Detached Eddy Simulation Turbulence Model Research Articles

Effects of the turbulent incoming flow condition on the wind–snow flow characteristics and accretion distribution in the cutting area of the high-speed railway

In addressing snowdrift disasters in high-speed railway cutting areas, this study utilized the improved delayed detached eddy simulation turbulence model, synthetic eddy method, and Eulerian two-phase flow model (Euler–Euler). Coupled with grid dynamic deformation techniques, the investigation facilitated extracting temporal variations in snow particle accumulation profiles within the cutting areas. The predicting accuracy of the numerical method was validated through the wind tunnel and wind–snow experimental data. Building upon this foundation, the study investigated how uniform and turbulent flow conditions impact snow distribution in the high-speed railway cutting areas. The analysis examined the turbulent flow structures, friction velocity distribution, and snow accretion contours within the cutting area. Findings indicate that the snow mass on the windward slope in the cutting area remains consistent under both flow conditions. However, there are significant differences in the rate of snow mass growth and the characteristics of snow accretion distribution. The disappearance of vortex structures at the base of the leeward slope is crucial for indicating the transition in the accumulation states of the surrounding snow particles. Under turbulent incoming flow conditions, the transition from large-scale vortex systems to vortex pairs on the leeward slope of the cutting is delayed, resulting in a postponed stable growth period for snow accretion. Additionally, strong erosion between the track bed and the leeward slope results in reduced snow accretion mass in these areas. In particular, at the simulation time of t = 150 min, under turbulent incoming flow conditions, the overall snow accumulation in the cutting decreased by 20.8% compared to the uniform incoming flow conditions. The snow mass in the leeward slope area decreased by 38.3%, and the track bed area experienced a 23.3% reduction in snow accumulation. In contrast, the snow quantity on the windward slope remained relatively consistent.

Performance, Aerodynamics, and Aeroacoustics of Side-by-Side Rotors Using High-Fidelity Computational Fluid Dynamics

This paper investigates the performance, aerodynamics, and aeroacoustics of NASA’s side-by-side urban air mobility (UAM) aircraft in hover conditions, utilizing high-fidelity computational fluid dynamics (CFD) simulations. High Performance Computing Modernization Program CREATE™-AV Helios is employed for the CFD simulations, while acoustic predictions are performed by PSU-WOPWOP. The simulations apply overset structured grids, the Spalart–Allmaras detached eddy simulation turbulence model, and adaptive mesh refinement (AMR) on the off-body mesh. The investigation covers four rotor overlap scenarios and five collective pitch angles, all maintaining a consistent rotational speed. Results indicate that as rotor overlap distance increases, a minor improvement in the figure of merit is noticed. Nevertheless, a substantial reduction in blade loading within the overlap region becomes apparent. The blade–vortex interaction (BVI) intensifies between both rotors as the overlap region broadens. The 25% overlap configuration is shown to yield the highest overall sound pressure level (OASPL) among all configurations due to the more pronounced BVIs at the entrance and exit of the overlap area. It is observed that the OASPL exceeds 62 dBA at an altitude of 500 ft (152.4 m) for all overlap cases, exceeding the suggested UAM aircraft noise guideline. Additionally, noise generated by all overlap scenarios is compared against various background noise levels. Results suggest that the noise produced by the side-by-side rotor configuration is not fully masked by any of the different background noise levels at an altitude of 500 ft (152.4 m).

Examination of air flow characteristics over an open rectangular cavity between the plates

Air flow characteristics of an open cavity have been numerically examined by using different turbulence models of Detached Eddy Simulation (DES), k-ε Realizable, k-ω Shear Stress Transport (SST) and Large Eddy Simulation (LES) based on an experimental study of open literature. Numerical results of transient analyses have been compared to experimental results at cavity length-based Reynolds number of Re = 4600. Pressure distributions, streamline patterns, streamwise velocity components, mean velocity values and their vectors have been given in terms of contour graphics. Moreover, velocity profiles have been presented. Pressure fluctuations have been triggered by flow separation and its reattachment. Due to upstream separation of boundary layer, there was curved boundary layer obtained between the outer potential-flow-like and the recirculation zones. As a result, negative velocity values are evidence for rotational flows affected by formation of secondary flows in the cavity. Furthermore, lower pressure region has been observed as a result of rotational flow which was powerful in the open cavity. Numerical results of DES and LES turbulence models are in good agreement with the results of reference study. As the numerical results obtained by LES turbulence model are approximately same with those of experimental reference study, LES turbulence model is mostly recommended. As an option to these turbulence models, k-ω SST model could be utilized for limited computer capacity. However, k-ε Realizable model is not sufficient for capturing rotational flows which are very effective in terms of present case.

Assessment of nonlinear-low Reynolds number/detached eddy simulation turbulence model for wake flow field simulation of a realistic automotive model

Since the results of wake flow simulation with commonly used turbulence models are unsatisfactory, by introducing a nonlinear Reynolds stress term and combining the DES (Detached Eddy Simulation) model, this paper further validates the nonlinear-LRN (Low Reynolds Number)/DES turbulence model which can predict the flow separation and the reattachment phenomenon more accurately. This model was verified by a wall-mounted hump flow case and was applied to the time-averaged and transient flow field structure analysis of a realistic automotive model with several widely used turbulence models. These simulation results were compared with experimental data, indicating that the nonlinear-LRN/DES model gives better agreement with the experiment and can predict the automobile wake flow structures and aerodynamic characteristics more accurately. Furthermore, the performance of the nonlinear-LRN/DES model in mesh with different refinements is compared, concluding that the new proposed model can obtain high accuracy in the coarse mesh.

Aerodynamic and aeroacoustic performance assessment of a vertical axis wind turbine by synergistic effect of blowing and suction

The power efficiency of Darrieus wind turbines significantly deteriorates during rotation caused by periodic dynamic stall at low tip speed ratios. This leads to a strong fluctuation in torque and a reduction in energy acquisition. This paper aims to improve the aerodynamic performance of an H-type Darrieus wind turbine using an innovative fluidic flow control technique based on the synergistic effect of blowing and suction. The aeroacoustic noise emissions accomplished with this enhancement are evaluated. The Improved Delayed Detached Eddy Simulation turbulence model and the Ffowcs Williams-Hawkings acoustic analogy method are adopted to simulate the instantaneous flow field and predict the far field noise. Following validation of the numerical approach using wind tunnel experimental data on a static airfoil, an orthogonal experimental design method was used to analyze and optimize the operating factors. The factors include suction position relative to leading-edge (Ls), blowing position relative to trailing-edge (Lb) and jet coefficient (Cμ). The results indicate that Cμ plays crucial role in determining the airfoil performance, while the role of Lb is almost negligible. Furthermore, the impact of optimal combination of these factors was analyzed based on three different control strategies. It is found that appropriate application of the active control solution can eliminate the wind turbine’s negative torque, avoid excessive alternating load on the rotor and improve the energy extraction efficiency. The aeroacoustic noise estimation shows that the active device can reduce the noise emission by moderating pressure fluctuation, stabilizing the flow field and influencing the vortex shedding. Similarly, the proposed active control solution can reduce the wind turbine noise level by up to 6.56 dB by modifying the sound pressure spectra at frequencies between 100 and 1000 Hz.

Comparative evaluation of DES and SAS turbulence models for incompressible flow in a Venturi tube

Computational fluid dynamics employs the Large Eddy Simulation (LES) and Reynolds-averaged Navier-Stokes (RANS) models and combines both models to perform a hybrid RANS/LES simulation of the transient state flow field, within which there are the Detached Eddy Simulation (DES) and Scale-Adaptive Simulation (SAS) models. In the present work, the objective is to evaluate the DES and SAS turbulence models for flow (water) in a Venturi tube, for 2D computational domains. The domain was discretized for grids with quadrilateral cells, and the flow field was studied for four flow rates. The results of the pressure flow field simulations for the DES and SAS models were compared with experimental data reported in the literature, which fit the experiments. However, the DES k-ε model presented a negative pressure drop for a region of the flow adjacent to the wall, at the entrance of the throat section, the other models DES S-A, DES SST k-ω and SAS presented positive pressures. The discharge coefficients yielded values in the range of 0.94-0.951, which were lower than the experiments, for errors in the range of 1.57-2.76%.

Assessing Spillway Modifications by Detached Eddy Simulation Turbulence Model

Spillway associated with flip bucket structure is a famous way to evacuate the floods at dams and dissipate the flood destructive energy in order to protect the dam structure. Several spillways suffered from damages where most affected zones were located downstream the flip bucket along the flow path towards the plunge pool. Poor geometrical design is a very important main reason of the damages occurring at the downstream of the spillway. It mainly occurred when the water jet failed to achieve safe impingement location. This study proposes solutions to damages that occurred in an existing dam. These proposed solutions were tested through numerical investigations of the existing and modified spillway configuration. This investigation was originated with the aspiration of decreasing the damage of real Spillway. From the literature, previous experimental results of spillway modelling were extracted and analyzed. The results were replicated numerically by ANSYS FLUENT software. The experimental results were contrasted against the simulated results. The contrast reflected that DES results are reasonably applicable. Confident with its results, it was applied to real spillway (Base case) and results were obtained. Firstly, the existing conditions were simulated to ensure that the model mimic the existing conditions accurately. Secondly, several modifications to the existing conditions were introduced to the model and then simulated to define the most applicable solution. Detached eddy simulation is used for accurately simulation of the flow physical model. The criteria of the modification properties are developed in the way of avoiding the existing erosion problem.

Detached eddy simulation of turbulent flow fields over steep hilly terrain

The verification and validation (V&V) of multi-fidelity turbulence models is crucial for the efficient and reliable industrial application of computational fluid dynamics (CFD). This study aims to evaluate the performance of detached eddy simulation (DES) modeling approaches embedded with various unsteady Reynolds Averaged Navier-Stokes (URANS) models in simulating the turbulent flow fields over a steep-sloped hill. And the high-fidelity numerical methodology based on the DES model was proposed and analyzed in the context of URANS modes, mesh resolution and sampling duration. In comparison with the experimental data, the shear stress transport (SST) k-ω based DDES (Delayed Detached eddy simulation) turbulence model (DDES SST k-ω) could better predict the profiles of mean velocity and turbulence fluctuation, while the Spalart-Allmaras (S-A) based DES and DDES turbulence models show its stronger capability to reproduce the spectral characteristics. Additionally, in contrast to DES S-A and DDES S-A, a larger time-averaged separation bubble was predicted by DES SST k-ω and DDES SST k-ω. Moreover, the instantaneous flow patterns such as the formation and evolution of turbulent eddies in the hill wake could be reasonably reproduced by DES S-A and DDES S-A. Furthermore, the numerical results obtained from DES models were sensitive to URANS modes and vertical mesh resolution but trivially affected by the increase in sampling time after reaching the statistical convergence.

Effect of incoming boundary layer thickness on the flow dynamics of a square finite wall-mounted cylinder

This paper investigates the flow dynamics around a square finite wall-mounted square cylinder of two different aspect ratios (AR=H/D=6 and 10, H and D are the height and width, respectively) under six boundary layer thickness (δ/D=0.5–4.0, δ is the boundary layer thickness) using the improved delayed detached eddy simulation turbulence model. The Reynolds number based on the square cylinder width and incoming velocity is Re=1.4×104. The aerodynamic force, surface mean and fluctuating pressure, wake closure length, Reynolds stresses, turbulence production, and the anisotropy of turbulence are analyzed in detail to investigate the effect of AR and δ/D combinations. The Q-criterion vortex identification method is adopted to identify and classify the vortex structures of different AR and δ/D combinations. The parametric diagram based on AR and δ/D to predict the dipole/quadrupole wake structure proposed in Yauwenas et al. [“The effect of aspect ratio on the wake structure of finite wall-mounted square cylinders,” J. Fluid Mech. 875, 929–960 (2019)] is refined based on results of the present study (quadrupole wake for AR=6 with δ/D≥1.5 and AR=10 with δ/D≥1.0). The instantaneous space–time pressure distribution along the whole span proves the existence of cellular shedding (AR=10 for δ/D=0.5–2.0), and the spectral proper orthogonal decomposition technique is adopted to investigate the cellular shedding phenomenon and its distribution. With the increase in δ/D from 2.0 to 4.0, the cellular shedding process is absent, implying the inhibitory effect of increasing boundary layer thickness on the cellular shedding.

Study on the effect of dimple position on drag reduction of high-speed maglev train

Abstract Transient numerical simulations were carried out by placing dimples at the top, sides and bottoms of the tail car streamline area of a high-speed maglev train. The results of an improved delayed detached eddy simulation turbulence model using three-dimensional compressible Navier-Stokes and shear-stress transport K-Omega double equations were compared to the results of a wind tunnel test to verify the numerical simulation accuracy, within 5% of the ground truth, which is an acceptable precision range. The results show that dimples arranged on the streamline area atop the train tail car affected the locations at which the airflow at the top and bottom of the train met and weakened the strength of the wake. The aerodynamic drag and lift coefficient decreased by 3.40% and 4.27%, respectively. When the dimples were arranged on the streamline area at the sides or bottoms of the train tail car, they had little effect on the top of the tail car, so they did not destroy the balance of the airflow at the top and bottom. They also had little influence on the development of wake topology. Therefore, the aerodynamic drag and lift of the train changed little.

Numerical analysis for wake flow field of Ahmed model based on a nonlinear-LRN/DES turbulence model

PurposeThis paper aims to propose a precise turbulence model for automobile aerodynamics simulation, which can predict flow separation and reattachment phenomena more accurately.Design/methodology/approachAs the results of wake flow simulation with commonly used turbulence models are unsatisfactory, by introducing a nonlinear Reynolds stress term and combining the detached Eddy simulation (DES) model, this paper proposes a nonlinear-low-Reynolds number (LRN)/DES turbulence model. The turbulence model is verified in a backward-facing step case and applied in the flow field analysis of the Ahmed model. Several widely applied turbulence models are compared with the nonlinear-LRN/DES model and the experimental data of the above cases.FindingsCompared with the experimental data and several turbulence models, the nonlinear-LRN/DES model gives better agreement with the experiment and can predict the automobile wake flow structures and aerodynamic characteristics more accurately.Research limitations/implicationsThe nonlinear-LRN/DES model proposed in this paper suffers from separation delays when simulating the separation flows above the rear slant of the Ahmed body. Therefore, more factors need to be considered to further improve the accuracy of the model.Practical implicationsThis paper proposes a turbulence model that can more accurately simulate the wake flow field structure of automobiles, which is valuable for improving the calculation accuracy of the aerodynamic characteristics of automobiles.Originality/valueBased on the nonlinear eddy viscosity method and the scale resolved simulation, a nonlinear-LRN/DES turbulence model including the nonlinear Reynolds stress terms for separation and reattachment prediction, as well as the wake vortex structure prediction is first proposed.

Numerical and experimental research on the erosion of solid-liquid two-phase flow in transport butterfly valve based on DEM method

PurposeThe purpose of this paper is to analyze the mechanism of particle erosion in butterfly valve pipelines under hydraulic transportation conditions. The results will affect the sealing and safety of butterfly valve pipelines and hopefully serve as reference for the anti-erosion design of butterfly valve pipelines.Design/methodology/approachThrough the discrete element method (DEM) simulation that considers the force between particles, the detached eddy simulation (DES) turbulence model based on realizable k-epsilon is used to simulate the solid-liquid two-phase flow-induced erosion condition when the butterfly valve is fully opened. The simulation is verified by building an experimental system correctness. The solid-liquid two-phase flow characteristics, particle distribution and erosion characteristics of the butterfly valve pipeline under transportation conditions are studied.FindingsThe addition of particles may enhance the high-speed area behind the valve. It first increases and then decreases with increasing particle size. With increasing particle size, the low-velocity particles change from being uniformly distributed in flow channel to first gathering in the front of the valve and, then, to gathering in lower part of it. Fluid stagnation at the left arc-shaped flange leads to the appearance of two high-speed belts in the channel. With increasing fluid velocity, high-speed belts gradually cover the entire valve surface by focusing on the upper and lower ends, resulting in the overall aggravation of erosion.Originality/valueConsidering the complexity of solid-liquid two-phase flow, this is the first time that the DEM method with added inter-particle forces and the DES turbulence model based on realizable k-epsilon has been used to study the flow characteristics and erosion mechanism of butterfly valves under fully open transportation conditions.

Computational aeroacoustic modeling using hybrid Reynolds averaged Navier–Stokes/large‐eddy simulations methods with modified acoustic analogies

AbstractThis study employs a numerical approach to investigate noise mechanisms in tandem cylinders. The tandem cylinder configuration is an excellent model for aircraft landing gear. The study is therefore primarily motivated by the need to understand the aircraft landing gear as a primary contributor to airframe noise during approach and landing. Fluctuations in the flow properties induced by turbulent flow around the tandem cylinders are computed and analyzed. A hybrid improved delayed detached eddy simulation turbulence model is employed to compute the boundary layer flow as well as to compute the fluctuations in the flow properties. The numerical methodologies for the turbulent flow computations are implemented on the OpenFOAM software platform. The study also considers the impact of neglecting the volume sources and employs two modified versions of Curle's and Ffowcs‐Williams and Hawkings analogies. The results are in good agreement with published numerical and experimental data in the existing literature.

Aerodynamic response of high-speed trains under crosswind in a bridge-tunnel section with or without a wind barrier

High-speed trains experience a sharp transition to a strong crosswind environment when running in a bridge–tunnel section due to the perennial prevailing crosswind in the canyon, and this sudden transition seriously affects train safety. In this study, a 3D computational fluid dynamics numerical model of the train–tunnel–bridge–wind barrier is established based on the delayed detached eddy simulation turbulence model and porous media theory. A dynamic analysis model of wind–train–bridge coupling is adopted. The effects of wind barrier with a height of 3 ​m and porosity of 30% on the aerodynamic coefficient, flow field structure and running safety of high-speed trains under crosswind in the bridge–tunnel section are studied. Results indicate that the sharp change effect of the aerodynamic coefficient is significantly weakened by more than 50% by the wind barrier. The aerodynamic fluctuation amplitudes in the bridge–tunnel section are 1.25–5.5 times higher than those in the bridge section. The difference in pressure distribution in the longitudinal direction is significantly reduced because of the obstruction and diversion of the wind barrier and the space limitation on the windward side. Accordingly, the change in amplitude of the aerodynamic coefficients in the bridge–tunnel section is reduced, and so is the safety of train operation. The bridge–tunnel section is the weak link of safety control. Using a wind barrier with the same parameters for the bridge and bridge–tunnel sections is unreasonable, that is, the parameters should be separately designed.

Dynamic Analysis of Cavitation Tip Vortex of Pump-Jet Propeller Based on DES

When a pump-jet propeller rotates at high speeds, a tip vortex is usually generated in the tip clearance region. This vortex interacts with the main channel fluid flow leading to the main energy loss of the rotor system. Moreover, operating at a high rotational speed can cause cavitation near the blades which may jeopardize the propulsion efficiency and induce noise. In order to effectively improve the propulsion efficiency of the pump-jet propeller, it is mandatory to research more about the energy loss mechanism in the tip clearance area. Due to the complex turbulence characteristics of the blade tip vortex, the widely used Reynolds averaged Navier–Stokes (RANS) method may not be able to accurately predict the multi-scale turbulent flow in the tip clearance. In this paper, an unsteady numerical simulation was conducted on the three-dimensional full flow field of a pump-jet propeller based on the DES (detached-eddy-simulation) turbulence model and the Z-G-B (Zwart–Gerber–Belamri) cavitation model. The simulation yielded the vortex shape and dynamic characteristics of the vortex core and the surrounding flow field in the tip clearance area. After cavitation occurred, the influence of cavitation bubbles on tip vortices was also studied. The results revealed two kinds of vortices in the tip clearance area, namely tip leakage vortex (TLV) and tip separation vortex (TSV). Slight cavitation at J = 1.02 led to low-frequency and high-frequency pulsation in the TLV vortex core. This occurrence of cavitation promotes the expansion and contraction of the tip vortex. Further, when the advance ratio changes into J = 0.73, a third type of vortex located between TLV and TSV appeared at the trailing edge which runs through the entire rotational cycle. This study has presented the dynamic characteristics of tip vortex including the relationship between cavitation bubbles and TLV inside the pump-jet propeller, which may provide a reference for the optimal design of future pump-jet propellers.

Effect of Shaft Diameter on the Hydrodynamic Torque of Butterfly Valve Disk.

In this study, computational fluid dynamics (cfd) software and detached eddy simulation turbulence model were used to simulate butterfly valves with different designs. The effects of shaft diameters on the value and the fluctuation of valve disk torque were studied, and the physical reason was discussed. The simulation results were verified by comparing with the experimental data. The findings revealed that with the closing of the valve, the hydraulic torque of the valve disk first increases and then decreases. Meanwhile, the torque decreases gradually with the increase of the shaft diameter. The variation of torque is caused by the change of pressure on both sides of the valve disk. The result also indicates that the fluctuation of torque is induced by the flow separation phenomenon occurs on the valve disk. The fluctuation is significant for the valve opening from 20% to 60%. The strength of the torque fluctuation is greater for the smaller shaft diameter. This study provides a theoretical basis for the design and optimization of butterfly valves.

Numerical Simulation of Haze-Fog Particle Dispersion in the Typical Urban Community by Using Discrete Phase Model

The haze-fog particle dispersion in urban communities will cause serious health and environmental problems, which has aroused society attention. The aim of the present investigation is to reveal the underlying mechanisms of haze-fog particle dispersion via Computational Fluid Dynamics (CFD) method, and then to provide a groundwork for the optimal spatial arrangement of urban architecture. The Delayed Detached-eddy Simulation turbulence model (DDES) and Discrete Phase Model (DPM) are utilized to investigate the wind flow distribution and the particle dispersion around the building group. The numerical results show that the particle dispersion is dominated by the incoming wind flow, the layout of architectural space and the type and distribution of vortex. The ‘single body’ wake pattern and the vortex impingement wake pattern are identified in the wind flow field, which have different effects on the distribution of haze-fog particle. The cavity formed by the layout of the building group induces primary vortex and secondary vortex, which will make it more difficult for the particles entering the square cavity to flow out. Moreover, the concentration of the particle in the rear of the buildings is relatively low due the effect of attached vortices.

Turbulence modeling approaches on unsteady flow structures around a semi-circular cylinder

This paper presents the flow field around a Semi-Circular Cylinder (SCC), a basic geometry for fluid mechanics, obtained numerically by Computational Fluid Dynamics (CFD) at Reynolds number of 3.2 × 104. Time-averaged flow characteristics such as vorticity, velocity fields with streamline patterns, Reynolds stress correlations, Turbulence Kinetic Energy (TKE), pressure and drag coefficients have been examined by using Reynolds-averaged Navier-Stokes (RANS), Detached Eddy Simulation (DES) and Large Eddy Simulations (LES) turbulence models. Several RANS approaches with turbulence modeling including SST k-ω, k-ε and their derivatives were used, while in the LES simulations, the dynamic Smagorinsky subgrid-scale model was employed. The obtained results of the present study have been compared with an experimental study from the open literature. The results of the investigation showed that when numerical findings are compared to experimental ones, the results of LES and DES turbulence models were found to be more consistent than those obtained by using k-ε and SST k-ω models. Time-averaged flow patterns have been found to be considerably symmetric with respect to the axis passing through the end of the curved surface of the SCC. The deviation of LES and DES results, and experimental results was less than 7% in terms of the drag coefficient.

Numerical study on the transformation of an internal solitary wave propagating across a vertical cylinder

A finite volume method based 3D Cartesian grids is developed to study the effect of cylinder diameter on the evolution of a depression internal solitary wave (ISW) with different initial wave amplitudes across a vertical cylinder. This method solves the Navier–Stokes equations using an Improved Delayed Detached Eddy Simulation turbulence model. Numerical results reveal that the significant symmetric vortex shedding on the front and rear side of the cylinder during the wave-obstacle interaction, especially at the large diameter. Moreover, the vortex shedding affects the variations of the horizontal force on different depths. Results also show transmitted wave amplitude and total net horizontal force increase with initial wave amplitude and the diameter of the cylinder, in which the former is less significant than the latter, while the converse is true for the run-up height along the cylinder. In addition, the total horizontal forces on the cylinder reverse direction from compression to tension in the upper layer and conversely it is also true in the lower layer during the propagation of a depression ISW.

Numerical study of the stratification erosion benchmark for NPPs containment using CFD code GASFLOW-MPI

During a severe accident involving a Nuclear Power Plant (NPP), a hydrogen-enriched stratification may occur in the dome of the containment. This is due to the gaseous density difference of a steam/hydrogen turbulent jet released from the break. This accident scenario is one of the high-ranking phenomena identified in the Phenomena Identification and Ranking Table (PIRT) for containment safety thermal hydraulics. In this work, the temporal evolution of the stratification erosion is analysed using the parallel CFD code GASFLOW-MPI with comparison to the experimental data from the THAI TH-20 benchmark, where helium is used as a hydrogen simulate. Identifying a suitable turbulence model is an important issue for a successful stratification erosion simulation due to turbulent mixing between the stratification and turbulent jet featured with the sharp velocity and density gradients. Three well-known turbulence models: 1. The classic k-ε model, 2. Large Eddy Simulation (LES) model and 3. Detached Eddy Simulation (DES) model are employed to evaluate their performances. Due to the negative buoyancy effect resulting from the light gas layer, the jet gradually loses its initial momentum and cannot penetrate the stratified layer in depth. The interaction Froude (Fr) number which is used to categorize the mixing mechanism is close to 1 in this work, indicating that the erosion process of the stratified layer is dominated by both the molecular diffusion and the momentum transport. The experimental data measured at different heights in THAI TH-20 benchmark, from the pure air region at the lower containment region up into the large helium concentration gradient region, and further up to high helium concentration region at the top of the containment. The calculated results show that the LES and DES turbulence models can capture time histories of helium concentration at all of the sensor points accurately, while the k-ε turbulence model predictions are always delayed due to its overestimating the turbulence mixing in the stratified layer. Moreover, compared with the k-ε model, more detailed turbulent eddy structure is resolved by the LES model and DES model, whose frequency satisfies the negative −5/3 energy decay rate for a wide region of frequency. Therefore, the LES and DES models are recommended for the stratification erosion simulation. This can provide guidance for nuclear engineers analysing hydrogen distribution phenomena during severe accidents.