مطالعه شبیه سازی بزرگ اختلاط آشفته در اتصال تی

A potential cause of thermal fatigue failures in energy cooling systems is identified with cyclic stresses imposed on a piping system. These are generated due to temperature changes in regions where cold and hot flows are intensively mixed together. A typical situation for such mixing appears in turbulent flow through a T-junction, which is investigated here using Large-Eddy Simulations (LES). In general, LES is well capable in capturing the mixing phenomena and accompanied turbulent flow fluctuations in a T-junction. An assessment of the accuracy of LES predictions is made for the applied Vreman subgrid-scale model through a direct comparison with the available experimental results. In particular, an estimation of the minimal mesh-resolution requirements for LES is examined on the basis of the complementary RANS simulations. This estimation is based on the characteristics turbulent scales (e.g., Taylor micro-scale) that can be computed from LES or RANS simulations.

Development and validation of modelling approaches for turbulent mixing is an important issue implicitly connected with the nuclear reactor safety. Turbulent mixing in reactor cooling systems can potentially lead to appearance of the thermal fatigue phenomena (Chapuliot et al., 2005). Detailed information about the amplitudes and frequencies of the flow temperature fluctuations in the pipelines is highly desired in order to prevent possible damages. Usual strategy applied to thermal fatigue predictions involves Computational Fluid Dynamics simulations to determine the temperature fluctuations, which serve as an input for the structural mechanics analyses. Turbulent mixing of hot and cold fluid streams in a T-junction is a challenging test case for validation of applied numerical methods. At high Reynolds number, accurate flow predictions require considerably large computational effort due to the amount of various flow-scales that need to be numerically resolved. In the available literature, a number of numerical experiments in T-junctions can be found (see Walker et al., 2009, and references therein). Simulations with five different physical conditions and two configurations of T-junctions were presented by Hu and Kazimi (2006). This benchmark study of high cycle temperature fluctuations showed the applicability of LES in prediction of turbulent flow features in a T-junction. Similar conclusions were drawn by Coste et al. (2006), where an influence of secondary flow on temperature averages and fluctuations is demonstrated. The secondary flow was imposed by an additional elbow attached to one of the T-junction legs. Simulations with the standard and dynamic Smagorinsky models were compared by Merzari and Ninokata (2007). An advantage of the dynamic model over the Smagorinsky method was shown as the dynamic procedure better resolves the small-scale turbulence flow features. The utility of both models application for predictions of the mean flow features was demonstrated there. Optimal operating conditions for a flow in a mixing-T were studied by Hosseini et al. (2008). Various jet rising mechanisms as a result of the T-junction geometry were examined there. An extensive literature survey that covers joint US–Japanese and European research programmes on this subject was prepared by Walker et al. (2009). Although thermal fatigue topic obtained escalating attention in the recent years, detailed validation of applied numerical methods is still needed in order to determine their accuracy and range of application. In the previous work (Kuczaj and Komen, 2010), we investigated accuracy of the flow predictions coming from two LES models (Vreman, 2004 and Smagorinsky, 1963) applied for modelling of turbulent mixing in a T-junction. An engineering estimation of the required computational mesh resolution based on the ‘a posteriori’ computed Taylor micro-scale length was provided. In this paper, we concentrate on two issues. First we try to answer the question whether ‘a priori’ estimated turbulence length scales from RANS simulations can be useful for predictions of the required LES mesh resolution. Other estimation approaches for the required grid resolution will be also briefly discussed. Afterwards, flow predictions obtained with the applied LES model on various mesh resolutions will be analyzed. Comparison with the experimental data (Andersson et al., 2006) provides a conclusive opportunity for assessing suitability of the obtained results in prediction of turbulent mixing. We pay special attention to the resolution that needs to be used for accurate flow predictions. We will show that large differences occur in predictions of turbulence quantities (mean and fluctuations) that are obtained on the basis of the same LES model but with under-resolved mesh density. The organization of this paper is as follows. In Section 2 we introduce the experimental and computational setup. Minimal LES mesh-resolution requirements are examined with complementary RANS simulations in Section 3. Section 4 is devoted to verification of the applicability of applied modelling methods and their numerical validation through a direct comparison of results with the experimental data. Concluding remarks are collected in Section 5.

In this paper, we studied applicability of the Large-Eddy Simulations for thermal fatigue prediction purposes. Recently developed SGS model (Vreman, 2004) was applied to modelling of the thermal mixing in a T-junction. Results obtained with this eddy-viscosity subgrid-scale model were compared with the experimental data. For the considered cases, we found that in order to obtain numerical solutions close to the experimental findings, the required mesh resolution must resolve the Taylor micro-scale length (Δ ∼ λ/3) or should be of the order of Taylor micro-scale obtained from the RANS simulations (Δ ∼ λR). Based on this engineering estimation, we can safely conclude that meeting this requirement allows estimating flow characteristics with sufficient accuracy for the practical applications downstream of the mixing zone. Our LES results show that finer meshes must be used in order to accurate capture fluctuations in the shear layer close to the center of the mixing zone.