Numerical Simulation of the Vortical Structures in a Lobed Jet Mixing Flow

Numerical Simulation of the Vortical Structures in a Lobed Jet Mixing Flow Nathan J. Cooper * and Parviz Merati Department of Mechanical and Aeronautical Engineering, Western Michigan University, Kalamazoo, MI 49008 and Hui Hu Department of Aerospace Engineering, Iowa State University, Ames, IA 50011-2271 Numerical simulations of an incompressible jet mixing flow exhausted from a circular lobed mixer/nozzle were presented and validated against whole-field quantitative Dual-Plane Stereoscopic PIV (DP-SPIV) measurement results of the same flow. The numerical simulations were conducted using a Reynolds Averaged Navier-Stokes approach with a modest number of unstructured tetrahedral cells and four widely used turbulence models to predict the vortex structures in a lobed jet mixing flow. The predictability of the turbulence models to the lobed jet mixing flow were assessed and quantified based on quantitative comparisons of the numerical predictions with the DP-SPIV measurement results. It is found that the numerical simulations agree with the experimental measurements reasonably well in terms of streamwise vorticity and azimuthal (spanwise) vorticity. Although all the four turbulent models investigated over predict the magnitude of the turbulent kinetic energy significantly (about 50% to 130% over predicted), the numerical simulations are found to agree with the experimental results in predicting the locations of the regions with high turbulent kinetic energy, and the trends associated turbulent kinetic energy production and decay with downstream distance. It is found that the k-ε Realizable turbulence model provides the most accurate prediction of the lobed jet mixing flow among the turbulence models compared. Nomenclature D = inner diameter of lobed nozzle, D=40 mm H = lobe height, H=15 mm 1 k = turbulent kinetic energy, k = ( u i u i ) 2 2 U 0 ρ U ( x, y, z) k( x, y, z) dydz K(x) = averaged turbulent kinetic energy, K( x) = ρ U ( x, y, z) dydz Re D = Reynolds number from nozzle inner diameter, D U, V, W = mean velocity components, m/s U 0 = mass-flow averaged axial velocity at nozzle exit-plane, U 0 = 20 m/s x, y, z = Cartesian coordinate system µ = absolute viscosity, 1.789 10-5 kg/m-s ρ = density of air, 1.225 kg/ m 3 * Graduate Research Assistant, Department of Mechanical and Aeronautical Engineering, AIAA Student Member; Email: nate.cooper@wmich.edu. Professor and Chair, Department of Mechanical and Aeronautical Engineering. Assistant Professor, Department of Aerospace Engineering, Corresponding Author; Email: huhui@iastate.edu. 1

θ in = average inward penetration angle, 14 θ out = average outward penetration angle, 25 ω x D w v = non-dimensional x or streamwise vorticity, ω x = U 0 y z ω y D u w = non-dimensional y vorticity, ω y = U 0 z x ω z D v u = non-dimensional z vorticity, ω z = U 0 x y ω in-plane = non-dimensional azimuthal (spanwise) vorticity, in plane 2 y ω = ω + ω 2 z T I. Introduction HERE are numerous aerospace applications where methods to enhance fluid mixing between co-flowing streams have been identified as critical or enabling technologies. Examples include low emissions combustors, ejectors for high lift or jet noise reduction, infrared suppressor nozzles, turbofan core-bypass mixers and supersonic combustion ramjets. It is widely recognized that a powerful mechanism for mixing enhancement is through introduction of strong streamwise vortices between the co-flowing streams. Lobed mixers/nozzles, which are essentially splitter plates with corrugated trailing edges, provide an extraordinary means for efficiently generating strong streamwise vortices and associated cross-stream circulation in co-flowing streams. Due to their outstanding performance of mixing enhancement, lobed mixers/nozzles have been given great attention by many researchers in past years. Paynter & Birch 1 investigated the benefit of incorporating a lobed nozzle/mixer into a turbofan exhaust system to decrease jet noise. They found that lobed nozzles/mixers not only decreased jet noise but also increased thrust output. Further detailed investigations for optimizing the geometry of lobed nozzles/mixers for bulk mixing enhancement and performance improvement of turbofan exhaust systems were conducted through the NASA Energy Efficient Engine (E 3 ) program 2 initiated in 1978. Additional mixing benefits were gained by further modifications to the lobed trailing edge, which include notches or scallops in the lobed trailing edge, cutback lobes, increased penetration and lobe number. Besides the continuous efforts to optimize the geometry of lobed mixers/nozzles for better mixing enhancement performance and widen the applications of lobed mixers/nozzles, extensive studies about the mechanism by which the lobed mixers/nozzles substantially enhance mixing have also been conducted in past years. Based on pressure, temperature, and velocity measurements of the flow field downstream of a lobed mixer/nozzle, Paterson 3, 4 revealed the existence of large-scale streamwise vortices in lobed mixing flows induced by the special geometry of the lobed nozzle. Large-scale streamwise vortices were suggested to be responsible for the enhanced mixing. Werle et al. 5 and Eckerle et al. 6 found that the streamwise vortices in lobed mixing flows follow a three-step process by which the streamwise vortices form, intensify, and then break down, and suggested that the high turbulence resulting from the vortex breakdown improved the overall mixing process. Elliott et al. 7 suggested that there are three primary contributors to the mixing processes in lobed mixing flows. The first is the azimuthal (spanwise) vortices, which occur in any free shear layers due to the Kelvin-Helmholtz instability. The second is the increased interfacial contact area due to the convoluted trailing edge of the lobed nozzle/mixer. The last element is the streamwise vortices produced by the special geometry of the lobed nozzle/mixer. Based on pulsed laser sheet flow visualization with smoke and three-dimensional velocity measurements with hot film anemometer, McCormick and Bennette 8 suggested that it is the interaction of the azimuthal (spanwise) Kelvin-Helmholtz vortices with the streamwise vortices that produce the high levels of mixing. Although unsteady vortical and turbulent structures in lobed mixing flows have been suggested to play important roles for the mixing enhancement, the detailed quantitative experimental data like whole-field velocity and vorticity distributions to reveal the evolution and interaction characteristics of the unsteady vortical and turbulent structures have not become available until recent years. By using modern whole-field flow diagnostic techniques like Planar Laser Induced Fluorescence (PLIF), Particle Image Velocimetry (PIV), Stereoscopic Particle Image Velocimetry (SPIV) and Dual-plane Stereoscopic PIV techniques, Hu et al. 9-12 conducted a series of studies to address the evolution and interactive characteristics of various vortical and turbulent structures in lobed jet mixing flows instantaneously and quantitatively. 2

There is a great deal of interest in accurately simulating lobed mixing flows. One obvious reason for doing so is that designing an optimum mixer/nozzle by experimental evidence alone can be expensive and nontrivial due to the wide range of mixer parameters involved. By exploiting computational approaches, some of these costs can be reduced dramatically. Furthermore, numerical simulations can provide detailed information of fluid quantities that are often difficult or impossible to measure experimentally. Such quantities can be helpful to identify the underlying physics and further our understanding of lobed mixer flows. Several computational studies of lobed mixing flows have been conducted by solving the Reynolds-Averaged Navier Stokes (RANS) equations. The previous studies include the work of Kotumos and McGuick 13, Tsui and Wu 14, and Salman et al 15. It is well known that RANS equations introduce additional unknowns in that they require extra equations in order to solve. These equations incorporate various assumptions in order to estimate the Reynolds stresses and form a closed set of equations. Therefore, RANS simulations may require thorough validation against experimental data in order to use the results to predict physically meaningful flows. Salman et al. 15 simulated a lobed mixing flow down stream a two dimensional lobed splitter plate by using a standard k-ε turbulence model to close the RANS equations. Based on the validation of their simulation results against the experimental measurements of McCormick and Bennett 8, they evaluated the applicability of the standard k-ε turbulence model to predict the planar lobed mixing flow. The results of Salman et. al. 15 revealed that the standard k-ε model generally predicted flow structures correctly albeit with lower magnitudes of momentum thickness and higher values of streamwise circulation. In should be noted that many practical applications of lobed mixers are in the form of the circular nozzles instead of two-dimensional splitter plates. When a two dimensional lobed splitter plate is used, the vertical components of the secondary flow velocities are greater than the horizontal components, whereas their magnitudes are comparable in size for a three-dimensional situation 16, as shown in the measurements of Hu et al. 12. This effect could result in a more complex evolution and interaction of vortex structures in lobed jet mixing flows that may be more numerically challenging. The main objective of the present study is to assess and quantify the applicability of numerically simulating circular lobed jet mixing flows with a modest number of unstructured tetrahedral cells and commonly used turbulence models. In the present paper, we will present our numerical simulation results of streamwise and azimuthal (spanwise) vortex structures in a circular lobed jet flow by solving a three-dimensional incompressible form of the Reynolds-Averaged Navier-Stokes (RANS) equations. The Reynolds stresses are modeled using four widely used two-equation turbulence models: The k-ε standard turbulence model, k-ε Realizable turbulence model, k-ω standard turbulence model, and k-ω Shear Strain Turbulence (SST) model. Numerical simulation results will be compared with the Dual-plane Stereoscopic PIV measurement results of Hu et al. 12 quantitatively to evaluate the level of agreement between the predictions and the experimental results of the circular lobed jet mixing flow. The applicability of the four widely used RANS turbulence models is assessed and quantified based on these comparisons. II. Computational Modeling and Procedure A. Geometry and flow conditions Figure 1 shows the geometry of the circular lobed nozzle/mixer used in the present study. The nozzle inner diameter (D) upstream of the lobed trailing edge is 40 mm. Six lobed structures with straight parallel walls form the trailing edge of the test nozzle. Lobe width and height are 6 mm and 15 mm respectively. The nozzle has a contraction ratio of 1.38 corresponding to exit-plane area of 908mm 2. Outer and inner penetration angles measure 17 and 22 along the inner surface, and 9 and 28 from the outer surface due to thickness variation between circular pipe and trailing edge. Reference velocity, U 0 taken from Hu et al. 12 is 20 m/s at the nozzle exit. The fluid medium is air, and is passed through a settling chamber with honeycomb mesh structures prior to entering the test nozzle. The Reynolds number (Re D ) of the lobed jet flow is 5.478 10 4 based on nozzle inner diameter, D. 3

Figure 1. Modeled nozzle/mixer geometry (dimension in mm) B. Computational Modeling Specific modeling assumptions are made to set the initial and boundary conditions in the present numerical simulation. The lobed jet mixing flow is assumed incompressible based on maximum experimental velocities recorded of 20 m/s at the nozzle exit. The incoming velocity in the circular pipe upstream the lobed trailing edge is calculated by applying mass conservation of exit velocity and nozzle contraction ratio. For the present study, the inlet velocity is set to a constant 14.5 m/s one nozzle diameter upstream of the nozzle exit plane. The turbulence quantities at the inlet are estimated to be 3% of the incoming velocities. The computational domain encompasses one lobe period (60 ) of the nozzle and extends four nozzle diameters downstream of the exit-plane. Figure 2 shows the nozzle/mixer and the domain modeled. Constant static pressure is maintained across the outer domain extents with rotationally periodic boundary condition imposed on each side. Figure 2. Computational domain and boundary conditions. In the present study, the numerical simulations were conducted by using a commercial CFD solver, Fluent V6.1. The numerical scheme incorporated uses a threedimensional incompressible form of the Reynolds-Averaged Navier-Stokes (RANS) equations similar to Salman et. al.. 15 The Reynolds stresses are modeled using four widely used two-equation turbulence models. The k-ε standard turbulence model 17, k-ε Realizable turbulence model, k-ω standard turbulence model, and k-ω Shear Strain Turbulence (SST) models are chosen for in the present study to form a closed set of the RANS equations. A 4

thorough description of the models and constants used is given in reference 18. The four turbulence models were intentionally chosen to represent those commonly used in industry. The aforementioned models are preferred because of low computational cost and adequate stability required for industrial applications. The nozzle and domain are modeled and meshed using commercial software packages, Pro/Engineer and Gambit. Figure 3 shows the tetrahedral surface meshes on the lobed surface. Unstructured tetrahedral cells are attractive in circular lobed jet simulations as they can be generated quickly for complicated geometries typical of lobed nozzle/mixer without creating highly skewed cells. This is important when the goal of the study is to determine the best lobe shape for a practical application. It is noted that the cells are quite coarse in the boundary layer and cannot physically resolve the flow near the boundary; thus, it depends upon standard wall functions incorporated into the mathematical model to approximate the viscous sublayer. The meshes were refined until no significant changes between cases with different mesh density were observed. The final mesh number was 434,437 cells in the present study. Computations were performed on a windows based single processor Pentium 4 class 3.1Ghz workstation with 3Gb of physical memory. Model convergence is obtained within 12-15 hours (5000 7000 iterations) after initialization. Convergence is determined through monitoring flow properties at the exit plane and at the constant pressure downstream boundary. Averaged velocity and turbulent kinetic energy are recorded at these boundaries and the solution is converged when values remain constant to at least five significant digits. Figure 3. The tetrahedral surface mesh on the lobed surface. III. Results and Discussion Figure 4 to Fig. 6 compares the results of the present numerical simulations with k-ε Realizable and k-ω SST turbulence models with the DP-SPIV experimental measurements of Hu et al. 12 with respect to streamwise vorticity, azimuthal (spanwise) vorticity and turbulent kinetic energy distributions. The standard forms of the each turbulence model is found to show similar behavior, and therefore not presented here. It should be noted that the numerical simulation results shown in the figures are created by duplicating and periodically rotating the computational domain (periodic on 60 intervals) through a full 360. Figure 7 to 9 give a more detailed quantitative comparison between numerical predictions and the experimental results. The maximum streamwise vorticity, maximum azimuthal vorticity and averaged turbulent kinetic energy is plotted versus downstream distance and is used to evaluate the effect of each turbulence model on predicting the measured quantities. A. Streamwise vorticity distributions From the mean streamwise vorticity distributions shown in Fig. 4, it can be seen that the numerical simulations agree with the DP-SPIV measurement results quite well in revealing the evolution of large-scale streamwise vortex structures generated by the special geometry of the lobed nozzle. The signature of the lobed nozzle is evident by six pairs of large-scale counter rotating streamwise vortices that correspond to each of the six lobes at the nozzle trailing edge. The streamwise vortex structures are found to grow in size, and expand radially as they travel downstream. The strength, contrary to the size, decays rapidly with increasing downstream distance. In the X/D=0.5 cross plane, both turbulence models predict a more elliptic pattern than the experimental results which tend to bulge azimuthally near the lobe peaks. Values of maximum streamwise vorticity at the center of each 5

streamwise vortex core predicted by numerical simulations are slightly smaller (within 10%) than the experimental results, i.e. the numerical simulations under predict the mean streamwise vorticity slightly at this cross plane. The signature of the lobed nozzle is still distinguishable in the mean streamwise vorticity distributions of both numerical simulations and the experimental results in the X/D=1.0 cross plane. Strength of the streamwise vortices for the numerical simulations is nearly identical to experimental results at this location. The streamwise vortices predicted expand azimuthally and closely resemble those revealed in Hu et al. s experimental measurements. The numerical simulation results agree with the experimental data in predicting the streamwise vortex size as well as the radially stretching of the streamwise vortices. Both numerical simulation and measurement results show that the signature of the lobed nozzle is less pronounced in the X/D = 2.0 cross plane compared with upstream sections. The numerical simulation results with both k-ε Realizable turbulence model and k-ω SST turbulence model indicate similar streamwise vorticity strength although they are found to be smaller in size in the k-ω SST model prediction. In this cross plane, the numerical simulation results are found to agree with experimental results quite well in the terms of both vortcity strength and distribution pattern. B. Azimuthal (spanwise) vorticity distributions In order to reveal the azimuthal vortex structures in lobed jet mixing flows more clearly, Hu et al. 11 introduced a quantity ω in plane, which is defined as in plane 2 y ω = ω + ω 2 z 6. We adopted the same definition in the present study in order to make a quantitative comparison of the present numerical simulations with Hu et al. s experimental results. From the azimthal (spanwise) vorticity distribution shown in Fig. 5, it can be seen that the numerical simulations agree reasonably well with the DP-SPIV measurement results of Hu et al. 11 in predicting azimuthal vortex structures in the lobed jet mixing flow. As it is expected, the azimuthal vortex structures form a vortex ring at the exit of the lobed nozzle which appear to outline the lobed trailing edge in Fig. 5A. As suggested by McCormick and Bennett 8, the azimuthal vortex ring is pinched off (Fig. 5B) due to the interaction with the streamwise vortex structures in the lobed mixing flow. This azimuthal ring is found to break down at further downstream locations. The broken fragments of the azimuthal vortex ring corresponding to the lobe troughs reconnect to form a wreath-like-structure in the center of the jet flows, while the broken fragments corresponding to the six lobe peaks form six crescents surrounding the wreath-like-structure (Fig. 5C). Compared with the experimental results, the k-ε Realizable turbulence model is found to over predict (about 10%) the maximum magnitude of the azimuthal vorticity in the X/D=0.5 cross-plane, while the k-ω SST turbulence model under predicts by about 8% of the value. The locations of the maximum azimuthal (spanwise) vorticity are found to be at lobe peaks and troughs, which are in good agreement with the experimental results. These results actually verify the conjectures of Hu et al 8 quantitatively that the interaction of the azimuthal vortex structures with the streamwise vortex pairs in lobed jet flows will result in uneven stretching of the spanwise vortices in lobe peaks and troughs. In the X/D = 1.0 cross plane, the magnitudes and locations of maximum spanwise vorticity are nearly identical for the numerical predictions and the experimental measurements. Greatest azimuthal vorticity magnitudes remain located at the lobe peaks and lobe troughs, while the azimuthal vortex ring is found to pinch off at each lobe peak. The magnitude of the azimuthal vorticity is also shown to dissipate as it travels downstream. The k-ω SST turbulence model is found to predict a faster dissipation of the azimuthal vorticity compared with k-ε Realizable turbulence model. Therefore, in the X/D = 2.0 cross plane, the k-ω SST turbulence model predicts a smaller size for the wreath-like-structure at the center of the jet flow compared with prediction of the k-ε Realizable turbulent model and the experiment results. C. Turbulent kinetic energy distributions Figure 6 shows the results of the numerical simulations compared with Hu et al. s experimental data for turbulent kinetic energy distributions in the lobed jet mixing flow. Generally, the numerical simulations agree with the experimental measurements globally to predict turbulent kinetic energy in terms of the distribution patterns and their locations. However, the numerical simulations over predict the turbulent kinetic energy significantly in each of the three investigated cross planes. The turbulent kinetic energy distributions in the X/D = 0.5 cross plane retain the signature of the lobed nozzle presented in above comparisons. As it is expected, the regions with higher turbulent kinetic energy are found to concentrate in the mixing layers between the core jet and the ambient flow for both the numerical simulation and

experiment results. The contour plots indicate distinct differences in magnitude between the two numerical simulations and experimental measurements with the k-ε Realizable model closer predicting the measured turbulent kinetic energy over the k-ω SST model. The numerical simulation results in the X/D = 1.0 cross plane indicate high turbulent kinetic energy distribution patterns in the same regions as the experimental results. Once again, the numerical simulations over predict the magnitudes of the turbulent kinetic energy in this cross plane. It is found that the numerical simulation with k-ω SST turbulence model predicts larger magnitudes of the turbulent kinetic energy compared with the k-ε Realizable model. In the X/D = 2.0 cross plane, it is found that the regions with higher turbulent kinetic energy levels concentrate more at the outer regions than at the inner side of the lobed jet flow. Both k-ε Realizable and k-ω SST models over predict turbulent kinetic energy in the lobe peaks, while the k-ω SST model is also found to over predict the maximum values at the interface of the core jet region. D. The Maximum Streamwise Vorticity The decay profile of the maximum streamwise vorticity in the lobed jet mixing flow versus downstream distance is plotted in Fig. 7. Compared with the experimental data, the k-ω standard turbulence model is found to provide the best prediction of the maximum streamwise vorticity near the nozzle exit (X/D = 0.25), which is within 2% of the experimental value, while the other turbulence models over predict the maximum streamwise vorticity in the range of 5% to 12%. It is found that the streamwise vorticity is under predicted by all the turbulence models in the regions of 0.5 < X/D < 2.0 with the lowest amount of error given by k-ε Realizable turbulence model with an averaged value of 11%. Conversely, the k-ω standard model, which predicts the first mean streamwise vorticity correctly, is found to provide the worst prediction of the streamwise decay among the four investigated turbulence models, yielding about 50% under prediction of the streamwise vorticity at the X/D = 2.0. Further downstream (X/D > 2.0), all the turbulence models continue to under predict the measured values with an increasingly higher error margin. At x/d=4.0, the predicted values tend to a single value approximately 93% lower than the measured values. E. The Maximum Azimuthal (spanwise) Vorticity The decay profiles of the maximum azimuthal vorticity values predicted by the four turbulence models along with the experimental data are given in Fig. 8. All turbulence models tested over predict the maximum azimuthal vorticity near the exit of the lobed nozzle. At the downstream region of X/D > 0.50, the k-ω standard turbulence model, the k-ω SST turbulence model and the standard k-ε turbulent model provide almost the same predictions, under predicting the maximum azimuthal vorticity by an average of 18%. Interestingly, the k-ε Realizable turbulence model is found to stand out from other turbulence models by over predicting the maximum azimuthal vorticity between 10% and 53% in the near field region (X/D < 2.0), but provides good predictions (within 5%) at farther downstream regions of X/D > 2.0 F. Averaged Turbulent Kinetic Energy (K) The change of the averaged turbulent kinetic energy in the lobed jet mixing flow with the downstream distance is shown in Fig. 9. The absolute values of the numerical predictions and the experiment results are plotted in Fig. 9(a). Fig. 9(b) shows the same data as those used in Fig. 9(b) but normalized by the maximum values for each data set. From the Fig. 9 (a), it can be seen that all the four turbulence models over predict averaged turbulent kinetic energy in the lobed jet mixing flow. At the regions near the exit of the lobed nozzle, the k-ω standard model predicts the greatest magnitude of mass-flow averaged turbulent kinetic energy which deviates from the measurement data by about 500% at the X/D = 0.25. Interestingly, the k-ω standard turbulence model gives the closest prediction of massflow averaged turbulent kinetic energy further downstream with the predicted values approaching the experimental data downstream at X/D > 3.0. The predictions of other three models appear overlapped and follow a single profile downstream of X/D > 2.5. Among the four investigated turbulence models, the k-ε Realizable turbulence model provides the closest prediction to the experimental data. The k-ε Realizable turbulence model over predicts the experimental data by approximately 50%. From Figure 9(b), it can be seen that although all turbulence models over predict the averaged turbulent kinetic energy in the lobed jet mixing flow significantly, numerical predictions and experimental results follow similar trends with respect to turbulent kinetic energy production and decay with downstream distance. The averaged turbulent kinetic energy is found to increase rapidly near the exit of the lobed nozzle, followed by a slower, moderate growth rate to its maximum value. Once the averaged turbulent kinetic energy reaches its peak, it begins to decrease slowly with increasing downstream distance. As suggested by Hu et al. 12, such trend of the averaged turbulent kinetic energy production and decay with downstream distance is closely related to the behavior of the 7

steamwise vortices and azimuthal vortex structures in lobed mixing flows: The lobe nozzle/mixer generates largescale streamwise vortices at the downstream of the lobed trailing edge. Due to the stirring effect of the large-scale streamwise vortices, the core jet flow and ambient flow mix intensely in the regions near the exit of the lobed nozzle. The large-scale streamwise vortices generated by the lobed nozzle dissipate very rapidly as the downstream distance increases (as shown in Fig. 7), i.e, the stirring effect of the streamwise vortices weaken. Therefore, the growth rate of the averaged turbulent kinetic energy slows down, and reaches its peak value. Further downstream at (X/D>2.0), the streamwise vortices have become so weak (less than 1/10 of those at the nozzle exit) that their contribution to mixing is minimal. Since most of the turbulent kinetic energy in the lobed jet mixing flow has dissipated due to intensive mixing upstream, the turbulent kinetic energy is found to decrease at further downstream locations. Such trends of the averaged turbulent kinetic energy agree with the findings of McCormick and Bennett 8 in a two-dimensional planar lobed mixing flow, who suggested that the majority of the enhanced mixing caused by the lobed mixer occurs within the first six lobed heights. From the Fig. 9(b), it is evident that the k-ε Realizable turbulence model provides the best prediction among the four studied turbulence models. Compared with the experimental data, the k-ε Realizable turbulence model not only predicts the location of the maximum averaged turbulent kinetic energy correctly, but also gives reasonable estimation of its rapid growth to X/D = 1.5 ~ 2.0 and its slow rate of decay further downstream. IV. Conclusion Numerical simulations of a circular lobed jet mixing flow by solving a three-dimensional incompressible form of the Reynolds-Averaged Navier-Stokes (RANS) equations with unstructured tetrahedral cells were conducted to assess and quantify its ability to simulate lobed jet mixing flows. The Reynolds stresses were modeled by using four widely used turbulence models: k-ε Standard, k-ε Realizable, k-ω Standard, and k-ω Shear Strain Turbulence (SST) models. The numerical simulation results were validated by using quantitative whole-field DP-SPIV measurement results of the same circular lobed jet mixing flow to determine the level of agreement between the numerical predictions and the experiment measurements. The applicability of the four commonly used RANS turbulence models to lobed jet mixing flows was also evaluated based on the qualitative comparisons of the numerical predictions with the experimental results. The most notable result of the present study is that a RANS approach with a commonly used two-equation turbulence model and a moderate density of unstructured cells can successfully predict many significant features about the evolution and interaction of the large-scale streamwise vortices and azimuthal (spanwise) vortical structures in lobed jet mixing flows. The numerical simulation results agree with the experimental data reasonably well in terms of streamwise vorticity and azimuthal (spanwise) vorticity. Although all the four turbulence models over predict the turbulent kinetic energy in the lobed jet mixing flow significantly, the numerical simulation predictions follow a trend similar to the experimental results regarding the production and decay of the averaged turbulent kinetic energy with downstream distance. The k-ε Realizable turbulence model is found to provide the most accurate predictions among the four turbulence models investigated in the present study. References 1 Paynter, G. C. and Birch, S. C., An Experimental and Numerical Study of the 3-D Mixing Flows of a Turbofan Engine Exhaust System, 15 th Aerospace Sciences Meeting, AIAA-77-204, Los Angeles, CA., 1977. 2 Rowe, R. K. and Huchar, A. P., Engineering Efficient Engine (E 3 ) Scaled Mixer Performance Report Final Report, NASA CR-167947, No. 1982. 3 Paterson, R. W., Turbofan Mixer Nozzle Flow Field Vol. 1- A Benchmark Experimental Study, NASA CR-3492, 1982. 4 Paterson, R. W., Turbofan Mixer Nozzle Flow Field Benchmark Experimental Study, Journal of Engineering for Gas Turbines and Power, Vol. 106, July 1984, pp. 692-698. 5 Werle, M. J., Paterson, R. W., and Presz, W. M., Flow Structure in a Periodic Axial Vortex Array, 25 th Aerospace Sciences Meeting, AIAA-87-0610, Reno, NV., 1987. 6 Eckerle, W. A., Sheibani, H. and Awad, J., Experimental Measurement of the Vortex Development Downstream of a Lobed Forced Mixer, Journal of Engineering for Gas Turbines and Power, Vol. 114, January 1992, pp. 63-71. 8

7 Elliott, J. K., Manning, Y. J., Qiu, Y. J., Greitzer, E. M., Tan, C. S. and Tillman, T. G., Computational and Experimental Studies of Flow in Multi Lobed Forced Mixers, 28 th AIAA/SAE/ASME Joint Propulsion Conference, Nashville, TN., 1992. 8 McCormick, D.C. and Bennett, J. C., Vortical and Turbulent Structure of a Lobed Mixer Free Shear Layer, AIAA Journal, Vol. 32, No. 9, 1994, pp. 1852-1859. 9 Hu H., Kobayashi T., Saga T. and Taniguchi N., PIV and LIF Measurements on the Lobed Jet Mixing Flows, Experiments in Fluids, Vol.29, No.7, 2000, pps141-157. 10 Hu, H., Saga, T., Kobayashi, T. and Taniguchi, N., A Study on a Lobed Jet Mixing Flow by Using Stereoscopic PIV Technique, Physics of Fluids, Vol.13, No. 11, 2001, pp3425-3441. 11 Hu, H., Saga, T., Kobayashi, T. and Taniguchi, N., Mixing Process in a Lobed Jet Flow, AIAA Journal, Vol. 40, No. 7,, 2002, pp1339-1345. 12 Hu, H., Saga, T., Kobayashi, T. and Taniguchi, N. Simultaneous Measurements of All Three Components of Velocity and Vorticity Vectors in a Lobed Jet Flow by Means of Dual-plane Stereoscopic Particle Image Velocimetry, Physics of Fluids, Vol. 14, No. 7, 2002, pp2128-2138. 13 Koutmos, P., and McGuirk, J. J., Turbofan Forced Mixer/Nozzle Temperature and Flow Field Modeling, International Journal of Heat andmass Transfer, Vol. 32, No. 6, 1989, pp. 1141 1153. 14 Tsui, Y.-Y., and Wu, P.-W., Investigation of the Mixing Flow Structure in Multilobe Mixers, AIAA Journal, Vol. 34, No. 7, 1996, pp. 1386 1391. 15 Salman, H., Page, G. J., and McGuirk, J. J., Prediction of Lobed Mixer Vortical Structures with a k-ε Turbulence Model, AIAA Journal, Vol. 41, No. 5, 2003, pp. 878-887. 16 Belovich, V. M., and Samimy, M., Mixing Processes in a Coaxial Geometry with a Central Lobed Mixer-Nozzle, 34 th Aerospace Sciences Meeting, AIAA-96-0118, Reno, NV., 1996. 17 Launder, B. E. and Spalding, D. B., The Numerical Computation of Turbulent Flows, Computer Methods in Applied Mechanics and Engineering, Vol. 3, 1974, pp. 269-289. 18 FLUENT, Software and Documentation, Ver. 6.1.18, Fluent, Inc, Centerra Park Lebanon, NH, 2003. 9

` A. X/D=0.50 cross plane B. X/D=1.00 cross plane C. X/D=2.00 cross plane Figure 4. Normalized streamwise vorticity distributions in the lobed jet mixing flow 10

A. X/D=2.00 cross plane B. X/D=2.00 cross plane C. X/D=2.00 cross plane Figure 5. Azimutal (spanwise) vorticity distributions in the lobed jet mixing flow 11

A. X/D=2.00 cross plane B. X/D=2.00 cross plane C. X/D=2.00 cross plane Figure 6. Turbulent kinetic energy distributions in the lobed jet mixing flow 12

Figure 7. Decay of the maximum streamwise vorticity Figure 8. Decay of the maximum azimuthal vorticity (a). Absolute values (b). Normalized values Figure 9. Average turbulent kinetic energy profile 13