The previous results were all derived using a Cartesian coordinate system.

We can use results

  dV

[1A-1]

Again we have a physical balance equation. The time rate of change in Φ equals the net inflow
plus the source. The net outflow, which is the negative of the net inflow, takes place through the
surface area, Σ, that encloses the volume Ω. The flow can take place by convection or through a
“diffusive” flux such as viscous momentum transport or conduction heat transfer. We can express
this verbal balance as the following integral equation.

 
t
  dV    v  ndS   d  ndS   S dV
t 
[1A-2]

In this equation, v is the velocity vector and n is a (dimensionless) unit normal vector that is
perpendicular to the surface and points outward. The dot products ρφv∙n and dφ∙n are the
convective and diffusive fluxes of Φ flowing out of the volume, Ω. 1 The source term, Sφ has the
same definition as in equation [1-5]. The surface integral can be converted to a volume integral
using the Gauss divergence theorem.

 div fdV

  f  ndA

[1A-3]

This equation applies to any vector, f. In particular, we can use this equation for the vector ρφv,
in equation [1A-2] to obtain the following result


t  dV

  div ( v )dV

  div (d )dV   S dV
 
[1A-4]

If we consider the limit of equation [1A-4] as the volume element, Ω, shrinks to zero, the
integrands in the equation must satisfy the overall equation. This gives the usual differential
equation.


 div( v )  div (d )  S [1A-5]
t

The terms in equations [1A-4] and [1A-5] do not depend on the specific coordinate system used.
Thus, they can apply to any coordinate system. The integral balance equation in [1A-4] is used
as the starting point for numerical methods in general coordinate systems.

1
Recall that the dot product v∙n gives the component of v in the direction of n. Since n is defined
to be normal to the surface, v∙n is the component of the velocity, v, that is in the direction leaving
the volume. The term ρφv∙n i, which has units of (phi units)/(area)/(time) represents a flux of Φ.
Integrating this flux over the entire surface enclosing the volume gives the net outflow, the
negative of which is the net inflow. A similar argument applies to the diffusive flux, dφ, which has
the same units as ρφv. Typically this flux is given as a gradient such as the heat flux, q, which is
given by Fourier’s Law as q = -kT.
Page 2 ME 692, L. S. Caretto, July, 31, 2019 CFD Equations

The mass conservation equation is the simplest one to consider. For this equation, φ = 1 and
there are no diffusive flux or source terms. The mass conservation or continuity versions of
equations [1A-4] and [1A-5] are then written as follows.


t  dV

  div( v )dV

0 [1A-6]

If we consider the limit of equation [1A-4] as the volume element, Ω, shrinks to zero, the
integrands in the equation must satisfy the overall equation. This gives the usual differential
equation.


 div( v )  0 [1A-7]
t

The diffusive flux vector is typically expressed as a gradient of some other function such as the
Fourier Law for heat transfer. (However, as noted above ion the discussion between equations
[1-59] and [1-60], the actual heat flux may be much more complex than the simple Fourier law.)
The previous discussion in the Cartesian coordinate system may be also reproduced in vector
from for the generalized coordinate system used here.

In equation [1-29] we defined the stress components, σ ij, for a Newtonian fluid, in terms of
Cartesian coordinates. We can express these nine components as a stress tensor, S. We first
define the deformation or rate-of strain tensor, which has the following components in Cartesian
coordinates.

1   ui  u j 

eij   [1A-8]
2  x j xi 


Using this definition, we can rewrite equation [1-29] in its tensor form for Cartesian coordinates as
follows

  11  12  13    p 0 0 
   
S   21  22  23    0  p 0  
   0  p 
 31  32  33   0
[1A-9]
 1 0 0  e11 e12 e13 
 2     
      0 1 0   2   e21 e22 e23 
 3    e 
 0 0 1  31 e32 e33 

In this equation we have displayed the stress tensor, S, using its Cartesian components. We can
write the general form of the stress tensor, for an arbitrary coordinate system, by defining the
identity tensor, I, and the deformation of rate-of-strain tensor, D. For any coordinate system, we
can write the “diffusive” flux for a Newtonian fluid as its stress tensor, using the following
equations.

  2  
S    p         I  2 D [1A-10]
  3  
 2
 
D  1 grad v   grad v 
T
 [1A-11]

In this definition, grad v is the gradient of a vector, which is a second order tensor. We will find it
convenient to separate the pressure term from those containing the viscosity. To this end we
define the viscous stress tensor, T, as follows.2

 2 
T       I  2 D so that S   pI  T [1A-12]
 3 

The stress and deformation tensors, written in equations [1A-10] to [1A-12] do not depend on the
coordinate system. We can write equation [1A-4] as a momentum balance equation in which we
regard the quantity φ as the velocity vector, v. This makes the equation a vector equation and the
term vv is called a dyadic product. It represents a second-order tensor that has nine
components. The integral momentum equation can be written for the vector velocity in two forms,
without and with the Gauss divergence theorem. Both results are shown below where B
represents the body force term divided by the density; i.e., B has dimensions of acceleration.


t  vdV

  vv  ndS   S  ndS   BdV
  
[1A-13]


t  vdV

  div ( vv )dV

  div(S)dV

  BdV

[1A-14]

For a general scalar equation, where the diffusive flux is given by the equation dφ = –Γ(φ)grad φ,
the integral balance equations become


 dV   v  ndS    grad  ndS   S dV
( )

t
[1A-15]
   


 dV  div ( v )dV  div (  grad ) dV   S dV
( )
  [1A-16]
t    

The integral conservation equations derived in this appendix are used in finite-volume
approaches to CFD. They are particularly useful in applications to complex geometries where a
simple rectangular coordinate system is not readily applicable.