Analysing Transient Flows

• Simplifying analysis to steady-state conditions is not always appropriate

‒ Aerodynamic vortex shedding, rotating machinery, free surface and bubble
dynamics in multiphase flows, reactive flows, unsteady heat transfer…

Suspended
proppant
transport in a Hydrogen
subsurface creation via
fracture methane
network pyrolysis
MECH3780 | Semester I, 2024 © The University of Queensland CRICOS code 00025B 28
Transient Analysis
• Transient analysis implies the generation of
sequential converged solutions separated
by a time step
‒ Multiple iterations within each time step
‒ Good convergence behaviour required,
just as in steady state analysis
‒ Monotonic residual change is ideal
• Choice of (implicit) time step will influence
convergence behaviour
‒ Too large → possibly diverging residuals
‒ Try to choose a time step that reduces
residuals by 2~3 orders of magnitude
‒ Flow physics also governs time step…

MECH3780 | Semester I, 2024 © The University of Queensland CRICOS code 00025B 29

Transient Time Step
• Time step must be small

QUANTITY OF INTEREST
enough to resolve important
time-dependent features
‒ Frequency of vortex shedding
‒ Rate of temperature change
‒ Vibration of structural parts
‒ Transport of particles TIME (s)

‒ Evolution of free surface QUANTITY OF INTEREST

‒ Wake development
• Time step must also be small
enough to maintain solver
stability

TIME (s)
Difference in prediction of flow dynamics with different time steps (LEAP Australia)
MECH3780 | Semester I, 2024 © The University of Queensland CRICOS code 00025B 30
Time Step Selection
• Courant number can be used to estimate time step (number of cells fluid passes through in one step)
Characteristic flow velocity × ∆𝑡 𝑣∆𝑡
Courant Number = = Fluent time-stepping options
Typical cell size ∆𝑥 Fixed:
• Good for general purpose
• Typical values are 1~10, but in exceptions this might be higher or lower
analysis
• For a characteristic length and velocity: ∆𝑡 = 𝐿 • Enter a fixed time step and
3𝑣 number of steps
Fixed-periodic:
1 𝑛 • For rotating machinery
• For turbomachinery: ∆𝑡 =
10 𝜔 • Enter rotation frequency,
steps per rotation, number of
𝐿 1 𝜕𝜌 rotations
• For natural convection: ∆𝑡 = 𝛽=−
𝑔𝛽∆𝑇𝐿 𝜌 𝜕𝑇 CFL Based:
• Good where time and
𝜌𝐶 velocity scales change during
• For conduction in solids: ∆𝑡 = 𝐿 the simulation
𝑘 • Enter a target Courant
number and end time, solver
adjusts time steps
MECH3780 | Semester I, 2024 © The University of Queensland CRICOS code 00025B 31
Initialisation and Execution of Transient Analysis
• A converged steady state solution is advised as
the starting point of a transient analysis

RESIDUALS
• If a transient simulation is started from an
approximate initial guess, the initial transient
results will not be accurate
‒ The first few time steps may not converge
‒ A smaller time step may be needed initially to
maintain solver stability
‒ Ignore the first few cycles of periodic flows
until a repeatable pattern is obtained
‒ Use approximately 20 iterations per time step
‒ PISO has much lower computational cost
than Coupled and converges faster for
unsteady flows than SIMPLE or SIMPLEC TIME (s)

MECH3780 | Semester I, 2024 © The University of Queensland CRICOS code 00025B 32

Lecture Summary
Key takeaways from the lecture…
• The SIMPLE algorithm is used to address
pressure-velocity coupling for steady-state
• PISO algorithm is designed for transient flows
→ greatly reduces cost per iteration
• BCs should be used to minimise the cost
(computational) of a CFD analysis
• Standard combination is velocity inlet and
pressure outlet – robust, good convergence
• Keep BCs remote from regions of interest!
• Symmetry and periodicity can be leveraged to
reduce the size of the required domain
• Transient time step selection is constrained by
flow dynamics and stability requirements Remains of a heat exchanger in the Fairlight wreck on Moreton Island
MECH3780 | Semester I, 2024 © The University of Queensland CRICOS code 00025B 33