betaFeedback

WorksPhysics and Astronomy

Analytical Criterion for the Turbulent-to-Laminar Transition in Channel Flow

Abstract

An analytical criterion for the disappearance of turbulent flow in a channel is developed within the Alexeev hydrodynamic equations (AHE). The analysis builds on an analytical representation of the streamwise velocity as a superposition of a parabolic component and a nonlinear turbulent component coupled to the transverse velocity field. The critical Reynolds number is obtained from the balance of viscous dissipation of kinetic energy associated with the turbulent component. The resulting critical Reynolds number is expressed analytically as Rec=f(δ,γ,n)Re_c=f(δ,γ,n), where δδ is the Reynolds-independent AHE similarity parameter, γγ is the weighting coefficient of the turbulent component, and nn is the transverse-velocity mode number. The criterion describes the disappearance of the turbulent solution as the Reynolds number is decreased and therefore addresses the turbulent-to-laminar transition rather than the linear instability of the laminar solution. Unlike a criterion based on a single universal critical Reynolds number, the AHE criterion predicts a transition threshold that depends on the parameters characterizing the flow. For representative channel-flow parameters, the predicted critical Reynolds number is approximately Rec=1000–1200Re_c=1000\text{--}1200. The predicted threshold is compared with experimental observations of transition and of the persistence of laminar flow in channel flows. The analysis suggests that viscous dissipation provides a mechanism limiting the persistence of the turbulent solution below a critical Reynolds number.%

Introduction

A general criterion for transition to turbulence was established by O. Reynolds [16] using the concept of mechanical similarity of flows of viscous fluid: the flow will be laminar if Reynolds number Re<RecRe < Re_c. The Reynolds number provides a similarity parameter for viscous flows, and transition behavior is commonly characterized by a critical Reynolds number RecRe_c. It was found that the value of RecRe_c corresponding to the transition to turbulence is the smaller, the greater the intensity of the disturbances Monin and Yaglom (1979). If the degree of disturbance at the inlet is decreased significantly, one can delay the transition from laminar to turbulent flow until very high Reynolds numbers. Van Doorne [4] was able to keep laminar flow up to Re=60,000, Comolеt (1950) obtained laminar flow in a pipe up to Re=75,000, and Pfenninger (1961) up to Re=100,000. These results show that the Reynolds number itself is not a unique criterion for transition to turbulence.

We instead determine a lower Reynolds-number threshold associated with the disappearance of the turbulent/coherent branch of the AHE solution as the Reynolds number is decreased. The threshold is evaluated for specified amplitudes and transverse modes of the disturbance represented by VTV^T. When the minimizing value of γ\gamma becomes sufficiently small, the resulting velocity profile approaches the parabolic branch ULU^L.

In the analytical solution by Fedoseyev (2023), the Alexeev (Generalized) Hydrodynamic Equations (AHE) have been employed for turbulent flow in a channel. The AHE were derived from the Generalized Boltzmann Equation (GBE), which takes into account finite particle size Alexeev [1], while in the traditional Boltzmann equation, particles are treated as material points.

The approximate analytical solution was obtained for a mean turbulent flow velocity UU as a superposition of the laminar (parabolic) ULU^{L} and turbulent (superexponential) UTU^{T} solutions,

U=γUT+(1−γ)UL,(1)U = \gamma U^{T} + (1-\gamma)U^{L}, \tag*{(1)}

where the coefficients γ\gamma and (1−γ)(1-\gamma) were introduced. The expressions for UTU^{T} and ULU^{L} were explicitly provided giving

U=U0[γ(1−e1−ey/δ)+(1−γ)4y(L−y)/L2](2)U = U_{0}\left[\gamma\left(1-e^{1-ey/\delta}\right)+(1-\gamma)4y(L-y)/L^{2}\right] \tag*{(2)}

in a 2D channel. The coordinate yy is normalized by the full channel width, so that 0≤y≤10 \le y \le1, with the centerline at y=1/2y=1/2; by symmetry, the analytical solution may be evaluated on 0≤y≤1/20 \le y \le1/2. The velocity scale is the centerline velocity U0U_{0}. All parameters are nondimensional. The Eq. (2) presents the laminar flow velocity if γ=0\gamma=0, and the turbulent flow if γ>0\gamma>0. The parameters δ\delta is

δ=τ0ν/L0,(3)\delta= \sqrt{\tau_{0}\nu}/L_{0}, \tag*{(3)}

where τ0\tau_{0} is the relaxation time, or timescale, a material property for particular liquid or gas used in the experiments, ν\nu is the kinematic viscosity, and L0L_{0} is the hydrodynamic scale. The nondimensional τ\tau, a timescale coefficient for the fluctuation terms in AHE, is expressed as

τ=τ0L0−1U0=δ2Re,(4)\tau= \tau_{0}L_{0}^{-1}U_{0}=\delta^{2}Re, \tag*{(4)}

where Re=U0L0/νRe=U_{0}L_{0}/\nu denotes the Reynolds number.

The analytical solution for transverse velocity of turbulent flow was obtained in Fedoseyev (2024a), and the amplitude is chosen to produce a solution UTU^{T} with the amplitude close to unity Fedoseyev [5]:

VT(y)=−sin⁡(nπy)nπδ2Re,(5)V^{T}(y)=-\frac{\sin(n\pi y)}{n\pi\delta^{2}Re}, \tag*{(5)}

where nn is an integer.

As an example, the analytical solution UU (red line) for the experiments of Wei (1989) is shown in Figure 1 in (U+,y+)(U^{+},y^{+}) coordinates. The experimental velocity is shown as points for four Reynolds numbers. The parameter y+=yuτ/νy^{+}=yu_{\tau}/\nu where uτu_{\tau} is so called friction velocity, yy is the absolute distance from the wall, and ν\nu is the kinematic viscosity. One can interpret y+y^{+} as a local Reynolds number. The friction velocity uτu_{\tau} is defined as

Figure 1. Streamwise velocity profiles for turbulent channel flow. Experimental data from Wei and Willmarth (1989) are shown by symbols. The AHE solution UU at Re=22776Re = 22776 (red line) is compared with the corresponding data (red dots). The laminar ULU^{L} (green line) and turbulent UTU^{T} (blue line, Re=22776Re = 22776) components of the superposition UU are plotted normalized to unity and shown only to illustrate the decomposition of the model. The logarithmic law of the wall, U+=(1/κ)log⁡(y+)+BU^{+} = (1/\kappa)\log(y^{+}) + B, is shown for reference (olive-green line).

uτ=τw/ρ,(6)u_{\tau}=\sqrt{\tau_{w}/\rho}, \tag*{(6)}

where wall shear stress τw\tau_{w}, τw=ρνdUdy\tau_{w}=\rho\nu\frac{dU}{dy} at y=0y=0, and the dimensionless velocity is given by U+=u/uτU^{+}=u/u_{\tau}. The Figure 1 demonstrates that the superposition UU provides an excellent fit to the experimental mean velocity profile for γ=0.65\gamma=0.65 and δ=0.05\delta=0.05, Fedoseyev (2023).

The coefficient γ\gamma in Eq. (1), should not be regarded as an independent empirical fitting parameter. It is obtained from a variational principle based on the minimization of the viscous dissipation associated with the analytical velocity field within the volume swept by the fluid per unit time Fedoseyev (2024b). The resulting value determines the relative weight of the laminar and turbulent branches of the AHE solution. This interpretation is important because the analytical profile is therefore not constructed by fitting an arbitrary functional form to each experimental data set. Once the AHE solution and the minimum-dissipation criterion are specified, γ\gamma is determined by the model.

In this work we use the functional proposed in Fedoseyev (2024b), with the addition of the dissipation term related to VTV^{T}, which becomes increasingly important as Reynolds number decreases. The velocity VTV^{T} plays the role of a transverse disturbance, as discussed in Fedoseyev [6]. The goal is to determine the Reynolds number at which the minimizing value of γ\gamma becomes negligible as ReRe is decreased. The amplitude aa of VTV^{T} is also varied to examine the dependence of the threshold on disturbance amplitude.

The contents of the paper are as follows. Section 2 presents the AHE and its simplified form for turbulent channel flow to which the minimization principle is applied. Section 3 presents the analytical solutions for turbulent flow, formulates the minimization principle to obtain γ\gamma, and compares the resulting RecRe_{c} with experimental data for turbulent channel flows. Section 4 discusses the obtained results, followed by the conclusions.

Alexeev Hydrodynamic Equations

To proceed with the minimization principle, we have to present the governing equation used, the Alexeev Hydrodynamic Equations (AHE). The AHE are obtained from Generalized Boltzmann Transport Equation, Alexeev [2], by multiplying the latter by the standard collision invariants (mass, momentum, and energy), and integrating the result in the velocity space. The particles of finite size are considered. The obtained equations below are valid for incompressible viscous flow, and have the following non-dimensional form, Fedoseyev (2012):

∂V∂t+(V∇)V−Re−1∇2V+∇p−F=τ{2∂∂t(∇p)+∇2(pV)+∇(∇⋅(pV))},(7)\frac{\partial\mathbf{V}}{\partial t} + (\mathbf{V}\nabla)\mathbf{V} - Re^{-1}\nabla^{2}\mathbf{V} + \nabla p - \mathbf{F} = \tau\left\{ 2\frac{\partial}{\partial t}(\nabla p) + \nabla^{2}(p\mathbf{V}) + \nabla(\nabla\cdot(p\mathbf{V})) \right\}, \tag*{(7)}
∇⋅V=τ{2∂∂t(∇⋅V)+∇⋅((V∇)V)+∇2p−∇⋅F},(8)\nabla\cdot\mathbf{V} = \tau\left\{ 2\frac{\partial}{\partial t}(\nabla\cdot\mathbf{V}) + \nabla\cdot((\mathbf{V}\nabla)\mathbf{V}) + \nabla^{2}p - \nabla\cdot\mathbf{F} \right\}, \tag*{(8)}

where V\mathbf{V} and pp are nondimensional velocity and pressure respectively, Re=U0L0/νRe = U_{0}L_{0}/\nu - the Reynolds number, U0U_{0} - velocity scale, L0L_{0} - hydrodynamic length scale, ν\nu - kinematic viscosity, F\mathbf{F} is nondimensional body force and nondimensional timescale τ=τ0L0−1U0\tau= \tau_{0}L_{0}^{-1}U_{0}. Terms containing τ\tau are called the fluctuations (temporal and spatial) by Alexeev [1]. One can see the equations become the Navier-Stokes equations if τ→0\tau\to0.

Additional boundary condition was set for pressure on walls:

(∇p−F)⋅n=0,(9)(\nabla p - \mathbf{F}) \cdot\mathbf{n} = 0, \tag*{(9)}

where n\mathbf{n} is a wall normal.

The AHE is not a turbulence model, and no additional equations are introduced.

AHE for 2D Stationary Incompressible Channel Flow

The case of 2D incompressible fluid flow in channel is considered with the flow direction in xx. For the stationary analytical solution in Fedoseyev (2023), AHE (7), (8) were simplified by: (a) dropping all temporal derivatives, (b) dropping all the terms (with coefficient τ\tau) in the momentum equations, (c) the nonlinear terms were neglected in the fluctuations, (d) all the derivatives in xx were neglected, except for the pressure gradient px=constp_x=\mathrm{const}, so the Laplacian of pressure was ∇2p=pyy\nabla^2p=p_{yy}.

The resulting continuity equation of AHE model is as follows:

vy=τpyy(10)v_y=\tau p_{yy} \tag*{(10)}

while the momentum equations are:

vuy+px=Re−1uyy(11)v u_y+p_x=Re^{-1}u_{yy} \tag*{(11)}
vvy+py=Re−1vyy(12)v v_y+p_y=Re^{-1}v_{yy} \tag*{(12)}

where ReRe is Reynolds number, τ\tau is given by Eq.(4). The boundary conditions are as follows: u=0u=0, v=0v=0 and the normal derivative of pressure pn=0p_n=0 at the wall y=0y=0; u=U0u=U_0, and v=0v=0, ∂p∂y=0\frac{\partial p}{\partial y}=0 (the symmetry conditions) at y=L/2y=L/2. A symmetry about the centerline y=L/2y=L/2 is assumed, and the problem is solved in half of the domain.

Viscous Dissipation and Functional Minimization

Viscous Dissipation in Turbulent Channel Flow

The dissipation function of a Newtonian fluid with viscosity μ\mu in 2D channel is

E=2μ[(∂U∂x)2+(∂V∂y)2−13(∇⋅V)2]+μ[∂V∂x+∂U∂y]2(13)E=2\mu\left[\left(\frac{\partial U}{\partial x}\right)^2+\left(\frac{\partial V}{\partial y}\right)^2-\frac{1}{3}(\nabla\cdot\mathbf{V})^2\right]+\mu\left[\frac{\partial V}{\partial x}+\frac{\partial U}{\partial y}\right]^2 \tag*{(13)}
∇⋅V=∂U∂x+∂V∂y(14)\nabla\cdot\mathbf{V}=\frac{\partial U}{\partial x}+\frac{\partial V}{\partial y} \tag*{(14)}

where the UU and VV are the components of the dimensional velocity vector, V\mathbf{V} is the velocity vector, and xx, yy are streamwise and transverse coordinates. A similar equation was derived in Horne (1986) (p.6, Eq.(15)). In previous paper Fedoseyev (2024b) the VV-velocity terms were neglected due to their small values at high Reynold numbers, and now these terms are retained.

In the case of a developed flow in channel we neglect the derivatives with respect to xx. The nondimensional equations become

ε=2Re[(dVdy)2−13(dVdy)2]+1Re[dUdy]2(15)\varepsilon=\frac{2}{Re}\left[\left(\frac{dV}{dy}\right)^2-\frac{1}{3}\left(\frac{dV}{dy}\right)^2\right]+\frac{1}{Re}\left[\frac{dU}{dy}\right]^2 \tag*{(15)}

or

ε=43Re[dVdy]2+1Re[dUdy]2(16)\varepsilon=\frac{4}{3Re}\left[\frac{dV}{dy}\right]^2+\frac{1}{Re}\left[\frac{dU}{dy}\right]^2 \tag*{(16)}

Viscous Dissipation for Analytical Solution

In the case of turbulent flow, the velocity UU is provided in Fedoseyev [5] (Eq. (9),(10)).

The streamwise velocity is

U(y)=γUT(y)+(1−γ)UL(y)(17)U(y)=\gamma U^T(y)+(1-\gamma)U^L(y) \tag*{(17)}

and the transverse velocity is

V(y)=γVT(y)(18)V(y)=\gamma V^T(y) \tag*{(18)}

where ULU^L and UTU^T denote the laminar and turbulent/coherent contributions, respectively, and γ∈(0,1)\gamma\in(0,1) is a weighting parameter. Since the laminar solution satisfies VL=0V^L=0, only the turbulent/coherent component contributes to the transverse field.

For the simplest case of transverse velocity mode n=2n=2, and U0=1U_0=1, the analytical solutions were presented in Fedoseyev [5] in explicit form:

UT=1−exp⁡(1−ey/δ)(19)U^T=1-\exp\left(1-e^{y/\delta}\right) \tag*{(19)}

for y∈[0,0.5]y\in[0,0.5] and symmetrically in y∈[0.5,1]y\in[0.5,1], the VV-component of velocity

VT=−asin⁡2πy2πδ2Re(20)V^T=-\frac{a\sin2\pi y}{2\pi\delta^2Re} \tag*{(20)}

where aa is a coefficient we will vary from 1 to smaller and larger values, and the laminar velocity

UL=4y(1−y).(21)U^{L} = 4y(1-y). \tag*{(21)}

Let us estimate the viscous dissipation of every term in the Eq. (16). The velocity derivatives are

dUdy=[γδey/δ−ey/δ+1+(1−γ)4(1−2y)],(22)\frac{dU}{dy} = \left[\frac{\gamma}{\delta}e^{y/\delta-e^{y/\delta}+1} + (1-\gamma)4(1-2y)\right], \tag*{(22)}
dVdy=−aγycos⁡2πyδ2Re,(23)\frac{dV}{dy} = -\frac{a\gamma y\cos2\pi y}{\delta^{2}Re}, \tag*{(23)}
dULdy=4(1−2y).(24)\frac{dU^{L}}{dy} = 4(1-2y). \tag*{(24)}

For the dissipation in the channel volume formed by the channel width and a unit length along the channel, we integrate Eq. (16) for each of the terms:

DUstreamwise=1Re∫01∫01(dUdy)2dydx=1Re[γ2δ2I1(δ)+γ(1−γ)δI2(δ)+(1−γ)2I3],(25)DU_{\mathrm{streamwise}} = \frac{1}{Re}\int_{0}^{1}\int_{0}^{1}\left(\frac{dU}{dy}\right)^{2} dydx = \frac{1}{Re}\left[\frac{\gamma^{2}}{\delta^{2}}I_{1}(\delta)+\frac{\gamma(1-\gamma)}{\delta}I_{2}(\delta)+(1-\gamma)^{2}I_{3}\right], \tag*{(25)}
DV=1Re∫01∫01(dVdy)2dydx=a2γ2δ4Re3I4,(26)DV = \frac{1}{Re}\int_{0}^{1}\int_{0}^{1}\left(\frac{dV}{dy}\right)^{2} dydx = \frac{a^{2}\gamma^{2}}{\delta^{4}Re^{3}}I_{4}, \tag*{(26)}
DULAMI=1Re∫01∫01(dULdy)2dydx=1ReI3.(27)DU_{LAMI} = \frac{1}{Re}\int_{0}^{1}\int_{0}^{1}\left(\frac{dU^{L}}{dy}\right)^{2} dydx = \frac{1}{Re}I_{3}. \tag*{(27)}

Here

I1(δ)=∫01exp⁡[2(y/δ−ey/δ+1)]dy,(28)I_{1}(\delta) = \int_{0}^{1}\exp\left[2\left(y/\delta-e^{y/\delta}+1\right)\right]dy, \tag*{(28)}
I2(δ)=8∫01(1−2y)exp⁡(y/δ−ey/δ+1)dy,(29)I_{2}(\delta) = 8\int_{0}^{1}(1-2y)\exp\left(y/\delta-e^{y/\delta}+1\right)dy, \tag*{(29)}
I3=16∫01(1−2y)2dy=163,(30)I_{3} = 16\int_{0}^{1}(1-2y)^{2}dy = \frac{16}{3}, \tag*{(30)}
I4=∫01cos⁡2(2πy)dy=12.(31)I_{4} = \int_{0}^{1}\cos^{2}(2\pi y)dy = \frac{1}{2}. \tag*{(31)}

Figure 2 shows the Reynolds-number dependence of the viscous dissipation associated with the transverse turbulent velocity, VTV^{T}, the analytical AHE turbulent/coherent branch streamwise velocity, UU, and the laminar velocity component, ULU^{L}. The results are shown for γ=0.65\gamma=0.65 and two values of the AHE similarity parameter, δ=0.35\delta=0.35 and 0.40. The dissipation associated with the transverse turbulent velocity increases rapidly as the Reynolds number decreases and becomes comparable to the other dissipation contributions in the vicinity of Re≈103Re \approx10^{3}. The location of this balance changes with δ\delta, demonstrating the parameter dependence of the critical Reynolds number.

Figure 2. Viscous dissipation versus Reynolds number for γ=0.65\gamma=0.65, n=2n=2, and (a) δ=0.35\delta=0.35, (b) δ=0.40\delta=0.40.

The balance between laminar and turbulent dissipation determines whether the flow is predominantly laminar or turbulent. A more careful analysis is therefore needed.

Transverse Velocity Amplitude

The transverse velocity amplitude was chosen to produce a solution UU with the amplitude close to unity Fedoseyev [5]. It is at the lower end of the experimental amplitude range (0.01-0.025) observed for Re=15000Re=15000, δ=0.04\delta=0.04. The amplitude of VTV^{T} is about 0.0066 for this Reynolds number.

The streamwise velocity UTU^{T} is obtained as a solution for momentum equation Fedoseyev [5]:

VT(y)dUTdy−1Red2UTdy2=0.(32)V^{T}(y)\frac{dU^{T}}{dy}-\frac{1}{Re}\frac{d^{2}U^{T}}{dy^{2}}=0. \tag*{(32)}

Figure 3 shows two solutions for a=0.1a=0.1 and a=0.5a=0.5. Thus, for fixed ReRe and δ\delta, a finite transverse-velocity amplitude is required to obtain a nontrivial turbulent branch of the analytical solution.

The streamwise velocity $U^{T}$ and residual $F(x)$ for different transverse velocity amplitudes

Figure 3. The streamwise velocity UTU^{T} obtained for different transverse velocity amplitude (a) a=0.1a = 0.1; (b) a=0.5a = 0.5, for Re=15,000Re=15,000 and δ=0.04\delta= 0.04. Solution UTU^{T} is shown by the pink line, and F(x)F(x) is the residual of Eq. (32).

We also computed the solution UTU^T for a=1,2,4,8,20a=1,2,4,8,20. The results for the amplitudes of UTU^T are: A≈14.5,−5,2.8,11,60A \approx14.5,-5,2.8,11,60.

Similar computations for Re=1000Re=1000, for a=0.1,0.5,1,10,20a=0.1,0.5,1,10,20. The results are: A≈0,3E−08,−7.2,−5.5,11A \approx0,3E-08,-7.2,-5.5,11. For Re=1000Re=1000, the amplitude a=0.5a=0.5 is not enough to obtain a reasonable UTU^T, while amplitudes of 1 to 20 result in a similar solution. The sign of the resulting UTU^T does not affect the normalized mean-profile comparison, while its magnitude is subsequently absorbed into the normalization of the turbulent branch. Therefore, the reasonable range of amplitudes for VTV^T is 1 to 20.

Functional Minimization and the Critical Reynolds Number

We now determine the turbulent-flow weighting coefficient using a variational principle based on the minimization of the viscous dissipation functional associated with the analytical velocity field within the volume swept by the fluid per unit time Fedoseyev (2024b). The minimization principle and its justification were established in Fedoseyev (2024b); here we apply that principle to the present critical-Reynolds-number problem and extend the functional to include the transverse-velocity dissipation:

εT=1Uy(0)2∫0L0Uy2 dy∫0L0U dy.(33)\varepsilon_T=\frac{1}{U_y(0)^2}\int_0^{L_0} U_y^2\,dy\int_0^{L_0} U\,dy. \tag*{(33)}

Minimization of Eq. (33) with respect to γ\gamma determines the relative weight of the laminar and turbulent branches of the AHE solution, as established and supported by the comparisons with experimental velocity profiles in Fedoseyev (2024b).

The normalization by the square of the wall-normal velocity gradient at the wall removes the trivial dependence of the dissipation functional on the magnitude of the wall shear and provides a measure of dissipation relative to the shear scale of the corresponding velocity field.

The following logical step is to use a similar functional, but with the viscous dissipation to include the additional term related to VTV^T:

εT′=1Uy(0)2[DUstreamwise′+43DV′],(34)\varepsilon'_T=\frac{1}{U_y(0)^2}\left[DU'_{\mathrm{streamwise}}+\frac{4}{3}DV'\right], \tag*{(34)}

where the prime denotes the flow-weighted dissipation functional, with the additional factor ∫01U(y) dy\int_0^1 U(y)\,dy corresponding to the volume swept by the fluid per unit time:

DUstreamwise′=(1Re∫01U(y) dy)∫01(dUdy)2dy,(35)DU'_{\mathrm{streamwise}}=\left(\frac{1}{Re}\int_0^1 U(y)\,dy\right)\int_0^1\left(\frac{dU}{dy}\right)^2dy, \tag*{(35)}
DV′=(1Re∫01U(y) dy)∫01(dVdy)2dy.(36)DV'=\left(\frac{1}{Re}\int_0^1 U(y)\,dy\right)\int_0^1\left(\frac{dV}{dy}\right)^2dy. \tag*{(36)}

As the Reynolds number is reduced, the minimum of the functional shifts toward smaller γ\gamma. We define the practical critical Reynolds number RecRe_c as the Reynolds number at which the minimizing value satisfies γ≤0.02\gamma\leq0.02.

Channel flow experiment by Pasch, Leister, and Gatti et al.

A turbulent channel flow experiment reported in Pasch (2023) was conducted at Re=14 000Re=14\,000 for the full channel height, using air. The minimum of εT\varepsilon_T was γ=0.65\gamma= 0.65 for δ=0.027\delta= 0.027 (δ0≈0.7\delta_0 \approx0.7 mm), the obtained analytical solution is compared with the measured velocity profile in Figure 4(a). The agreement between the analytical solution and experimental data is within experimental uncertainty across the wall-normal range.

Figure 4. (a) Comparison of streamwise velocity in turbulent channel experiment Pasch (2023), Re=14000Re = 14000, (blue circles) and analytical solution UU with coefficient γ=0.65\gamma= 0.65 (red line), δ=0.027\delta= 0.027. (b) Minimization of εT′\varepsilon'_T is achieved at γ=0.07\gamma= 0.07 for Re=1500Re = 1500, a=10a = 10. (c) Streamwise velocity solutions at Re=1500Re = 1500 for γ=0.65\gamma= 0.65, a=0a = 0; γ=0.07\gamma= 0.07, a=10a = 10; and γ=0.02\gamma= 0.02, a=20a = 20.

Minimization of εT′\varepsilon'_T is achieved at γ=0.07\gamma= 0.07 for Re=1500Re = 1500, a=10a = 10, Figure 4(b). Streamwise velocity solutions at Re=1500Re = 1500 for different γ=0.65,0.07,0.02\gamma= 0.65, 0.07, 0.02 are shown in Figure 4(c). For γ=0.02\gamma= 0.02 the UU-soliton nearly coincides with the laminar solution ULU^L. Increasing the amplitude of the transverse component shifts the minimum toward smaller γ\gamma, reducing the contribution of the turbulent component to the resulting streamwise profile.

Wei and Willmarth channel flow experiments.

The experiments of Wei and Willmarth Wei and Willmarth (1989) Wei (1989) were conducted in turbulent channel flow over the range 2970≤Re≤395822970 \le Re \le39582 using distilled water.

Figure 5(a) shows the comparison between experimental data and the analytical solution at Re=2970Re = 2970 in coordinates (y+,U+)(y^+, U^+). The laminar component ULU^L (parabolic profile), the turbulent component UTU^T, and the combined solution UU are shown. The parameter values γ=0.65\gamma= 0.65 and δ=0.04\delta= 0.04 are used.

Figure 5. (a) Comparison of streamwise mean velocity from Wei and Willmarth (1989) experiments in (y+,U+)(y^+, U^+) coordinates at Re=2970Re = 2970 (circles) with UU (solution of the Alexeev hydrodynamic equations, red line). The laminar ULU^L (green line) and turbulent UTU^T (blue line) components of the superposition UU are plotted normalized to unity, so that each component reaches approximately U0U_0 at the channel centerline. (b) Minimization of εT′\varepsilon_T^{\prime} is achieved at γ=0.095\gamma= 0.095 for Re=1000Re = 1000, a=10a = 10. (c) Streamwise velocity solutions at Re=1000Re = 1000 for γ=0.65\gamma= 0.65, a=0a = 0; γ=0.095\gamma= 0.095, a=10a = 10; and γ=0.025\gamma= 0.025, a=20a = 20.

Minimization of εT′\varepsilon'_T is achieved at γ=0.095\gamma= 0.095 for Re=1000Re = 1000, a=10a = 10, Figure 5(b). Streamwise velocity solutions at Re=1000Re = 1000 for different γ=0.65,0.095,0.025\gamma= 0.65, 0.095, 0.025 are shown in Figure 5(c). For γ=0.025\gamma= 0.025 the UU-soliton nearly coincides with the laminar solution ULU^L. Increasing the amplitude of the transverse component shifts the minimum toward smaller γ\gamma, reducing the contribution of the turbulent component to the resulting streamwise profile.

Dependence on the AHE similarity parameter δ\delta

The simulation for different δ\delta are provided in Figure 6. Even for smaller amplitude a=1a = 1 the streamwise velocity is close to laminar within 2.5%2.5\%.

Figure 6. (a) Minimization of εT′\varepsilon_T' achieved at γ=0.025\gamma= 0.025 for Re=1000Re = 1000, δ=0.005\delta= 0.005, a=1a = 1. (b) Minimization of εT′\varepsilon_T' achieved at γ=0.580\gamma= 0.580 for Re=1000Re = 1000, δ=0.1\delta= 0.1, a=1a = 1.

A Reynolds-number sweep was performed for δ=0.025\delta= 0.025, a=10a = 10, and n=2n = 2. The computed values are

ReRe500600700800900100011001200130014001500
γ\gamma0.0050.0100.0150.0200.0200.0250.0350.0400.0450.0500.060

Table 1.

The resulting variation is shown in Figure 7. The Reynolds number is based on the centerline velocity. Using γ≤0.02\gamma\le0.02 as the practical criterion for disappearance of the turbulent branch gives Rec≈800Re_c \approx800–900 for this parameter set. The sensitivity to the practical cutoff can be estimated from the same Reynolds-number sweep: taking γc=0.015\gamma_c = 0.015, 0.020, and 0.025 gives Rec≈700Re_c \approx700, 800–900, and 1000, respectively. Thus, changing the cutoff within this range changes the estimated threshold by approximately 100–200 in Reynolds number, without changing its order of magnitude or the physical interpretation of the result.

Figure 7. Plot of γ\gamma versus ReRe, δ=0.025\delta= 0.025, a=10a = 10.

A systematic sweep in δ\delta was performed for a=10a = 10 and n=2n = 2, using δ=0.005\delta= 0.005, 0.01, 0.02, 0.04, 0.05, 0.08, and 0.1. The resulting values of ReRe at which γ≤0.02\gamma\le0.02 are

δ\delta0.0050.010.020.040.050.080.10
RecRe_c900032001200450320180140

The result is shown in Figure 8. The Reynolds number is based on the centerline velocity. Thus, for the present fixed values of aa and nn, the predicted threshold decreases strongly as δ\delta increases.

Figure 8. Plot of RecRe_c versus δ\delta, a=10a = 10.

Discussion

The calculations support an interpretation of the turbulent-to-laminar transition as the disappearance of the turbulent/coherent branch of the AHE solution when the Reynolds number is decreased. For the n=2n = 2 mode,

VT(y)=−asin⁡(2πy)2πδ2Re,V^T(y) = -\frac{a\sin(2\pi y)}{2\pi\delta^2 Re},

so the transverse velocity required to sustain the analytical turbulent component increases as 1/Re1/Re. Since viscous dissipation contains the square of the velocity gradient multiplied by 1/Re1/Re, the transverse contribution scales as

DV∝a2γ2δ4Re3.D_V \propto\frac{a^2\gamma^2}{\delta^4 Re^3}.

Thus, decreasing ReRe produces a rapidly increasing dissipation penalty for maintaining a nonzero transverse component.

The minimization of the flow-weighted dissipation functional provides a way of determining the relative weight γ\gamma of the two AHE branches. At the higher Reynolds numbers considered in the experimental comparisons, the minimum occurs near γ=0.65\gamma= 0.65, consistent with the velocity-profile comparisons for the Pasch and Wei–Willmarth experiments. At lower Reynolds numbers the additional transverse-velocity contribution changes the location of the minimum. For example, γ=0.07\gamma= 0.07 is obtained for Re=1500Re = 1500, a=10a = 10 in the Pasch parameter set, while γ=0.095\gamma= 0.095 is obtained for Re=1000Re = 1000, a=10a = 10 in the Wei–Willmarth parameter set. As γ\gamma becomes small, the resulting streamwise profile approaches the parabolic branch ULU^L.

For δ=0.025\delta= 0.025, a=10a = 10, and n=2n = 2, the Reynolds-number sweep gives γ≤0.02\gamma\le0.02 at Re=800Re = 800–900900. We use this value of γ\gamma as a practical numerical criterion rather than asserting that the numerical minimization reaches exactly γ=0\gamma= 0. The transition is therefore represented by the continuous weakening of the turbulent/coherent contribution in the present calculation.

The dependence on δ\delta is particularly strong. For a=10a = 10 and n=2n = 2, the calculated threshold decreases from Rec=9000Re_c = 9000 at δ=0.005\delta= 0.005 to Rec=140Re_c = 140 at δ=0.1\delta= 0.1. The present results therefore do not support a universal critical Reynolds number within the AHE formulation. Instead, the threshold depends on the characteristic AHE similarity parameter and, in general, on the transverse-velocity amplitude and mode number. The dependence on aa follows directly from the a2a^2 factor in the transverse dissipation, while the dependence on nn enters through the spatial structure and amplitude of VTV^T.

The present RecRe_c should be distinguished from the critical Reynolds number of the classical Orr–Sommerfeld stability problem. The latter concerns the onset of linear instability of an initially laminar Poiseuille solution as ReRe is increased. The present calculation addresses the opposite direction: it follows an already existing AHE turbulent/coherent solution as ReRe is decreased and determines when the turbulent contribution becomes negligible according to the dissipation-minimization criterion.

The experimental observations discussed in the Introduction, including the persistence of laminar flow at Reynolds numbers substantially above the conventional transition range when inlet disturbances are suppressed, demonstrate the strong dependence of laminar-to-turbulent transition on disturbance conditions. They are therefore not direct measurements of the present turbulent-to-laminar threshold. The comparison is used instead to emphasize that a transition threshold need not be a single universal Reynolds number.

The present analysis is deliberately limited to channel flow and to the parameter ranges investigated here. A systematic determination of the dependence of RecRe_c on the transverse mode number nn and disturbance amplitude aa, together with comparisons over a broader set of channel-flow experiments, remains a subject for future work.

Conclusions

An analytical criterion for the disappearance of the turbulent/coherent branch in channel flow has been developed within the Alexeev hydrodynamic equations. The streamwise velocity is represented as a superposition of the parabolic branch ULU^L and the nonlinear AHE branch UTU^T, while the latter is coupled to a transverse velocity VTV^T.

The additional transverse contribution to viscous dissipation is essential at decreasing Reynolds number. For the transverse mode considered here,

DV∝a2γ2δ4Re3,D_V \propto\frac{a^2\gamma^2}{\delta^4 Re^3},

so that the dissipation associated with maintaining the transverse structure increases rapidly as ReRe decreases. Functional minimization consequently shifts the optimal turbulent weight γ\gamma toward zero.

Using γ≤0.02\gamma\leq0.02 as a practical numerical criterion for disappearance of the turbulent branch, the representative calculations give RecRe_c of order 10310^3. The sensitivity test described above gives Rec≈700Re_c \approx700–10001000 when the cutoff is varied from γc=0.015\gamma_c = 0.015 to 0.0250.025. For δ=0.025\delta= 0.025, a=10a = 10, and n=2n = 2, the threshold is approximately Rec=800Re_c = 800–900900. A systematic variation of δ\delta at fixed a=10a = 10 and n=2n = 2 gives thresholds from Rec=9000Re_c = 9000 at δ=0.005\delta= 0.005 to Rec=140Re_c = 140 at δ=0.1\delta= 0.1.

The resulting threshold is therefore parameter dependent and should not be interpreted as a universal critical Reynolds number. The present criterion describes the disappearance of an already existing turbulent/coherent AHE branch as Reynolds number is decreased, rather than the linear instability of the laminar Poiseuille solution. It consequently addresses a different transition question from the classical Orr–Sommerfeld stability criterion.

The calculations support viscous dissipation of the transverse velocity component as a mechanism limiting the persistence of the AHE turbulent/coherent solution at sufficiently low Reynolds number. Further work can examine systematically the dependence of the threshold on the transverse mode number and disturbance amplitude.

References

  1. [1]ALEXEEV B.V. 1994 The generalized Boltzmann equation, generalized hydrodynamic equations and their applications. Phil. Trans. Roy. Soc. London, A. 349 (1994), 417-443.DOI
  2. [2]ALEXEEV B.V. 2004 Generalized Boltzmann Physical Kinetics, Elsevier.
  3. [4]VAN DOORNE C.W.H., WESTERWEEL J. 2007 Measurement of laminar, transitional and turbulent pipe flow using Stereoscopic-PIV, February 2007, Experiments in Fluids 42(2), DOI: 10.1007/s00348-006-0235-5.DOI
  4. [5]FĖDOSEYEV A. 2026 (a) Analytical kink-type solutions and streak-like structures in turbulent channel flow. Physics of Fluids 38, 085135, https://doi.org/10.1063/5.0337174
  5. [6]FĖDOSEYEV A. 2026 (b) Controlling turbulent boundary layer flow through the transverse velocity component. In Boundary Layer Flows – Control Techniques and Technological Applications (ed. Zambri Harum), IntechOpen. https://www.intechopen.com/online-first/1229297 (last accessed September 1, 2026).DOI
  6. [16]Reynolds, O. 1883 An experimental investigation of the circumstances which determine whether the motion of water shall be direct or sinuous, and the law of resistance in parallel channels, Phil. Trans. Roy. Soc. London, 174, 935-982DOI

Paper details

vlnc:W_tzck5gy2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents