickness and oscillation, of the vapor film evolution during the quenching process were obtained through an image processing program developed on the MATLAB platform.The experimental results indicated that the quenching rate was higher in seawater than in deionized water, and the vapor film collapsed at a temperature higher than the Leidenfrost temperature in deionized water. The trend appeared more significant with increasing subcooling of water. The comparison of the quenched spheres suggested the surface of the sphere in seawater achieved higher degree of discoloration and roughening than that in deionized water, probably due to the additives of salt which change water properties. The image processing and analysis revealed that the vapor film had different thickness profile along the upper and lower hemispheres, and the averaged film thickness was smaller in seawater than in de-ionized water during the stage of stable film boiling. The vapor film was thinning and oscillating with time, and its fluctuations appeared different frequencies and amplitudes at the upper and lower locations, which may explain the mechanism of the earlier collapse of vapor film in the quenching process of a high-temperature sphere in seawater.

Evaluation of jet breakup length with a CFD code under steam generation condition in a pre-flooded cavity

When the reactor vessel is penetrated in a severe accident of light water reactor, the molten fuel-coolant interaction including the jet breakup occurs and the jet breakup length becomes one of the important parameters. Most numerical studies on jet breakup process have been carried out using dedicated computer codes. Some researchers are trying to apply commercial CFD codes to their investigations on comprehensive jet breakup process. However, the complexity of the phenomena limits the CFD application only to hydrodynamic aspects. In the present study, numerical analysis of jet breakup under vapor generation is pursued using the STAR-CCM + code. The obtained CFD prediction of the MATE09 experiment shows jet breakup progression patterns consistent to the images taken in the experiment. Further, the predicted positions of leading head, which determine the jet breakup length, are in good agreement with the MATE 09 data. The investigation of hydrodynamic effects on the jet breakup with higher jet velocity results in a stronger shear force and earlier jet breakup process even though there exists the vapor pocket around the corium jet. In future studies, the effect of vapor intensity on the jet breakup length would be investigated further by changing other parameters.

A numerical study on the levitation system for droplet preparation in a fuel-coolant interaction experiment

The MISTEE facility at KTH was designed to investigate the process and phenomena of a molten droplet falling into a water pool that may be encountered in fuel-coolant interactions (FCI) during a severe accident of light water reactors. An aerodynamic levitation mechanism is proposed to hold the molten droplet during its preparation (melting and heating up to a prescribed temperature) in an induction furnace. The crucible is flushed with argon through an injection nozzle at the bottom to prevent the droplet from falling out of the crucible. A numerical simulation of the aerodynamic levitation system is performed in the present study with the objective of determining and optimizing the design. The problem was simplified as an isothermal two-phase flow in an axisymmetric geometry. The simulation is realized through ANSYS Fluent v17 platform, which employs the VOF method to track interfaces between two phases and the SST k-omega model to describe turbulence flow of argon gas. The numerical model is validated against tests performed in the MISTEE facility after mesh sensitivity study. It is then applied to investigate the impacts of various parameters on the facility levitation capability and the droplet stability. According to the simulation results, stable molten droplets can be obtained in the designed experimental setup. The simulation also provides the appropriate values of argon inlet velocity and sample mass at which a stable droplet can be obtained inside the crucible before its discharge. Either higher or lower inlet velocity will destabilize the formation of the droplet. Considering the temperature-dependent melt properties, both surface tension and viscosity affect the movement and deformation of the molten droplet. The wettability of melt on the crucible wall is critical to droplet formation, and it is found that a poor wettability can ensure the levitation of droplet.

DAVINCI-SP tests for debris bed formation and spreading in a pool with center conical structure under two-phase condition

In case of severe accidents in light water reactors (LWRs), the coolability of relocated corium from the reactor vessel is a significant safety issue. The failure in cooling and stabilizing the molten core in the containment vessel can threaten the integrity of the containment boundary. In a flooded cavity, a porous corium debris bed is expected to develop on the bottom of the cavity pool due to the melt jet breakup and fragmentation. The geometric configuration of the porous debris bed is important for accurate coolability assessment under accident conditions. Earlier, Kim et al. (2016a, 2016b) presented an analytical model to estimate the spreading of the debris bed in terms of two-phase flow and the debris injection parameters. Here we are testing the effect of conical bottom surface in passively dispersing the debris particle to obtain a levelled bed formation. For this, the previous model for debris bed development is modified to simulate the development of debris bed over conical bottom surface. The existing ‘The Debris Bed Research Apparatus for Validation of the Bubble-Induced Natural Convection Effect Issue (DAVINCI) - Spreading (SP) experimental facility was modified with a center conical structure. Five kilograms of stainless steel simulant debris were injected stepwise from the top of the water level, while air bubbles simulating the two-phase vapor flow were injected from the bottom of the particle catcher plate. The test results from the current study to describe the spreading of the debris bed in terms of two-phase flow, debris injection parameters, and conical bottom surface geometrical parameters are discussed. Subsequently, the model for characteristics length and side slope angle of debris bed were modified using the empirical parameters from the tests. Comparison between the tests results and modified model prediction are presented. Lastly, the effectiveness of the conical bottom surface on particle dispersion is discussed.

Experimental investigation of quenching phenomena in high temperature conical debris bed – Part 1: MONET tests

In case of severe accidents in light water reactors (LWRs), the coolability of relocated corium from the reactor vessel is a significant safety issue. In a flooded cavity, a porous corium debris bed is expected to develop on the bottom of the cavity pool due to the melt jet breakup and fragmentation. The debris bed needs to be quenched by water either flooding from the top or flooding from the bottom until continuous cooling is established. If it is not quenched, this mass will begin to re-melt due to decay heat and leads to Molten Corium Concreate Interaction (MCCI). Most of the studies reported have conservative assumptions of one-dimensional homogeneous particulate beds with top or bottom flooding condition of coolant. Understanding of debris beds quenching with multi-dimensional characteristics are limited. To investigate the quenching behavior of the conical debris bed by means of experiments, a test facility MCCI-mitigatiON through passive cooling Effect Test (MONET) apparatus is established. Specifically, the objective is to understand the mechanism of heat transfer and quench front movement inside the conical heated debris bed during quenching by cooling water supplied from the bottom of the bed. Two different particle bed bottom wall boundary conditions are studied; closed bottom wall and open bottom wall to understand the effect of water ingression from the bottom wall. The particulate debris bed is formed with two different types of particles; Ø 3 mm SS 304 and Ø 3 mm alumina. In addition, the effect on the quench front movement inside the particle bed has been reported by using temperature measurements together with direct visualizations.

Experimental Investigation on Debris Bed Quenching With Additional Non-Condensable Gas Injection

Abstract Severe accidents of light water reactors with core degradation can lead to the formation of a, so-called, debris bed inside the reactor cavity. In the scenario of depleted residual water, the bed can partially melt and interact with the concrete underneath generating noncondensable gases (NCG) at the bottom of the particle bed, which will flow through the debris bed. The impact of additional gas on the quenching process can in principle be considered in thermal-hydraulic system codes such as ATHLET; however, there is still a need for experimental validation of respective models or verification of corresponding simulation results. Therefore, especially for the model validation of COCOMO-3D, which is implemented in ATHLET, a specific extension to the existing experimental database is required. Experimental results of the quenching behavior of a monodispersed particle bed at top-flooding cooling condition with additional NCG injection, utilizing the new built-up test facility FLOAT (flooding facility with gas injection), are presented.

A new potential interface tension model for MPS method avoiding unphysical particle cohesion

Improvement of the capability to accurately track the interface between immiscible phases/components is of great importance to the numerical analysis with complex multi-component and multi-phase flow dynamics, which might be involved in severe accidents of light water reactors, such as Molten Core-Concrete Interaction (MCCI). In order to model the interface tension between immiscible phases/components, Zhu et, al., precedingly developed a potential interface tension model in MPS method based on Kondo et al.‘s earlier model for surface tension and succeeded to stably model the interface tension qualitatively by introducing different attractive forces for different pairs of particle types. However, it has been known that pressure is overestimated in Zhu et, al.‘s potential interface tension model. Moreover, this study has shown that the fluid flow can be numerically hindered with Zhu et al.‘s potential interface tension model. These issues of Zhu et al.‘s potential interface tension model is caused by unphysical particle cohesion due to the inter-particle attractive forces determined by the surface tension coefficients, which are required even when the free surface is not considered. Meanwhile, in the framework of SPH method, Zhou et, al., proposed to model interface tension by introducing the repulsive force between different phases. Hence, this study aimed to develop a new potential interface tension model for MPS method which can overcome the conventional drawbacks of pressure overestimation or unphysical flow hindrance, by incorporating the repulsive force between immiscible fluid phases referring to Zhou et, al.‘s idea within SPH method.The interface tension in the present method is modeled as the repulsive force derived from the “repulsive potential” and is considered only near the interface to avoid the unphysical particle cohesion due to the inter-particle attractive forces. The amplitude coefficient in the repulsive potential is determined by the methodology similar to that proposed by Kondo, et, al.With the newly developed interface tension model, the common issue of numerical flow hindrance that occurred in the previous interface tension models in MPS method has been resolved. The validation of the newly developed interface tension model shows good agreement with theoretical evaluations of Laplace pressure and the droplet deformation ratio in a shear flow. Moreover, the new interface tension model has been validated against a 3-dimensional immiscible fluid stratification test, in which the new interface tension model shows good agreement with the test results and improvement of the pressure profile evaluations compared to that by Zhu et al.‘s model. Further extension of the present model so as to incorporate surface tension and wall wetting is one of the most important issues which require future development.

Reflooding of degraded cores in ASTEC V2.1: Modelling and validation on PEARL experiments

The injection of water into the reactor vessel at a late stage of a severe accident (core reflooding) is one of the key Severe Accident Management measure to stop the progression of the accident in Light Water Reactors. Therefore, an appropriate modelling of the physical phenomena involved in this process is of paramount importance for the enhancement of safety of these nuclear power plants.This article presents the improvements performed in the ASTEC V2.1 integral severe accident code in order to obtain a comprehensive reflooding model, which is valid regardless the core damage state. The main development consists in replacing, in presence of debris, the single momentum conservation equation (plus the drift flux relation) by two momentum conservation equations (one for each phase) with specific porous friction terms. Moreover, the calculation of the quench front position and the heat transfers downstream of the quench front have been harmonized.The physical models are validated against experimental data from the PEARL facility where a large heated debris bed is quenched. A wide range of thermal–hydraulic (pressure, injection velocity) and geometrical parameters (bed particle diameter and bypass thickness) is investigated. Generally, ASTEC V2.1 provides a good agreement for tests involving either a one or two-dimensional quench front progression. However, conversion ratios of produced steam to injected water seem to be overestimated in two-dimensional tests with fine particle diameters, which raises questions about the modelling of capillarity in those situations.The validated version of the code is then used to gain further insights on degraded core reflooding: radial steam redistribution is mainly driven by the large pressure gradients generated within the bed, whereas water entrainment is driven by the interfacial drag between the redistributed steam and the liquid, both contributing to hinder the coolability of the debris bed.

Physics-informed machine learning-aided framework for prediction of minimum film boiling temperature

Minimum film boiling temperature (TMFB) is a crucial parameter in post-critical heat flux (CHF) condition by determining the collapse of stable vapor film on overheated surface. Although many correlations have been suggested with a wide range of experimental data to predict the TMFB considering various parameters, they have failed to accommodate the universal quenching data due to limited regression capability addressing effects of multivariate parameters. Even though several data-driven machine learning models (DDML) showed improved prediction performance, their performance are dramatically degraded in extrapolation conditions, which are unincluded in the training process. To overcome the inherent problems of conventional correlations and DDMLs, physics-informed machine learning-aided frameworks (PIMLAF) for TMFB were developed in this study by combining the conventional correlations (Henry model and Groeneveld-Stewart correlation) and machine learning techniques (multi-layer perceptron and random forest) to overcome their limited regression capability and ‘black-box’ characteristics. For the interpolation condition, it was observed that the Groeneveld-Stewart correlation provides better baseline for TMFB than Henry model, and random forest (RF) model has advantage discovering the pattern between TMFB and input variables with tuning capability. Through benchmark studies corresponding to extrapolation condition of the training dataset, it was concluded that hybrid approach of RF model and Groeneveld-Stewart correlation exhibits the best performance on TMFB prediction for both interpolation and extrapolation conditions compared to standalone machine learning models, Groeneveld-Stewart correlation, and MLP-based PIMLAF. The suggested RF-based PIMLAF is expected to enhance the TMFB prediction performance during reflood phase of loss of coolant accidents in light water reactors.

Experimental study and analysis on the interfacial drag of two-phase flow in porous media

To reduce the uncertainty in coolability estimate of debris bed in the situation of severe accident in light water reactors, especially study the law of interfacial drag (Fi), two-phase flow tests are carried out to clarify how different parameters affect Fi and the ratio of Fi to the pressure drop. The results show that Fi leads to a decline of the pressure drops in packed beds with 3 mm and 6 mm spheres. Under some setting conditions in this study, Fi illustrates a decline with the increasing gas flow velocity. It is inferred that the decrease of Fi is due to the decrease of gas–liquid contact area, and the contact area of annular flow is deduced.

Erosion of a helium-rich layer by a steam jet in the presence of an inclined grid: Comparison of the predictions by URANS, STRUC-URANS and LES

Severe accidents in Light Water Reactors (LWRs) involve the production of hydrogen, which may accumulate in the upper regions of the containment and form flammable mixtures with the initial air inventory. To study phenomena related to light gas layer removal, the H2P1_10_2 HYMERES test (Paranjape et al., 2019, 2020) was conducted in the PANDA vessel with the aim of investigating the erosion of a helium-rich layer by a steam jet deflected by an inclined grid. In this paper, we make use of the test data to validate computational fluid dynamics (CFD) simulations of such a transient. The simulation models involve the standard k- ε URANS, the more advanced second generation STRUC- ε URANS, and a coarse Large Eddy Simulation (CLES). All simulations are conducted with the same numerical grid consisting of 3.5 million hexahedral cells. Results show that all models capture well the main physical quantities, namely: the helium erosion rate, the temperatures and the velocities downstream of the obstructing grid. Fine details such as the turbulent kinetic energy (TKE) are better reproduced by the CLES, although the latter miss somewhat the amplitudes due to lack of spatial resolution, especially in the near-field downstream of the grid region. The STRUC- ε model replicates qualitatively the flow field fluctuations in sections of the domain where the spatiotemporal resolution is adequate. Mesh refinement in the jet region shows a substantial quantitative improvement, demonstrating that the STRUC- ε model is superior to traditional URANS when spatial resolution is increased. Finally, the simulations show unmistakably that radiation modeling is necessary to achieve reasonable predictions of the temperature field, as the radiative flux can account for a substantial fraction of the heat transfer to the vessel walls, in particular when convective currents are weak.