The Finite Volume Method

3. Unsteady Flows
 1. Formulation of unsteady convection-diffusion

𝜕
𝜌𝜙 + 𝑑𝑖𝑣 𝜌𝜙𝒖 = 𝑑𝑖𝑣 Γ 𝑔𝑟𝑎𝑑 𝜙 + 𝑆𝜙
𝜕𝑡

Rate of change term, = 0 for stationary flows

To predict transient problems, this term, must be retained in the discretization process.

the FV intergation of this equation over a CV must be augmented with a further intergation over a finite time step Δt

Steady-state convection-diffusion න𝒏 𝜌𝜙𝒖 𝑑𝐴 = න𝒏 Γ 𝑔𝑟𝑎𝑑 𝜙 𝑑𝐴 + න𝑆𝜙 𝑑𝑉

Unsteady-state :
𝑡+∆𝑡 𝑡+∆𝑡 𝑡+∆𝑡 𝑡+∆𝑡
𝜕
න න 𝜌𝜙 𝑑𝑡 𝑑𝑉 + න න 𝑛. 𝜌𝜙𝒖 𝑑𝐴 𝑑𝑡 = න න 𝑛. Γ 𝑔𝑟𝑎𝑑 𝜙 𝑑𝐴 𝑑𝑡 + න න 𝑆𝜙 𝑑𝑉 𝑑𝑡
𝐶𝑉 𝑡 𝜕𝑡 𝑡 𝐴 𝑡 𝐴 𝑡 𝐶𝑉
2. Unsteady 1D Diffusion
𝑑𝑇 𝑑 𝑑𝑇
Unsteady one-dimensional heat conduction : 𝜌𝑐 = 𝑘 +𝑆
𝑑𝑡 𝑑𝑥 𝑑𝑥

𝑡+∆𝑡 𝑡+∆𝑡 𝑡+∆𝑡

𝜕𝑇 𝜕 𝜕𝑇
න න 𝜌𝑐 𝑑𝑉 𝑑𝑡 = න න 𝑘 𝑑𝑉 𝑑𝑡 + න න 𝑆 𝑑𝑉 𝑑𝑡
𝑡 𝐶𝑉 𝜕𝑡 𝑡 𝐶𝑉 𝜕𝑥 𝜕𝑥 𝑡 𝐶𝑉

𝑒 𝑡+∆𝑡 𝑡+∆𝑡 𝑡+∆𝑡

𝜕𝑇 𝜕𝑇 𝜕𝑇
න න 𝜌𝑐 𝑑𝑡 𝑑𝑉 = න 𝑘𝐴 − 𝑘𝐴 𝑑𝑡 + න 𝑆 ∆𝑉 𝑑𝑡
𝑤 𝑡 𝜕𝑡 𝑡 𝜕𝑥 𝑒
𝜕𝑥 𝑤 𝑡

𝑡+∆𝑡 𝑇𝑃0 at time t

𝜕𝑇
න න 𝜌𝑐 𝑑𝑡 𝑑𝑉 = 𝜌𝑐 𝑇𝑝 − 𝑇𝑃0 ∆𝑉
𝐶𝑉 𝑡 𝜕𝑡
𝑇𝑝 at time 𝑡 + ∆𝑡

𝜕𝑇
= 𝑇𝑝 − 𝑇𝑃0 Τ∆ 𝑡 1st order backward discretization scheme
𝜕𝑡
 𝑡+∆𝑡 𝑡+∆𝑡
𝑇𝐸 − 𝑇𝑃 𝑇𝑃 − 𝑇𝑊
𝜌𝑐 𝑇𝑝 − 𝑇𝑃0 ∆𝑉 = න 𝑘𝑒 𝐴 − 𝑘𝑤 𝐴 𝑑𝑡 + න 𝑆 ∆𝑉 𝑑𝑡
𝑡 𝛿𝑥𝑃𝐸 𝛿𝑥𝑊𝑃 𝑡

Now, we need to consider assumptions regarding the variation of 𝑇𝑃 , 𝑇𝐸 , 𝑇𝑊 in time

𝑡+∆𝑡
𝐼𝑇 = න 𝑇𝑃 𝑑𝑡 = 𝜃 𝑇𝑃 + 1 − 𝜃 𝑇𝑃0 ∆𝑡
𝑡

𝜃 0 1/2 1
𝐼𝑇 𝑇𝑃0 ∆𝑡 1 0 𝑇𝑃 ∆𝑡
(𝑇 + 𝑇𝑃 )∆𝑡
2 𝑃

𝑇𝑝 − 𝑇𝑃0 𝑇𝐸 − 𝑇𝑃 𝑇𝑃 − 𝑇𝑊 𝑇𝐸0 − 𝑇𝑃0 𝑇𝑃0 − 𝑇𝑊

0
𝜌𝑐 ∆𝑥 = 𝜃 𝑘𝑒 − 𝑘𝑤 + 1−𝜃 𝑘𝑒 − 𝑘𝑤 + 𝑆 ∆𝑥
∆𝑡 𝛿𝑥𝑃𝐸 𝛿𝑥𝑊𝑃 𝛿𝑥𝑃𝐸 𝛿𝑥𝑊𝑃
 ∆𝑥 𝑘𝑒 𝑘𝑤 𝑘𝑒 0
𝑘𝑤 0
𝜌𝑐 +𝜃 + 𝑇𝑃 = 𝜃𝑇𝐸 + 1 − 𝜃 𝑇𝐸 + 𝜃𝑇𝑊 + 1 − 𝜃 𝑇𝑊
∆𝑡 𝛿𝑥𝑃𝐸 𝛿𝑥𝑃𝐸 𝛿𝑥𝑃𝐸 𝛿𝑥𝑊𝑃
∆𝑥 𝑘𝑒 𝑘𝑤
+ 𝜌𝑐 − 1−𝜃 − 1−𝜃 𝑇𝑃0 + 𝑆 ∆𝑥
∆𝑡 𝛿𝑥𝑃𝐸 𝛿𝑥𝑊𝑃

0
𝑎𝑃 𝑇𝑝 = 𝑎𝑊 𝜃 𝑇𝑊 + 1 − 𝜃 𝑇𝑊 + 𝑎𝐸 𝜃 𝑇𝐸 + 1 − 𝜃 𝑇𝐸0 + 𝑎𝑃0 − 1 − 𝜃 𝑎𝑊 − 1 − 𝜃 𝑎𝐸 𝑇𝑃0 + 𝑏

𝑎𝑃 = 𝜃(𝑎𝑊 + 𝑎𝐸 ) + 𝑎𝑃0

∆𝑥 𝑘 𝑘
𝑎𝑃0 = 𝜌𝑐 ∆𝑡 𝑎𝐸 = 𝛿𝑥 𝑒 𝑎𝑤 = 𝛿𝑥 𝑤 b = 𝑆 ∆𝑥
𝑃𝐸 𝑊𝑃

The exact form of the final discretized equation depends on the value of 𝜃

𝜃 = 0, we only use 𝑇𝑃0 , 𝑇𝑊

0
, 𝑇𝐸0 at time t into the right side of the equation to evaluate 𝑇𝑃 at the new time. This scheme is called
explicit scheme.

0 < 𝜃 ≤ 1 temperatures at the new time are used in both sides of the equation, this is an implicit scheme

𝜃 = 1 defines a fully implicit scheme

𝜃 = 0.5 Crank-Nicolson scheme (1947)

3. explicit scheme 𝜃=0

0
𝑎𝑃 𝑇𝑝 = 𝑎𝑊 𝑇𝑊 + 𝑎𝐸 𝑇𝐸0 + 𝑎𝑃0 − (𝑎𝑊 + 𝑎𝐸 − 𝑆𝑃 𝑇𝑃0 + 𝑆𝑢

∆𝑥 𝑘 𝑘
𝑎𝑃 = 𝑎𝑃0 𝑎𝑃0 = 𝜌𝑐 ∆𝑡 𝑎𝐸 = 𝛿𝑥 𝑒 𝑎𝑤 = 𝛿𝑥 𝑤
𝑃𝐸 𝑊𝑃

for the coefficients to be positive, it is necessary that 𝑎𝑃0 − 𝑎𝑊 − 𝑎𝐸 > 0

for constant k and a uniform mesh, 𝛿𝑥𝑃𝐸 = 𝛿𝑥𝑊𝑃 = ∆𝑥

∆𝑥 2𝑘 ∆𝑥 2
𝜌𝑐 > 𝑜𝑢 ∆𝑡 < 𝜌𝑐
∆𝑡 ∆𝑥 𝟐𝑘

This inequality sets a maximum limit of the time step and represents a serious limitation for the explicit scheme.

It is becoming too expensive to improve spatial precision ∆𝑡 max requires reduction ∆𝑥2

Therefore, this method is not recommended for transient problems.

 4. Crank-Nicolson scheme (𝜃 = 1/2)

𝑇𝐸 + 𝑇𝐸0 0
𝑇𝑊 + 𝑇𝑊 0
𝑎𝐸 𝑎𝑊 0
𝑎𝑃 𝑇𝑝 = 𝑎𝐸 + 𝑎𝑊 + 𝑎𝑃 − − 𝑇 +𝑏
2 2 2 2 𝑃

1 1
𝑎𝑃 = 𝑎𝑃0 + 𝑎𝑊 + 𝑎𝐸 − 𝑆𝑃
2 2
∆𝑥 𝑘 𝑘 1
𝑎𝑃0 = 𝜌𝑐 ∆𝑡 𝑎𝐸 = 𝛿𝑥 𝑒 𝑎𝑤 = 𝛿𝑥 𝑤 𝑏 = 𝑆𝑢 + 2 𝑆𝑃 𝑇𝑃0
𝑃𝐸 𝑊𝑃

more than a single unknown value of T at the new time is present in the equation,
the method is implicit and the simultaneous equations for all nodes need to be solved at each time step

0.5 ≤ 𝜃 ≤ 1 unconditionally stable Tous les coefficients ≥ 0 for a realistic physical solution
and bounded results
∆𝑥 2
𝑎𝑊 + 𝑎𝐸
𝑎𝑃0 ≥ ∆𝑡 < 𝜌𝑐 Less severe limitation than that for the explicit diagram
2 𝑘

Crank-Nicolson scheme is based on a centered difference and consequently a sufficiently small 2nd order time precision ∆𝑡,
it is possible to have better precision than the explicit method
5. Implicit scheme (θ = 1)

𝑎𝑃 𝑇𝑝 = 𝑎𝑊 𝑇𝑊 + 𝑎𝐸 𝑇𝐸 + 𝑎𝑃0 𝑇𝑃0 + 𝑆𝑢

𝑎𝑃 = 𝑎𝑃0 + 𝑎𝑊 + 𝑎𝐸 − 𝑆𝑃

∆𝑥 𝑘 𝑘
𝑎𝑃0 = 𝜌𝑐 ∆𝑡 𝑎𝐸 = 𝛿𝑥 𝑒 𝑎𝑤 = 𝛿𝑥 𝑤
𝑃𝐸 𝑊𝑃

Both sides of the equation contain the temperatures at the new time,

The system of algebraic equations must be solved at each time step, after the solution T is assigned to T0 and the
procedure is repeated for the other remaining time steps.

The time advance procedure begins with an initial temperature field T0

All coefficients are positive, Implicit scheme unconditionally stable for any Δt

Accuracy of the scheme is 1st order, a small time step is suggested to ensure accurate results.
 6. Implicit scheme applied in 2D and 3D

The implicit method (θ = 1) is recommended for CFD calculations for its superior stability.

𝑑𝜙 𝑑 𝑑𝜙 𝑑 𝑑𝜙 𝑑 𝑑𝜙
3D unsteady diffusion 𝜌𝑐 = Γ + Γ + Γ +𝑆
𝑑𝑡 𝑑𝑥 𝑑𝑥 𝑑𝑦 𝑑𝑦 𝑑𝑧 𝑑𝑧

𝑎𝑃 𝑇𝑝 = 𝑎𝑊 𝑇𝑊 + 𝑎𝐸 𝑇𝐸 + 𝑎𝑆 𝜙𝑆 + 𝑎𝑁 𝜙𝑁 + 𝑎𝐵 𝜙𝐵 + 𝑎 𝑇 𝜙 𝑇 + 𝑎𝑃0 𝑇𝑃0 + 𝑆𝑢

∆𝑥
𝑎𝑃 = 𝑎𝐸 + 𝑎𝑊 + 𝑎𝑁 + 𝑎𝑆 + 𝑎 𝑇 + 𝑎𝐵 + 𝑎𝑃0 − 𝑆𝑝 𝑎𝑃0 = 𝜌𝑐
∆𝑡
𝑎𝑊 𝑎𝐸 𝑎𝑆 𝑎𝑁 𝑎𝐵 𝑎𝑇

Γ𝑤 Γ𝑒 1D 2D 3D
1D - - - -
𝐴 𝐴 ∆𝑉 ∆𝑥 ∆𝑥∆𝑦 ∆𝑥∆𝑦∆𝑧
𝛿𝑥𝑊𝑃 𝑤 𝛿𝑥𝑃𝐸 𝑒
𝐴 𝑒 = 𝐴𝑤 1 ∆𝑦 ∆𝑦∆𝑧
2D Γ𝑤 Γ𝑒 Γ𝑠 Γ𝑛 - -
𝐴𝑤 𝐴𝑒 𝐴𝑠 𝐴 𝐴𝑛 = 𝐴𝑠 - ∆𝑥 ∆𝑥∆𝑧
𝛿𝑥𝑊𝑃 𝛿𝑥𝑃𝐸 𝛿𝑦𝑆𝑃 𝛿𝑦𝑃𝑁 𝑛
𝐴 𝑡 = 𝐴𝑏 - - ∆𝑥∆𝑦
3D Γ𝑤 Γ𝑒 Γ𝑠 Γ𝑛 Γ𝑏 Γ𝑡
𝐴 𝐴 𝐴 𝐴 𝐴 𝐴
𝛿𝑥𝑊𝑃 𝑤 𝛿𝑥𝑃𝐸 𝑒 𝛿𝑦𝑆𝑃 𝑠 𝛿𝑦𝑃𝑁 𝑛 𝛿𝑧𝐵𝑃 𝑏 𝛿𝑧𝑃𝑇 𝑡
 7. Discretization of the unsteady convection-diffusion equation

𝜕
𝜌𝜙 + 𝑑𝑖𝑣 𝜌𝜙𝒖 = 𝑑𝑖𝑣 Γ 𝑔𝑟𝑎𝑑 𝜙 + 𝑆𝜙
𝜕𝑡

The HDS hybrid spatial discretization scheme is recommended for its preferred stability for the treatment of convection terms,
we present the implicit/hybrid scheme

𝜕 𝜕 𝜕 𝜕 𝜕 𝜕𝜙 𝜕 𝜕𝜙 𝜕 𝜕𝜙
𝜌𝜙 + 𝜌𝑢𝜙 + 𝜌𝑣𝜙 + 𝜌𝑤𝜙 = Γ + Γ + Γ +𝑠
𝜕𝑡 𝜕𝑥 𝜕𝑦 𝜕𝑧 𝜕𝑥 𝜕𝑥 𝜕𝑦 𝜕𝑦 𝜕𝑧 𝜕𝑧

𝑎𝑃 𝑇𝑝 = 𝑎𝑊 𝑇𝑊 + 𝑎𝐸 𝑇𝐸 + 𝑎𝑆 𝜙𝑆 + 𝑎𝑁 𝜙𝑁 + 𝑎𝐵 𝜙𝐵 + 𝑎 𝑇 𝜙𝑇 + 𝑎𝑃0 𝑇𝑃0 + 𝑆𝑢

𝑎𝑃 = 𝑎𝐸 + 𝑎𝑊 + 𝑎𝑁 + 𝑎𝑆 + 𝑎 𝑇 + 𝑎𝐵 + 𝑎𝑃0 + ∆𝐹 − 𝑆𝑝

∆𝑉
𝑎𝑃0 = 𝜌𝑃0 𝑐 𝑆∆𝑉 = 𝑆𝑢 + 𝑆𝑝 𝜙𝑃
∆𝑡
 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
- 𝐹𝑠 𝐹𝑠
𝑎𝑆 𝑚𝑎𝑥 𝐹𝑠 , 𝐷𝑠 + ,0 𝑚𝑎𝑥 𝐹𝑠 , 𝐷𝑠 + ,0
2 2
- 𝐹𝑛 𝐹𝑛
𝑎𝑁 𝑚𝑎𝑥 −𝐹𝑛 , 𝐷𝑛 − ,0 𝑚𝑎𝑥 −𝐹𝑛 , 𝐷𝑛 − ,0
2 2

- - 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

Face w e s n b t

F 𝜌𝑢 𝑤 𝐴𝑤 𝜌𝑢 𝑒 𝐴𝑒 𝜌𝑢 𝑠 𝐴𝑠 𝜌𝑢 𝑛 𝐴𝑛 𝜌𝑢 𝑏 𝐴𝑏 𝜌𝑢 𝑡 𝐴𝑡

Γ𝑤 Γ𝑒 Γ𝑠 Γ𝑛 Γ𝑏 Γ𝑡
D 𝐴 𝐴 𝐴 𝐴 𝐴 𝐴
𝛿𝑥𝑊𝑃 𝑤 𝛿𝑥𝑃𝐸 𝑒 𝛿𝑦𝑆𝑃 𝑠 𝛿𝑦𝑃𝑁 𝑛 𝛿𝑧𝐵𝑃 𝑏 𝛿𝑧𝑃𝑇 𝑡
8. Summary

Scheme Stability Accurancy Positive coeff

criterion
Explicit Conditionally stable 1st order ∆𝑡 < 𝜌𝑐 ∆𝑥 2 Τ2𝑘

Crank-Nicolson Unconditionally stable 2nd order ∆𝑡 < 𝜌𝑐 ∆𝑥 2 Τ𝑘

Implicit Unconditionally stable 1er ordre Always positive