Search results

The Navier‐Stokes equation and the species continuity equation have been solved numerically in a boundary fitted coordinate system comprising the geometry of a single strand bare tundish. The solution of the species continuity equation predicts the time evolution of the concentration of a tracer at the outlet of the tundish. The numerical prediction of the tracer concentration has been made with nine different turbulence models and has been compared with the experimental observation for the tundish. It has been found that the prediction from the standard k‐ε model, the k‐ε Chen‐Kim (ck) and the standard k‐ε with Yap correction (k‐ε Yap), matches well with that of the experiment compared to the other turbulence models as far as gross quantities like the mean residence time and the ratio of mixed to dead volume are concerned. It has been found that the initial transient development of the tracer concentration is best predicted by the low Reynolds number Lam‐Bremhorst model (LB model) and then by the k‐ε RNG model, while these two models under predict the mean residence time as well as the ratio of mixed to dead volume. The Chen‐Kim low Reynolds number (CK low Re) model (with and without Yap correction) as well as the constant effective viscosity model over predict the mixing parameters, i.e. the mean residence time and the ratio of mixed to dead volume. Taking the solution of the k‐ε model as a starting guess for the large eddy simulation (LES), a solution for the LES could be arrived after adopting a local refinement of the cells twice so that the near wall y⁺ could be set lower than 1. Such a refined grid gave a time‐independent solution for the LES which was used to solve the species continuity equation. The LES solution slightly over predicted the mean residence time but could predict fairly well the mixed volume. However, the LES could not predict both the peaks in the tracer concentration like the k‐ε, RNG and the Lam‐Bremhorst model. An analysis of the tracer concentration on the bottom plane of the tundish could help to understand the presence of plug and mixed flow in it.

The purpose of this paper is to simulate the behaviour of the symmetrical turn‐up vortex amplifier (STuVA) to obtain insight into its maximum through‐flow operation within the eight‐port STuVA, and understand the relation between its design parameters and flow characteristics. Furthermore, it is to test the performance of different turbulent models and near‐wall models using the same grid, the same numerical methods and the same computational fluid dynamics code under multiple impingement conditions.

Three turbulence models, the standard k‐ε, the renormalization group (RNG) k‐ε model and the Reynolds stress model (RSM), and three near‐wall models have been used to simulate the confined incompressible turbulent flow in an eight‐port STuVA using unstructured meshes. The STuVA is a special symmetrical design of turn‐up vortex amplifier, and the simulation focused on its extreme operation in the maximum flow state with no swirling. The predictions were compared with basic pressure‐drop flow rate measurements made using air at ambient conditions. The effect of different combinations of turbulence and near‐wall models was evaluated.

The RSM gave predictions slightly closer to the experimental data than the other models, although the RNG k‐ε model predicted nearly as accurately as the RSM. They both improved errors by about 3 per cent compared to the standard k‐ε model but took a long time for convergence. The modelling of complex flows depends not only on the turbulence model but also on the near‐wall treatments and computational economy. In this study a good combination was the RSM, the two layer wall model and the higher order discretization scheme, which improved accuracy by more than 10 per cent compared to the standard k‐ε model, the standard wall function and first order upwind.

The results of this paper are valid for the global pressure drop flow rate. It should be desirable to compare some local information with the experiment.

This paper provides insight into the maximum through‐flow operation within the eight‐port STuVA to understand the relation between its design parameters and flow characteristics and study the performance of turbulence and near wall models.

This paper aims to predict turbulent flow and heat transfer through different channels with periodic dimple/protrusion walls. More specifically, the performance of various low-Re k-ε turbulence models in prediction of local heat transfer coefficient is evaluated.

Three low-Re number k-ε turbulence models (the zonal k-ε, the linear k-ε and the nonlinear k-ε) are used. Computations are performed for three geometries, namely, a channel with a single dimpled wall, a channel with double dimpled walls and a channel with a single dimple/protrusion wall. The predictions are obtained using an in house finite volume code.

The numerical predictions indicate that the nonlinear k-ε model predicts a larger recirculation bubble inside the dimple with stronger impingement and upwash flow than the zonal and linear k-ε models. The heat transfer results show that the zonal k-ε model returns weak thermal predictions in all test cases in comparison to other turbulence models. Use of the linear k-ε model leads to improvement in heat transfer predictions inside the dimples and their back rim. However, the most accurate thermal predictions are obtained via the nonlinear k-ε model. As expected, the replacement of the algebraic length-scale correction term with the differential version improves the heat transfer predictions of both linear and nonlinear k-ε models.

The most reliable turbulence model of the current study (i.e. nonlinear k-ε model) may be used for design and optimization of various thermal systems using dimples for heat transfer enhancement (e.g. heat exchangers and internal cooling system of gas turbine blades).

The Navier‐Stokes equation and the species continuity equation have been solved numerically in a boundary fitted coordinate system comprising the geometry of a large scale industrial size tundish. The solution of the species continuity equation predicts the time evolution of the concentration of a tracer at the outlets of a six strand billet caster tundish. The numerical prediction of the tracer concentration has been made with six different turbulence models (the standard k‐ε, the k‐ε RNG, the Low Re number Lam‐Bremhorst model, the Chen‐Kim high Re number model (CK), the Chen‐Kim low Re number model (CKL) and the simplest constant effective viscosity model (CEV)) which favorably compares with that of the experimental observation for a single strand bare tundish. It has been found that the overall comparison of the k‐ε model, the RNG, the Lam‐Bremhorst and the CK model is much better than the CKL model and the CEV model as far as gross quantities like the mean residence time and the ratio of mixed to dead volume are concerned. However, the k‐ε model predicts the closest value to the experimental observation compared to all other models. The prediction of the transient behavior of the tracer is best done by the Lam‐Bremhorst model and then by the RNG model, but these models do not predict the gross quantities that accurately like the k‐ε model for a single strand bare tundish. With the help of the above six turbulence models mixing parameters such as the ratio of mix to dead volume and the mean residence time were computed for the six strand tundish for different outlet positions, height of advanced pouring box (APB) and shroud immersion depth. It was found that three turbulence models show a peak value in the ratio of mix to dead volume when the outlets were placed at 200 mm away from the wall. An APB was put on the bottom of the tundish surrounding the inlet jet when the outlets were kept at 200 mm away from the wall. It was also found that there exists an optimum height of the APB where the ratio of mix to dead volume and the mean residence time attain further peak values signifying better mixing in the tundish. At this optimum height of the APB, the shroud immersion depth was made to change from 0 to 400 mm. It was also observed that there exists an optimum immersion depth of the shroud where the ratio of mix to dead volume still attains another peak signifying still better mixing. However, all the turbulence models do not predict the same optimum height of the APB and the same shroud immersion depth as the optimum depth. The optimum height of the APB and the shroud immersion depth were decided when two or more turbulence models predict the same values.

The Navier‐Stokes equation and the species continuity equation have been solved numerically in a boundary fitted coordinate system comprising the geometry of a large scale industrial size tundish. The solution of the species continuity equation predicts the time evolution of the concentration of a tracer at the outlet of a single strand bare tundish. The numerical prediction of the tracer concentration has been made with three different turbulence models; (a standard k‐ε, a k‐ε RNG and a Low Re number Lam‐Bremhorst model) which favorably compares with that of the experimental observation for a single strand bare tundish. It has been found that the overall comparison of k‐ε model with that of the experiment is better than the other two turbulence models as far as gross quantities like mean residence time and ratio of mixed to dead volume are concerned. However, it has been found that the initial transient development of the tracer concentration is best predicted by the Lam‐Bremhorst model and then by the RNG model. The k‐ε model predicts the tracer concentration much better than the other two models after the initial transience (t>40 per cent of mean residence time) and the RNG model lies in between the k‐ε and the Lam‐Bremhorst one. The numerical study has been extended to a multi strand tundish (having 6 outlets) where the effect of outlet positions on the ratio of mix to dead volume has been studied with the help of the above three turbulence models. It has been found that all the three turbulence models show a peak value for the ratio of mix to dead volume (a mixing parameter) when the outlets are placed 200 mm away from the wall (position‐2) thus signifying an optimum location for the outlets to get highest mixing in a given multi strand tundish.

The purpose of this paper is to evaluate the capacity of two equation turbulence models to reproduce mean and fluctuating quantities in the case of both natural convection and isothermal flows.

Numerical predictions of mean velocity profiles, air and wall temperatures as well as turbulent kinetic energy by three different two equation models (standard k‐ε, renormalisation group k‐ε and shear‐stress transport‐k‐ω) are compared with corresponding experimental values.

The prediction of mean velocities and temperatures is in all cases satisfactory. On the other hand, the prediction of turbulent quantities is less precise.

The three models under consideration in this paper can be used for engineering applications such as HVAC calculations.

The flow inside an axisymmetric diffuser with a curved surface centre body is numerically simulated using different turbulence models, namely a high‐Reynolds number k‐ε in conjunction with wall function turbulence model, a high‐Reynolds number k‐ε with one‐equation turbulence model, a low‐Reynolds number k‐ε turbulence model, a RNG turbulence model and an anisotropic turbulence model. For the separation and reattachment positions, the comparisons made between the various numerical predictions and experimental measurements show that the high‐Reynolds number k‐ε with one‐equation turbulence model is superior to other models in the present study.

Heat transfer due to natural convection inside a closed cavity must be modeled to include the effects of turbulence if the Rayleigh number is sufficiently large. This study assesses the performance of several commonly used numerical turbulence models such as k‐ε, Renormalized Group k‐ε and Reynolds stress model, in predicting heat transfer due to natural convection inside an air‐filled cubic cavity. The cavity is maintained at 307 K on one side and 300 K on the opposite side with a linear temperature variation between these values on the remaining walls. Two cases are considered, one in which the heated side is vertical, and the other in which it is inclined at 45° from the horizontal. Rayleigh numbers of 10⁷, 10⁸, 10⁹ and 10¹⁰ are considered. Results of the three turbulence models are compared to experimentally determined values or values from correlations. It was found that the standard k‐ε model was the most effective model in terms of accuracy and computational economy.

Seeks to examine the performance of conventional turbulence models modelling strongly swirling flows within a Symmetrical Turn up Vortex Amplifier, with adjustment of the turbulence model constants to improve agreement with experimental data.

First, the standard k‐ε model and the Reynolds Stress Model (RSM) were used with standard values of model constants, using both the first order upwind and the quadratic upstream interpolation for convective kinetics (QUICK) schemes. Then, the swirling effect was corrected by adjusting the model coefficients.

The standard RSM with the QUICK did produce better predictions but still significantly overestimated the experimental data. Much improved simulation was obtained with the systematic adjustment of the model constants in the standard k‐ε model using the QUICK. The physical significance of the model constants accounted for changes of the eddy viscosity, and the production and destruction of k and ε.

More industrial cases could benefit from this simple and useful approach.

The constant adjustment is regular and directed, based on the eddy viscosity and the production and destruction of k and ε. The regularity of the effect of the model constants on the solutions makes it easier to quickly adjust them for other industrial applications.

In this paper a numerical simulation, based on finite difference techniques, has been developed in order to analyse turbulent natural and mixed convection of air in internal flows. The study has been restricted to two‐dimensional cavities with the possibility of inlet and outlet ports, and with internal heat sources. Turbulence is modelled by means of two‐equation k‐ε turbulence models, both in the simplest form using wall functions and in the more general form of low‐Reynolds‐number k‐ε models. The couple time average governing equations (continuity, momentum, energy, and turbulence quantities) are solved in a segregated manner using the SIMPLEX method. An implicit control volume formulation of the differential equations has been employed. Some illustrative numerical results are presented to study the influence of geometry and boundary conditions in cavities. A comparison of different k‐ε turbulence models has also been presented.