Geoscience Reference
In-Depth Information
cascade of kinetic energy from large scales to small ( Chapters 6 , 7 ). This cascade
dynamically “couples” all the eddies in a turbulent flow. Equation (1.35) shows
that the ratio of energy-containing and dissipative eddy sizes increases as the 3/4
power of the large-eddy Reynolds number R t , so a turbulent flow of large R t has
a huge number of interacting eddies. This has thwarted all attempts to solve the
turbulence equations analytically. Even at relatively small R t the “bookkeeping”
for these interactions overwhelms even our largest computers.
1.7 Numerical modeling of turbulent flows
Today's numerical calculations of large- R t turbulent flows (our daily weather fore-
casts, for example) do not use the basic fluid equations. Instead they use approximate
forms of the averaged equations first derived byOsborneReynolds ( 1895 ). Reynolds
averaged over a region of space surrounding a point; the average over an ensemble
of realizations of the flow was introduced later. For now we shall not be specific
about the type of averaging; it can be time, space, or ensemble averaging. If the
average commutes with differentiation (as we shall see, most averages do) we can
write the averaged form of the Navier-Stokes equation (1.26) as
∂u i
∂t +
∂u i u j
∂x j
1
ρ
∂p
∂x i ,
=−
(1.38)
with the overbar denoting the average. We have assumed the averaged viscous term
is negligible (Problem 1.8) . In writing (1.38) we have used the zero-diverg ence
property of u j to bring it into the derivative. B ut the averaging has created a u i u j
term in (1.38) ; in turbule nt flo w it differs from u i u j (Problem 1.17) and, hence, is
an unknown. If we write u i u j as
u i u j + u i u j
u i u j =
τ ij
ρ
u i u j
=
u i u j
,
(1.39)
the averaged equation (1.38) becomes
∂u i
∂t +
∂u i u j
∂x j
1
ρ
∂p
∂x i +
1
ρ
∂τ ij
∂x j .
=−
(1.40)
Averaging the Navier-Stokes equation has produced Eq. (1.40) for the average
velocity field, but the equation contains a new term involving a turbulent stress τ ij .
We'll see that this is a subtly different quantity for ensemble and space averages,
but in each case it is typically called the Reynolds stress . Experience has shown that
in turbulent flow it is always important.
 
Search WWH ::




Custom Search