The Finite Volume Method

2. Convection Diffusion problems

Examples
1. Convection Diffusion definition:

In problems where fluid flow plays a significant role, convection effects must be taken into consideration.
In nature, diffusion always manifests itself with convection
and for this we examine methods for predicting combined convection-diffusion.

putting a piece of sugar in a cup of tea

and mixing with a spoon

is a quite well convection-diffusion combination

2. Convection Diffusion formulation:

The steady-state convection-diffusion equation can be derived from the transport equation for a general property 𝜙
with the omission of the transient term.

𝑑𝑖𝑣 𝜌𝜙𝒖 = 𝑑𝑖𝑣 Γ 𝑔𝑟𝑎𝑑 𝜙 + 𝑆𝜙

Formal integration over a control volume gives:

න𝒏 𝜌𝜙𝒖 𝑑𝐴 = න𝒏 Γ 𝑔𝑟𝑎𝑑 𝜙 𝑑𝐴 + න𝑆𝜙 𝑑𝑉

Divergence theorem: which allows a volume integral to be changed into a surface integral
 3. One dimensional Steady State convection Diffusion:

In the absence of the source term, the convection-diffusion of a property 𝜙 conveyed with a one-dimensional flow field u
is governed by:

𝑑 𝑑 𝑑𝜙
𝑑𝑖𝑣 𝜌𝜙𝒖 = 𝑑𝑖𝑣 Γ 𝑔𝑟𝑎𝑑 𝜙 𝜌𝜙𝒖 = Γ
𝑑𝑥 𝑑𝑥 𝑑𝑥

The flow must satisfy the continuity equation:

𝑑𝑖𝑣 𝜌𝒖 = 0 𝑑
𝜌𝒖 = 0
𝑑𝑥
Integrating the transport equation over the control volume gives:

𝑑𝜙 𝑑𝜙
𝜌𝑢𝐴𝜙 𝑒 − 𝜌𝑢𝐴𝜙 𝑤 = ΓA − ΓA
𝑑𝑥 𝑒
𝑑𝑥 𝑤

Integrating the continuity equation gives: 𝜌𝑢𝐴 − 𝜌𝑢𝐴

𝑒 𝑤

The convective flux per unit area The diffusive flux per unit area
Γ
𝐹 = 𝜌𝑢 𝐷=
𝛿𝑥
 By adopting a centered approach for the diffusion coefficient and a regular mesh, we will have:

𝐹𝑒 𝝓𝒆 − 𝐹𝑤 𝝓𝒘 = 𝐷𝑒 𝜙𝐸 − 𝜙𝑃 − 𝐷𝑤 𝜙𝑃 − 𝜙𝑊

𝐹𝑒 − 𝐹𝑤 = 0

It is assumed that the flow field is known, which makes the evaluation of the convective flow 𝑭𝒆 𝒂𝒏𝒅 𝑭𝒘 possible.

To solve the previous equation, we need to calculate the transported property 𝝓𝒆 and 𝝓𝒘 to the respective faces e and w

Several discretization schemes are proposed for this :

 4. Centered Difference Scheme CDS a linear interpolation seems to be a logical remedy for evaluating 𝝓𝒆 and 𝝓𝒘 .

ϕ𝐸 + ϕ𝑃 ϕ𝑊 + ϕ𝑃
for a uniform mesh: ϕ𝑒 = ϕ𝑤 =
2 2

𝐹𝑒 𝜙𝑒 − 𝐹𝑤 𝜙𝑤 = 𝐷𝑒 𝜙𝐸 − 𝜙𝑃 − 𝐷𝑤 𝜙𝑃 − 𝜙𝑊

𝐹𝑒 𝐹𝑤
ϕ𝐸 + ϕ𝑃 − ϕ𝑊 + ϕ𝑃 = 𝐷𝑒 𝜙𝐸 − 𝜙𝑃 − 𝐷𝑤 𝜙𝑃 − 𝜙𝑊
2 2
𝐹𝑤 𝐹𝑒 𝐹𝑤 𝐹𝑒
𝐷𝑤 − + 𝐷𝑒 + ϕ𝑃 = 𝐷𝑤 + 𝜙𝑊 + 𝐷𝑒 − 𝜙𝐸
2 2 2 2
𝐹𝑤 𝐹𝑒 𝐹𝑤 𝐹𝑒
𝐷𝑤 + + 𝐷𝑒 − + 𝐹𝑒 − 𝐹𝑤 ϕ𝑃 = 𝐷𝑤 + 𝜙𝑊 + 𝐷𝑒 − 𝜙𝐸
2 2 2 2

𝑎𝑃 = 𝑎𝑊 + 𝑎𝐸 + 𝐹𝑒 − 𝐹𝑤 𝑎𝑃 𝜙𝑝 = 𝑎𝑊 𝜙𝑊 + 𝑎𝐸 𝜙𝐸

The previous equation is similar to that of pure diffusion, only the new coefficients contain additional terms taking into
account convection.

To solve a 1D convection-diffusion problem, it is necessary to construct a discretized equation for each node.
The resolution method remains the same as before.
CDS test

A property ϕ is transported by convection and diffusion through a 1D domain of length L. as boundary conditions
ϕ0 = 1, ϕ𝐿 = 0.

Using 5 nodes and the CDS scheme, test a speed of 0.1 m/s and 2.5 m/s

u = 0.1 m/s

Comparison of CDS solutions with the analytical solution

 5. Properties of discretization schemes

The failure of the centered scheme in certain cases involving the combination of convection and diffusion leads
us to explore in more detail the properties of discretization schemes.

Theoretically, the numerical solution is independent of the discretization scheme used when the number of
nodes considered tends towards infinity, whereas in practice this number is always finite.

A good discretization scheme must verify three essential properties

1 Conservativeness:

the flow leaving through the face e of a volume i must be equal to the flow entering through the face w of the volume i+1.

This property is generally satisfied by the finite volume method FVM

 2 Boundedness

Iterative numerical techniques are used to solve a large set of discretized equations for rach node.

To obtain a converged solution, Scarborough (1958) define a sufficient condition

𝑎′𝑃 = 𝑎𝑝 − 𝑆𝑝

+ Another essential condition : All coefficients of the discretized equations must have the same sign (generally positive).

3 Transportiveness 𝜌𝑢 𝑐𝑜𝑛𝑣𝑒𝑐𝑡𝑖𝑣𝑒 𝑓𝑙𝑢𝑥

𝑃𝑒 = 𝑃𝑒𝑐𝑙𝑒𝑡 𝑁𝑢𝑚𝑏𝑒𝑟 =
ΓΤ𝑑𝑥 𝑑𝑖𝑓𝑓𝑢𝑠𝑖𝑣𝑒 𝑓𝑙𝑢𝑥

No convection and pure diffusion 𝑃𝑒 = 0

No diffusion and pure convection 𝑃𝑒 → ∞

The property of transportivity is linked to the nature of the transport

translated by the Peclet number and the direction of the flow
 6. Assenssment of CDS scheme

- The CSD verifies the conservativeness property

- The limitation is not always verified since 𝑎𝐸 can become negative (case of a strongly convective flow (u =2.5 m/s, 5 nodes)

𝐹
For 𝑎𝐸 > 0 𝑃𝑒𝑒 = 𝐷𝑒 < 2
𝑒

if 𝑃𝑒𝑒 > 2 𝑎𝐸 < 0 Violation of the conditions of limitation

and can lead to impossible physical solutions

- The transportivity property is not verified by the CSD since the direction of the flow is not involved in the formulation
of the scheme.

- The truncation error of the Taylor series of the CDS scheme is of the second order,
but its instability for Pe > 2 makes it unacceptable.

For a given fluid (ρ, Γ= μ) The use of CDS remains limited to flows at low Re ( a dominance of diffusion (low speed) or
a fine mesh (dx too small).
7. Upwind Difference Scheme UDS

One of the major disadvantages of the centered CDS scheme is its inability to identify the direction of flow.

To ensure the transportivity property, the Upwind scheme UDS assigns to the facet of the control volume the
value of the variable at the center of the neighboring volume upstream of the flow direction

Upwind propose :
𝝓𝒘 = 𝝓𝑾 𝑎𝑛𝑑 𝝓𝒆 = 𝝓𝑷 positive flow direction, 𝑢𝑤 > 0, 𝑢𝑒 > 0

𝝓𝒘 = 𝝓𝑷 𝑎𝑛𝑑 𝝓𝒆 = 𝝓𝑬 negative flow direction, 𝑢𝑤 < 0, 𝑢𝑒 < 0

𝐹𝑒 𝜙𝑒 − 𝐹𝑤 𝜙𝑤 = 𝐷𝑒 𝜙𝐸 − 𝜙𝑃 − 𝐷𝑤 𝜙𝑃 − 𝜙𝑊

𝐹𝑒 − 𝐹𝑤 = 0
When the flow is in the positive direction, When the flow is in the negative direction,
𝑢𝑤 > 0, 𝑢𝑒 > 0 (𝐹𝑤 > 0, 𝐹𝑒 > 0) 𝑢𝑤 < 0, 𝑢𝑒 < 0 (𝐹𝑤 < 0, 𝐹𝑒 < 0)

Upwind propose 𝝓𝒘 = 𝝓𝑾 𝒂𝒏𝒅 𝝓𝒆 = 𝝓𝑷 Upwind propose 𝝓𝒘 = 𝝓𝑷 𝒂𝒏𝒅 𝝓𝒆 = 𝝓𝑬

𝐹𝑒 𝜙𝑃 − 𝐹𝑤 𝜙𝑊 = 𝐷𝑒 𝜙𝐸 − 𝜙𝑃 − 𝐷𝑤 𝜙𝑃 − 𝜙𝑊 𝐹𝑒 𝜙𝐸 − 𝐹𝑤 𝜙𝑃 = 𝐷𝑒 𝜙𝐸 − 𝜙𝑃 − 𝐷𝑤 𝜙𝑃 − 𝜙𝑊

(𝐷𝑤 + 𝐷𝑒 + 𝐹𝑒 )𝜙𝑃 = 𝐷𝑤 + 𝐹𝑤 𝜙𝑊 + 𝐷𝑒 𝜙𝐸
𝐷𝑤 + 𝐷𝑒 − 𝐹𝑒 + (𝐹𝑒 − 𝐹𝑤 ) 𝜙𝑃 = 𝐷𝑤 𝜙𝑊 + 𝐷𝑒 − 𝐹𝑒 𝜙𝐸
𝐷𝑤 + 𝐹𝑤 + 𝐷𝑒 + (𝐹𝑒 − 𝐹𝑤 ) 𝜙𝑃 = 𝐷𝑤 + 𝐹𝑤 𝜙𝑊 + 𝐷𝑒 𝜙𝐸

𝑎𝑃 𝜙𝑝 = 𝑎𝑊 𝜙𝑊 + 𝑎𝐸 𝜙𝐸 𝑎𝑃 = 𝑎𝐸 + 𝑎𝑊 + 𝐹𝑒 − 𝐹𝑤
 𝐹𝑤 > 0, 𝐹𝑒 > 0 𝑎𝑊 = 𝐷𝑤 + 𝐹𝑤 𝑎𝐸 = 𝐷𝑒

𝐹𝑤 < 0, 𝐹𝑒 < 0 𝑎𝑊 = 𝐷𝑤 𝑎𝐸 = 𝐷𝑒 − 𝐹𝑒

𝑎𝑃 𝜙𝑝 = 𝑎𝑊 𝜙𝑊 + 𝑎𝐸 𝜙𝐸

𝑎𝑃 = 𝑎𝐸 + 𝑎𝑊 + 𝐹𝑒 − 𝐹𝑤

A common notation covering both directions of flow 𝑎𝑊 = 𝐷𝑤 + max(𝐹𝑤 , 0) 𝑎𝐸 = 𝐷𝑒 + max(0, −𝐹𝑒 )

 UDS 1er order TEST

The CDS failed to reproduce a reasonable result with the same mesh resolution. The USD provides more realistic
solutions with less accuracy near the B boundary.

The formulation of the USD is conservative, the coefficients are positive, hence the limitatibility and transportivity is
also ensured. Only the precision of USD, it is of order 1, which is insufficient

The major disadvantage of UDS is its erroneous results when the flow is not aligned with the mesh lines

The resulting error has the appearance of diffusion and is called false diffusion (numerical)
To identifie the false diffusion due to the UDS scheme
Consider a 2D convection diffusion problem with u = v = 2m/s and the boundary conditions assigned in the figure

consider pure convection without physical diffusion. In the absence of physical diffusion, the exact solution exhibits
No source term for ϕ and a stationary solution a sudden jump from 100 to 0 at the intersection of the two
diagonals. Numerical results show bad profiles
As the flow is parallel to the diagonal solid line, the
The error is obvious for the coarse mesh, a refinement of the
value of ϕ in all nodes above the diagonal must be
mesh tends to eliminate this false diffusion
100 and zero below
So the UDS scheme is not entirely desirable for calculations that require high precision.
 8. Hybrid Difference Scheme HDS

Splading's (1972) hybridUDS schema is based on the combination of the CDS and UDS schemas.

The CDS with 2nd order precision is used for low Peclet numbers (Pe < 2)
The UDS 1st order precision is used for large Peclet numbers (Pe ≥ 2).

For Pe | < 2: HDS is equivalent to CDS for the convection and diffusion terms
For |Pe | ˃ 2: HDS is equivalent to UDS for the convection terms and the diffusion terms = 0.

𝑎𝑃 𝜙𝑝 = 𝑎𝑊 𝜙𝑊 + 𝑎𝐸 𝜙𝐸 𝑎𝑃 = 𝑎𝐸 + 𝑎𝑊 + 𝐹𝑒 − 𝐹𝑤

𝐹𝑤 𝐹𝑒
𝑎𝑊 = 𝑚𝑎𝑥 𝐹𝑤 , 𝐷𝑤 + ,0 𝑎𝐸 = 𝑚𝑎𝑥 −𝐹𝑒 , 𝐷𝑒 − ,0
2 2
 HSD TEST

• The scheme is completely conservative

• Since its coefficients are always positive, it is unconditionally

bounded (limitatibility).

• It satisfies the transportivity condition using UDS for high Pe.

• The HDS scheme is very stable compared to high-order

schemes and provides realistic physical solutions.

• The disadvantage of HDS is its accuracy in terms of truncation

error of 1st order Taylor series.
 1.9 HDS scheme in multi-dimensional convection-diffusion

𝑎𝑃 𝜙𝑝 = 𝑎𝐸 𝜙𝐸 + 𝑎𝑊 𝜙𝑊 + 𝑎𝑁 𝜙𝑁 + 𝑎𝑆 𝜙𝑆 + 𝑎 𝑇 𝜙𝑇 + 𝑎𝐵 𝜙𝐵 𝑎𝑃 = 𝑎𝐸 + 𝑎𝑊 + 𝑎𝑁 + 𝑎𝑆 + 𝑎 𝑇 + 𝑎𝐵 + ∆𝐹

1D 2D 3D

𝐹𝑤 𝐹𝑤 𝐹𝑤
𝑎𝑊 𝑚𝑎𝑥 𝐹𝑤 , 𝐷𝑤 + ,0 𝑚𝑎𝑥 𝐹𝑤 , 𝐷𝑤 + ,0 𝑚𝑎𝑥 𝐹𝑤 , 𝐷𝑤 + ,0
2 2 2

𝐹𝑒 𝐹𝑒 𝐹𝑒
𝑎𝐸 𝑚𝑎𝑥 −𝐹𝑒 , 𝐷𝑒 − ,0 𝑚𝑎𝑥 −𝐹𝑒 , 𝐷𝑒 − ,0 𝑚𝑎𝑥 −𝐹𝑒 , 𝐷𝑒 − ,0
2 2 2
Face w e s n b t
- 𝐹𝑠 𝐹𝑠
𝑎𝑆 𝑚𝑎𝑥 𝐹𝑠 , 𝐷𝑠 + ,0 𝑚𝑎𝑥 𝐹𝑠 , 𝐷𝑠 + ,0
2 2 F 𝜌𝑢 𝑤 𝐴𝑤 𝜌𝑢 𝑒 𝐴𝑒 𝜌𝑢 𝑠 𝐴𝑠 𝜌𝑢 𝑛 𝐴𝑛 𝜌𝑢 𝑏 𝐴𝑏 𝜌𝑢 𝑡 𝐴𝑡

- 𝐹𝑛 𝐹𝑛
𝑎𝑁 𝑚𝑎𝑥 −𝐹𝑛 , 𝐷𝑛 − ,0 𝑚𝑎𝑥 −𝐹𝑛 , 𝐷𝑛 − ,0
2 2 Γ𝑤 Γ𝑒 Γ𝑠 Γ𝑛 Γ𝑏 Γ𝑡
𝐴 𝐴 𝐴 𝐴 𝐴 𝐴
D 𝛿𝑥𝑊𝑃 𝑤 𝛿𝑥𝑃𝐸 𝑒 𝛿𝑦𝑆𝑃 𝑠 𝛿𝑦𝑃𝑁 𝑛 𝛿𝑧𝐵𝑃 𝑏 𝛿𝑧𝑃𝑇 𝑡

- - 𝐹𝑏
𝑎𝐵 𝑚𝑎𝑥 𝐹𝑏 , 𝐷𝑏 + ,0
2

- - 𝐹𝑡
𝑎𝑇 𝑚𝑎𝑥 −𝐹𝑡 , 𝐷𝑡 − ,0
2

∆𝐹 𝐹𝑒 − 𝐹𝑤 𝐹𝑒 − 𝐹𝑤 + 𝐹𝑛 −𝐹𝑠 𝐹𝑒 − 𝐹𝑤 + 𝐹𝑛 −𝐹𝑠 + 𝐹𝑡 − 𝐹𝑏