Computational fluid dynamics (CFD) of a three-inlet-Y-junction duct connected to a U-bend: Influence of inlet velocities on air and water flows subjected to cold fronts and turbulence

 Corresponding author: Pape Tamsir Ndiaye Copyright © 2025 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution Liscense 4.0. Computational fluid dynamics (CFD) of a three-inlet-Y-junction duct connected to a U-bend: Influence of inlet velocities on air and water flows subjected to cold fronts and turbulence Pape Tamsir Ndiaye 1, *, Goumbo Ndiaye 2, Omar Ngor Thiam 1, 3, Momath Ndiaye 1, 4 and Oumar Drame 1, 3 1 Fluid Mechanics and Transfer Laboratory, Department of Physics, Sciences and Technologies Faculty, Cheikh Anta DIOP University, Dakar-Fann, Senegal. 2 The Water, Energy, Environment and Industrial Processes Laboratory of the Polytech Higher School, Cheikh Anta Diop University, Dakar, Senegal. 3 Research Group on Solar Energy and Transfers (GREST), Sciences and Technologies Faculty, Cheikh Anta DIOP University, Dakar-Fann, Senegal. 4 Department of the Ufr Hydraulics, Rural Engineering, Machinery and Renewable Energy,University of Sine Saloum Elhadji Ibrahima NIASS, Kaolack, Senegal. World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 Publication history: Received on 30 June 2025; revised on 09 August 2025; accepted on 11 August 2025 Article DOI: https://doi.org/10.30574/wjarr.2025.27.2.2838 Abstract This study presents the movement of air and water flowing in a Y-junction duct with three connected inlets of a U-bend with attention to the impact of cold fronts. Analyses were carried out to observe the behavior of the fluids (air and water) at the duct axis, after the mixing zone (at 70 mm), just upstream of the bend and downstream, at the outlet. The realizable k-ε viscous turbulence model coupled with the energy equation were solved numerically using Ansys Fluent 2024R2. Three cases of cold front formation were studied according to the inlet velocities. An extended cold front appears when the cold fluid enters slowly from inlets 2 and 3, while a high velocity at inlet 1 (case II) can extend the front to outlet CS3. The U-bend acts as a turbulence amplifier, but the response of the fluid in question (air and water) depends both on their physical properties (density, kinematic viscosity, and inertia) and their physicochemical characteristics. The complex dynamics of these flows are explained not only by mechanical phenomena but also by intermolecular interactions: water (a polar molecule), capable of forming hydrogen bonds and susceptible to cold fronts, reacts differently from air, a mixture of non-polar gases with weak intermolecular interactions. Keywords: Ansys Fluent; Air and water movement; CFD; Heat and mass transfer; realizable k-ε turbulence model; TJunction; U-bend 1. Introduction Essential to achieving optimal thermal comfort and acceptable indoor air quality, the mixing of air at different temperatures is an important issue. In heating, ventilation, and air conditioning (HVAC), air quality control considers the mastery of the movement of hot and cold air essential. Also, the mixing of water at different temperatures is considered during the construction of wastewater treatment plants, water treatment, and hydraulic engineering. The shape of ducts, tubes, channels, conduits, channels, and vents plays a major role in air and water dynamics for achieving acceptable indoor air or water quality, hot air control, wastewater management, plant construction, water treatment, and hydraulic engineering. Whenever it is necessary to divide or combine flows, joints (T, Y, etc.) are used in the main piping systems. World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 681 Along with joints, many other obstacles such as valves, turbines, pumps, elbows, elbow-junctions, contractions, expansions, meters may be present. They affect the overall efficiency by causing major and minor losses in the pipes. A lot of work on junctions of experimental and numerical CFD setups with tools such as ANSYS FLUENT, OPENFOAM are used for this purpose [1-16]. Athulya A. Sa et al. [1] presented a multiphase flow model through the T-junction and the phase redistribution phenomenon at the junction using ANSYS FLUENT software. Based on their study, the following conclusions were drawn: it can be seen that a larger amount of air pocket is formed in the lower part of the bypass arm, phase separation is negligible in the bypass arm due to the effect of gravity and by analyzing the multiphase flow, they saw that considering the fluid alone, phase separation is greater at the outlet than at the junction. Nimadge Mr.G.B. et al. [2] set the goal to study the steady and incompressible flow of a fluid through a T-junction and to get familiar with CFD by focusing on losses in piping systems, as the working fluid through pipes plays an important role in the operation of industries such as chemical industries, petroleum industries, etc. They prepared an experimental setup to obtain the reference data when the fluid passes through the T-junction of the pipe, the same data is used for CFD analysis using software such as ANSYS FLUENT. Khan Wasim et al. [3] numerically studied the plug formation mechanism during gas-liquid two-phase flow in a Tjunction microchannel by developing a two-dimensional (2D) model of the microchannel using ANSYS Academic Research CFD 18.2 software and adopting the volume of fluid (VOF) method to solve it. Their obtained results agree well with the experimental results. The plug length, pressure drop, and velocity variations inside the plugs were measured under different operating conditions. The effects of contact angle (0°–155°), fluid viscosity, and surface tension on the two-phase flow interaction parameters, as well as the effect of gas and liquid superficial velocities, were examined in detail. They found that the liquid film formed at low capillary count (Ca) was very thin and was observed only for fine meshes near the channel wall. DOROSHENKO Yaroslav et al. [4] carried out 3D modeling of the elbow and T-junction in the linear part of the gas pipeline, especially at the places where a complex movement of multiphase flows occurs and changes its direction. Based on the Lagrangian approach (Discrete Phase Model - DPM), numerical modeling methods were developed to simulate the multiphase flow movement in the elbow and T-junction of the linear part of the gas pipeline using the ANSYS Fluent R17.0 Academic software package. The mathematical model is based on solving the Navier-Stokes equations and closed discrete phase continuity and motion equations with the Launder-Sharma two-parameter turbulence model (k-e) with appropriate initial and boundary conditions. These results provide the opportunity for a comprehensive and in-depth study of the erosive wear of the elbow and T-junction of the linear part of the gas pipeline and adjacent pipeline sections, as well as an assessment of their strength and residual life. The velocity of liquid and solid particles, the impact angles, the diameters of condensed droplets and solid particles at the collision site were determined. Such studies open the prospect of a comprehensive and in-depth research on the erosive wear of elbows and T-junctions in the linear section of gas pipelines. RAID AHMED MAHMOOD [5] presented a CFD simulation study to predict and visualize the separation of two-phase flows in the vertical T-junction and a comparison between the CFD results and the associated experimental results to validate their CFD results. Using ANSYS 17.1 to perform the simulation, the geometry of the T-junction was generated by ANSYS modular design based on the dimensions of the experimental test section and then passed to the ANSYS mesh to generate a suitable mesh. A numerical study in a three-inlet Y-junction microchannel was carried out with two different numerical codes: ANSYS Fluent and Open-FOAM by Chiriac Eugen et al. [6]. Using the same inlet and boundary conditions, three main parameters are studied: velocity amplitude, pressure and vorticity amplitude. The microchannel was fabricated using PDMS (Polydimethylsiloxane) soft litography and were used for the validation of numerical simulations. The simulation results were analyzed and a comparison between the numerical codes was carried out with the aim of testing the capabilities of the two numerical codes in a simulation of flow mixing in a microchannel. Makarem M. A. [7] proposed a CFD simulation of CO2 capture in a microchannel by aqueous mixtures of MEA and [Bmim] BF4 modified with TiO2 nanoparticles as chemical additives. The flow hydrodynamics, mass transfer characteristics and CO₂ absorption performance of the proposed solvent were investigated in a T-shaped microchannel structure by steady-state computational fluid dynamics technique. To present a detailed model, the Navier-Stokes and continuity equations are combined with a two-phase laminar flow module accounting for mass transfer between heterogeneous phases. The effects of [Bmim]BF₄ and TiO₂ mass fractions on CO₂ loading, bubble formation, and velocity profile were then studied at different gas and liquid holdups, with an ionic liquid fraction ranging from 0% to 10% and World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 682 a nanoparticle fraction ranging from 0% to 0.1%. According to their simulation results, the solvent containing 10% [Bmim]BF4, 3% MEA, and 0.04% TiO2 exhibited the maximum purification fraction of 79.62%. Taha Enas Salman et al. [8] examined the influence of turbulent parameters on the characteristic centerline of a fluid flow through a T-junction connected to a Venturi tube via a pipe. The continuity equation, momentum equation and energy equation of water are modeled and solved using ANSYS FLUENT 2020R1 software while for turbulence, the standard type turbulent model (k-ε) is used. The model is obtained on the velocity distribution, pressure drop distribution, turbulent kinetic energy and turbulent dissipation rate. The results showed a divergence in the values of the effect of the T-junction on the Venturi meter regarding the velocity distribution and pressure drop, while they showed a similar behavior for the turbulence parameters. A three-dimensional (3D) numerical simulation of a liquid/liquid two-phase flow was carried out in a rectangular microchannel with a T-junction by SAID Mohammed et al. [9]. They used the volume of fluid (VOF) method with ANSYS Fluent to capture the interface between the two phases and the dynamic mesh adaptation technique along with the assumption of a symmetry plane helped them reduce the computational cost. Their study focused on the flow patterns and hydrodynamics of plugs and their results revealed six distinct flow patterns by dispersing water in a continuous silicone oil phase. By decreasing the flow rate ratio as well as the viscosity ratio, the liquid film thickness increases in the corners and side planes. In turn, this has a significant impact on the liquid film velocity and the plug velocity. Bushra Khatoon et al. [10] conducted numerical and experimental studies for the analysis of hydrodynamics and volumetric mass transfer coefficient in a cross-T junction microchannel for a two-phase gas-liquid flow system. Initially, the CO2-water hydrodynamic simulation was performed using ANSYS-FLUENT 2021 R2 with the fluid volume technique. The computational fluid dynamics model was validated by comparing the results with experimental data. The results obtained in numerical simulation and experimental work show that the total volumetric mass transfer coefficient (margin 0.1–0.8 1/s) increases with increasing gas velocity, but it decreases with increasing film thickness (0.01–0.05 mm) and temperature (T = 298.15 K and 303.15 K). It was observed that with increasing bubble velocities, the total mass transfer coefficient also increases, and with increasing film thickness, it shows the opposite effect due to the dominant parameters in the microchannel, i.e., surface tension. Wang Fuzhang et al. [11] presented a paper on the dynamics of unsteady air and water through a three-inlet T-shaped duct, focusing on the problem of cold fronts. They performed simulations to notice the nature of these fluids some distances after mixing (70 mm), before turning the duct and after turning the duct. They solved the feasible k-ε viscous model and energy equation numerically using Ansys Fluent 2022R1. The examination of residual independence and meshing was taken into account to check the convergence of the results. Considering the variation of inlet velocity, they examined three cases of cold front formation. When the inlet velocities of cold air/water and warm air/water are the same, the formation of the cold front is invisible at the early stage despite the fact that air flows faster than water. They found that the optimal average vortex kinetic energy per unit mass formed near the T-junction, not due to previously formed cold fronts, but due to the pressure recirculation at the bend affecting the velocity of the cold fluid substance coming from the two adjacent inlets (i.e., Inlet 2 and Inlet 3). However, like Wang Fuzhang et al. [11], we propose a CFD study and a comparative analysis between the dynamics of air and water through a Y-junction duct (not T-junction unlike Wang Fuzhang et al. [11]) followed by a three-inlet Ubend where mixing and cold fronts start at 70 mm from the three inlets. We will study the effects of the variation of the three inlet velocities of the fluids (air and water) cold or hot on the flow and the overall dynamics (velocity field, pressure variations, temperature and turbulence) and how they vary in the duct at a distance after mixing, before the bend and after the bend and affect the formation and evolution of cold fronts. We will also emphasize the difference between air and water. 2. Mathematical Formulation 2.1. Description of the physical model and simplifying assumptions: In this work, we considered two circular aluminum ducts of 40 mm diameter, where it was assumed that air and water circulate separately in each of the ducts in a Y-junction shape with three inlets followed by a U-bend. As shown in Figure 1, the mixing zone is located 70 mm from inlets 1, 2 and 3 in each of these ducts. The geometric structure in 3D mode, presents a circular surface (CS1) located 30 mm from the mixing domain, another circular surface located 100 mm from the mixing domain (CS2) and an outlet located 20 mm from the duct bend (CS3). The hot fluid (air or water) enters the domain through inlet 1 with a velocity InV1 and a temperature of 30 °C, while the cold fluid inlets are located through inlets 2 and 3 with velocities InV2 and InV3 at 5 °C, respectively. The physical properties of fluids are listed in Table 1. World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 683 Figure 1 Illustration of the geometric model (three-way Y-junction followed by a U-bend) Table 1 The physical properties of the fluid considered : air and liquid water Air liquid water Density ρ : [Kg/m3] 1.225 998.2 Viscosity µ : [Kg/(m.s)] 1.7894-× 10-5 1.003× 10-3 Thermal Conductivity k : [W/(m.K)] 0.0242 0.6 Specific Heat Cp : [J/(Kg.K] 1006.43 4182 To carry out this work, we admit some simplifying hypotheses such as: the fluids (water and air) are incompressible and Newtonian, the physical properties are quasi-constant, there is no phase change and the flow is fully turbulent and permanent. 2.2. Governing equations Although Reynolds Averaged Navier-Stokes (RANS) models have some limitations but deserve to be recognized as the most reliable for characterizing velocity fields of turbulent flows due to near-wall modeling, turbulence modeling (i.e., involving averaging the equations governing fluid motion over time or space to obtain an average flow field and additional terms representing turbulence effects), reliable for higher Reynolds numbers, inherently assumes that turbulent fluctuations can be decomposed into an average and fluctuating component, with the fluctuations decaying rapidly toward zero. Moreover, RANS provides reasonable results when interactions, separation, and recirculation are simple [12-13]. In this study, the feasible k-ε viscous model of RANS by Shih et al. [14] considered the eddy viscosity not as a constant, but as a function depending on the angular rotational velocity of the system, the average deformation and rotational velocities, as well as the turbulence properties (k and ε). The continuity (1.a) and momentum (1.b) equations of averaged Navier-Stokes called RANS are as follows: World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 684 ( ) ( ) 0 i i i i j ij i i j i i j u x p u u u f u u t x x x      =         + = − + −        ………(1) Since there are more unknowns than equations, a strategy to “close” the system a is required. The feasible k-ε closure model is accurate for studying the velocity distribution of turbulent flow through a curved channel, according to Shaheed et al. [15]. It captures rotation, vortices, strong curvature of the streamline, and the unknown dissipation ε that incorporates the mean square fluctuation of the vorticity. It is estimated by two equations (2) and (3): • The turbulent kinetic energy transport equation (k): : ( ) ( ) t j k b M j j k j k k ku P P Y t x x x             + = + + + − −          ……… (2) Where the physical quantities are: density (  ), mean velocity component ( j u ), molecular dynamic viscosity (  ), turbulent viscosity ( t  ), turbulent Prandtl number for k ( k  ), production of k due to shear ( k P ), production of k due to buoyancy ( b P ), dissipation rate of k (), contribution of expansion in compressible flows ( M Y ). • The dissipation rate transport equation (Ꜫ) : ( ) ( ) 2 1 2 1 3 t jb j j j u C S C C C P t x x x k k                     + = + + − +       +    (3) Where the physical quantities are: Turbulent Prandtl number for ε (   ), strain rate modulus S, 2ij ij S S S= avec 1 2 j i ij ji u u Sxx    =+     et   = • Closure: turbulent viscosity ✓ The turbulent viscosity is given by : 2 tC     = ✓ 1max 0.43, 3 C    = +  , k S  = , 22 3 ij ij t ij kS     = − + ✓ C  depends on the flow, it is not constant : 0 1 S Ck A A s   = + ✓ With 04.04A= et 6cos S A  = , which depends on the local structure of the flow 1 3 1cos 6 3 ij jk ki S S S S  − =  ; ij ij S S S= ; 0.09C  = ; 11.44C  = ; 21.9C  = ; 1.0 k  = ; 1.2   = • The energy conservation equation can be expressed as follows:: World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 685 ( ) jeff j j j TT Tu t x x x       +=        (4) In Eq. (4): ✓ T represents the temperature, ✓ λeff is an effective coefficient that includes the contribution of turbulent mixing in addition to molecular conduction and can be expressed as: Pr t eff t Cp   =+ , where k and Cp are the thermal conductivity and specific heat at constant fluid pressure and µt is the turbulent viscosity. ✓ Prt is a turbulent Prandtl number, we will take: Prt= 0.85. 3. Numerical Modeling and Validation: Since it is virtually impossible to obtain an exact analytical solution to the equations of our problem, we will resort to a numerical solution. The finite volume method was used [16]. It is more economical in terms of volume and computation time while ensuring high computational stability and is more efficient than classical finite difference methods for complex geometry problems. We opted for the QUICK scheme to discretize the convective terms. It minimizes numerical diffusion effects and is renowned for its accuracy in representing gradients, and is particularly effective for turbulent flows and high gradients. For the pressure-velocity coupling, we chose the coupled scheme [16]. Numerical tests are performed with convergence threshold residuals for the continuity, motion, k-epsilon, and energy equations equal to 10-6 in order to address both speed, accuracy, and better convergence of the calculations. Mesh independence was investigated and found to be satisfactory. We performed the numerical solution using Ansys Fluent 2024R2. Thanks to a satisfactory comparison with results of Wang Fuzhang et al. [11] and we validated our work. 4. Analysis and discussion of results Each inlet, whether hot or cold, has its own characteristics (section, temperature and speed, etc.), which implies distinct physical properties depending on the cases studied: case I (InV1= InV2= InV3 = 3), case II (InV1 =10; InV2 = InV3 =3) and case III (InV1 =3; InV2 = InV3 =10). These differences are reflected in particular by potential variations in the mass flow rate and the total heat transfer rate for each inlet, which directly influences the flow dynamics and the heat transfer phenomena in the duct. Tables 2 and 3 present the mass flow rate and the total heat transfer rate for each inlet of each case for air and water. Table 2 Mass flow rate [kg/s] Case (Inlet Velocities [m/s]) Inlet 1 [kg/s] Inlet 2 [kg/s] Inlet 3[kg/s] Motion of air Case I InV1 = InV2 = InV3 = 3 0.00461814120078 0.00461814120078 0.00461814120078 Motion of water 3.763125344176 3.763125344176 3.763125344176 Motion of air Case II InV1=10; InV2 = InV3 =3 0.01539380400259 0.00461814120078 0.00461814120078 Motion of water 12.5437511472533 3.763125344176 3.763125344176 Motion of air Case III InV1=3; InV2 = InV3 = 10 0.00461814120078 0.01539380400259 0.01539380400259 Motion of water 3.763125344176 12.5437511472533 12.5437511472533 The meeting of the jets from the three inlets or the addition of their physical quantities creates shocks or collisions in the mixing zone of the junction which will generate strong hydrodynamic and thermal interactions, disturbances and turbulence in the conduit. To evaluate and study its effects, the velocity, temperature distribution and turbulent kinetic energy profiles in the plane (x,y,z=0) and at sections CS1 (i.e., the circular surface 30 mm after mixing), CS2 (i.e., another circular surface 100 mm from the mixing domain) and CS3 (i.e., the circular outlet surface, located 20 mm after the duct bend) with three cases: case I (InV1= InV2= InV3 = 3), case II (InV1 =10 ; InV2 = InV3 =3) and case III (InV1 =3 ; InV2 = InV3 =10) are presented according to the contours (Figure 2-19). World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 686 Table 3 Total heat transfer rate at the inlets [W]: CaseInlet Velocities [m/s] Inlet 1 [W] Inlet 2[W] Inlet 3[W] Motion of air Case I InV1 = InV2 = InV3 = 3 23.23917924349 -92.9567169739598 -92.9567169739598 Motion of water 78,686.9509467201 -314,747.80378688 -314,747.80378688 Motion of air Case II InV1=10; InV2 = InV3 =3 77.4639308116332 -92.9567169739598 -92.9567169739598 Motion of water 262,289.836489067 -314,747.80378688 -314,747.80378688 Motion of air Case III InV1=3; InV2 = InV3 = 10 23.23917924349 -309.855723246533 -309.855723246533 Motion of water 78,686.9509467201 - 1,049,159.34595627 - 1,049,159.34595627 The fluid velocity appears to increase along the radial direction, with the maximum velocity occurring at the pipe axis (Figure 2-7). At the intersection, the horizontal jet is disturbed by the two inclined jets from the two inlets (inlet 2 and inlet 3). This causes velocity gradients (increased local velocity, intense mixing with high turbulence) and vortices in the mixing zone at the center of the junction. Flow acceleration is observed after the junction (mixing zone) and before the bend. At the bend, its centrifugal effect causes the peak velocity to be deflected toward the outer wall of the bend and decelerated toward the inner wall. When a fluid moves from a straight path toward a curved pipe (180° bend), the flow appears to be subject to a centrifugal force that pulls the fluid outward. This is compounded by a radial pressure variation that attempts to compensate for this centrifugal force. It creates transverse instability and generates secondary flows, in addition to the main flow. These secondary flows form two counter-rotating vortices called Dean vortices. Hence, the increase or maximization of velocity at the inner wall of the bend is due to the secondary movement (Dean vortex). The opposite effect occurs when the fluid leaves the bend towards the straight pipe. At the outer wall, the area with the largest radius, friction is greater, which can also lead to a slowdown. The turbulent flow speed of air and water through sections CS1 and CS2 appears to be almost identical for cases 1 (Figs. 2 and 3). They show an increasingly homogeneous distribution of velocity, higher in the center. In CS2, the effect of the bend begins to be felt with the decrease in velocity at the bottom. In CS3, after the bend the minimum velocity is rejected towards the inner wall (blue spot at the bottom which enlarges), the maximum velocity is rejected towards the outer wall, which is due to the recirculation of the fluid in the bend (Dean vortex effect). For Case II (InV1 = 10; InV2 = InV3 = 3), the mass flow rate of inlet 1 is higher than that of inlets 2 and 3. The collision of the latter two with the first is not strong enough to disturb it, hence the mixing zone of the junction is less agitated (Figure 4,5). The velocities in CS1 and CS2 are higher than those in Case I because its mass flow rate from inlet 1 is greater and more imposing. The effect of the U-bend will be felt in CS3. For Case III (InV1 = 3; InV2 = InV3 = 10), the cumulative mass flow rates of the three inlets are greater than those in Cases II and I, leading to much stronger collisions and much steeper velocity gradients, resulting in increased turbulence (Figure 6,7). The velocities in CS1 and CS2 are more significant than those in Cases I and II. The effects of the secondary flow of the Dean vortex are more pronounced in case III, followed by case II and case I. We have a better hydrodynamic mixing respectively for case III, case II, and case I. The blue spot at the bottom (slow fluid) on CS3 is much more significant for air than for water and much more persistent respectively for cases I, followed by case II and case III (hydrodynamic mixing). This is due to the fact that hydrogen bonds are more present in water than in air (the dynamic viscosity of water is greater than that of air). World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 687 Figure 2 Velocity distribution across the domain of air dynamics - Case I (InV1= InV2= InV3 = 3). Figure 3 Velocity distribution across the domain of water dynamics - Case I(InV1= InV2= InV3 = 3). Figure 4 Velocity distribution across the domain of air dynamics - Case II (InV1 =10 ; InV2 = InV3 =3). Figure 5 Velocity distribution across the domain of water dynamics - Case II (InV1 =10 ; InV2 = InV3 =3). Figure 6 Velocity distribution across the domain of air dynamics - Case III (InV1 =3 ; InV2 = InV3 =10). Figure 7 Velocity distribution across the domain of water dynamics - Case III (InV1 =3 ; InV2 = InV3 =10). Figure 8-13 shows the hot flow inlet 1 and the other two inlets (inlet 2 and inlet 3) which carry cold fluid. With the meeting of the cold flows on the hot flow, a clear thermal transition zone called cold fronts are formed and emerge from the mixing zone of the junction along the axis of the duct in the three cases (Case I, Case II and Case III). This formation of visible cold fronts is very important in the mixing domain. This is due to the significant changes observed in the temperature distribution with the velocities of each inlet. World Journal of Advanced Research and Reviews, 2025, 27(02), 680-692 688 Figure 8 Temperature distribution across the domain of air dynamics - Case I (InV1= InV2= InV3 = 3).. Figure 9 Temperature distribution across the domain of water dynamics - Case I (InV1= InV2= InV3 = 3). Figure 10 Temperature distribution across the domain of air dynamics - Case II (InV1 =10 ; InV2 = InV3 =3). Figure 11 Temperature distribution across the domain of water dynamics - Case II (InV1 =10 ; InV2 = InV3 =3). Figure 12 Temperature distribution across the domain of air dynamics - Case III (InV1 =3 ; InV2 = InV3 =10). Figure 13 Temperature distribution across the domain of water dynamics - Case III (InV1 =3 ; InV2 = InV3 =10). For Case I (InV1= InV2= InV3 = 3), the velocities of the cold flows are equal to those of the hot ones, we have a more moderate and controlled collision (Figure 8,9). For Figure 10, 11 (Case II (InV1 =10; InV2 = InV3 =3)), the heat of the hot flow from inlet 1 is more imposing, so the temperature is little disturbed around the axis of the duct even after the collision at the mixing zone. There is a dominance of the hot in the central zone, on the axis of the duct and a progressive cooling in the lateral zones. For Figure 12, 13 (Case III (InV1 =3; InV2 = InV3 =10)), the hot flow from inlet 1 is weak compared to the two cold flows, so thinner or weakened around the axis of the duct.