Dallas V, Vassilicos J C, Hewitt G F
Institute for Mathematical Sciences, Imperial College, London, United Kingdom.
Phys Rev E Stat Nonlin Soft Matter Phys. 2009 Oct;80(4 Pt 2):046306. doi: 10.1103/PhysRevE.80.046306. Epub 2009 Oct 6.
On the basis of various direct numerical simulations (DNS) of turbulent channel flows the following picture is proposed. (i) At a distance y from either wall, the Taylor microscale lambda is proportional to the average distance l(s) between stagnation points of the fluctuating velocity field, i.e., lambda(y)=B(1)l(s)(y) with B(1) constant, for delta(nu) << y < or approximately equal to delta, where the wall unit delta(nu) is defined as the ratio of kinematic viscosity nu to skin friction velocity u(tau) and delta is the channel's half-width. (ii) The number density n(s) of stagnation points varies with height according to n(s)=(C(s)/delta(nu)(3))y(+)(-1) where y(+) identical with y/delta(nu) and C(s) is constant in the range delta(nu) << y < or approximately equal to delta. (iii) In that same range, the kinetic energy dissipation rate per unit mass, equals 2/3(E(+)((u(tau)(3)/kappa(s)y) where E(+) is the total kinetic energy per unit mass normalized by u(tau)(2) and kappa(s)=B(1)(2)/C(s) is the stagnation point von Kármán coefficient. (iv) In the limit of exceedingly large Reynolds numbers Re(tau) identical with delta/delta(nu), large enough for the Reynolds stress -(uv) to equal u(tau)(2) in the range delta(nu) << y << delta, and assuming that production of turbulent kinetic energy balances dissipation locally in that range and limit, the mean velocity U(+), normalized by u(tau), obeys (d/dy)U(+) approximately equal to 2/3(E(+)/kappa(s)y) in that same range. (v) It follows that the von Kármán coefficient kappa is a meaningful and well-defined coefficient and the log law holds in turbulent channel/pipe flows only if E(+) is independent of y(+) and Re(tau) in that range, in which case kappa approximately kappa(s). (vi) In support of (d/dy)U(+) approximately equal to 2/3(E(+)/kappa(s)y), DNS data of turbulent channel flows which include the highest currently available values of Re(tau) are best fitted by E(+) approximately equal to 2/3(B(4)y(+)(-2/15)) and (d/dy(+))U(+) approximately equal to (B(4)/kappa(s))y(+)(-1-2/15) with B4 independent of y in delta(nu) << y << delta if the significant departure from -(uv) approximately equal to u(tau)(2) at these Re(tau) values is taken into account.
基于对湍流槽道流的各种直接数值模拟(DNS),提出了以下图景。(i)在距任一壁面距离为y处,泰勒微尺度λ与脉动速度场驻点之间的平均距离l(s)成正比,即对于δ(ν) << y ≤ δ,λ(y)=B(1)l(s)(y),其中B(1)为常数,这里壁面单位δ(ν)定义为运动粘度ν与摩擦速度u(τ)之比,δ为槽道半宽。(ii)驻点的数密度n(s)随高度变化规律为n(s)=(C(s)/δ(ν)(3))y(+)(-1),其中y(+) = y/δ(ν),且C(s)在δ(ν) << y ≤ δ范围内为常数。(iii)在同一范围内,单位质量的动能耗散率等于2/3(E(+)((u(τ)(3)/κ(s)y),其中E(+)是单位质量的总动能除以u(τ)(2)进行归一化,κ(s)=B(1)(2)/C(s)是驻点冯·卡门系数。(iv)在雷诺数Re(τ) = δ/δ(ν)极大的极限情况下,大到足以使雷诺应力-(uv)在δ(ν) << y << δ范围内等于u(τ)(2),并假设在该范围和极限内湍动能的产生在局部平衡耗散,则经u(τ)归一化的平均速度U(+)在同一范围内服从(d/dy)U(+) ≈ 2/3(E(+)/κ(s)y)。(v)由此可知,冯·卡门系数κ是一个有意义且定义明确的系数,并且对数律仅在该范围内E(+)与y(+)和Re(τ)无关时才适用于湍流槽道/管道流,在这种情况下κ ≈ κ(s)。(vi)为支持(d/dy)U(+) ≈ 2/3(E(+)/κ(s)y),如果考虑到在这些Re(τ)值下与-(uv) ≈ u(τ)(2)的显著偏差,则包含当前可用最高Re(τ)值的湍流槽道流的DNS数据最佳拟合为E(+) ≈ 2/3(B(4)y(+)(-2/15))和(d/dy(+))U(+) ≈ (B(4)/κ(s))y(+)(-1 - 2/15),其中B4在δ(ν) << y << δ范围内与y无关。