Transient CFD & Unsteady Flow

What Is It?

Transient CFD solves the time-dependent flow equations, advancing the flow field through time in discrete steps. Unlike steady CFD, which seeks a time-independent solution, transient CFD captures unsteady phenomena — vortex shedding, pulsed flow, moving boundaries, rotating machinery, gusts, buffeting and transient thermal behaviour. The time-dependent flow field is resolved at each time step, and the solution history is saved for post-processing. Transient CFD is more expensive than steady CFD but is essential for flows that are inherently unsteady.

Why It Matters

Many engineering flows are inherently unsteady — a steady solver may converge to a mean-like solution that misses the physics that actually matters. Vortex shedding behind a bluff body produces unsteady forces that cause vibration and fatigue. Pulsed flow in an engine produces time-dependent pressure and temperature. Moving boundaries (valves, pistons) change the flow domain. Rotating machinery (pumps, turbines) produces periodic flow features. For these problems, steady CFD is not just less accurate — it is the wrong approach. It may not converge, or it may converge to a solution that does not represent the real physics.

A steady solver may converge to a mean-like answer while missing the physics that actually matters. Vortex shedding, pulsed flow, buffeting and moving boundaries are inherently unsteady. A converged steady solution does not mean the flow is steady — it may mean the solver has averaged out the unsteadiness.

When Steady CFD Is Inadequate

• Vortex shedding — periodic eddy shedding behind a bluff body; inherently unsteady
• Moving boundaries — valves, pistons, doors changing the flow domain
• Rotating machinery — pumps, turbines, fans; periodic flow features
• Pulsed flow — engine intake/exhaust; pulsed injection
• Buffeting — unsteady shock oscillation on transonic wings
• Flow separation — separation may be unsteady even if the boundary conditions are steady
• Transient thermal — start-up, shut-down, thermal soak
• Gusts and manoeuvres — time-varying free-stream conditions

Physical Time Step

In transient CFD, the time step is the physical time interval between solutions. It must be small enough to resolve the fastest time scale of interest — the vortex shedding period, the blade passing period, the pulse duration. If the time step is too large, the unsteady features are not resolved and the solution is inaccurate. The time step is related to the mesh size and the flow velocity through the CFL (Courant–Friedrichs–Lewy) condition — information should not travel more than one cell per time step.

CFL condition (transient CFD):

CFL  =  V · Δt / Δx  ≤  CFL_max

where:
V     =  local flow velocity
Δt    =  time step
Δx    =  local cell size

CFL_max depends on the time integration scheme:
  Explicit:  CFL ≤ 1
  Implicit:  CFL can be larger (but accuracy degrades if too large)

The time step must resolve the physical time scale of interest

Temporal Resolution and Averaging

The time step determines the temporal resolution — the smallest time scale that can be captured. For periodic phenomena (vortex shedding, blade passing), the time step should be small enough to provide adequate resolution within one period — typically 20–100 steps per period. For statistical averaging (LES, DES), the simulation must run for many periods after the initial transient to converge the statistics. The averaging period must be long enough that the mean quantities are stable — too short an averaging period produces unreliable statistics.

Statistical Stationarity and Periodic Response

For some unsteady flows, the solution reaches a statistically stationary state — the mean quantities are constant even though the instantaneous fields fluctuate. For periodic flows, the solution reaches a periodic state — the flow repeats with a fixed period. In both cases, the initial transient (the start-up of the simulation) must pass before the meaningful data is collected. The initial transient may take several flow-through times to dissipate. Collecting data during the initial transient produces incorrect statistics. The engineer must identify when the simulation has reached statistical stationarity or periodicity and collect data only after that point.

Moving Mesh and Sliding Mesh

For flows with moving boundaries — valves, pistons, rotating machinery — the mesh must move or deform to represent the changing geometry. Moving mesh (also called dynamic mesh) deforms the mesh to accommodate boundary motion. Sliding mesh is a specialised technique for rotating machinery — a rotating mesh block slides relative to a stationary mesh block, with the interface between them handled by a general grid interface. The sliding mesh approach captures the rotor-stator interaction — the periodic disturbance of the rotor on the stator flow — that a steady mixing-plane approach averages out.

Force Histories

One of the primary outputs of transient CFD is the force history — the time-varying aerodynamic force on the body. For vortex shedding, the force oscillates at the shedding frequency. For rotating machinery, the force has a periodic component at the blade passing frequency. For buffeting, the force is unsteady and may be random. The force history is used for structural analysis — the unsteady aerodynamic loads are applied to a structural model to predict vibration, fatigue and acoustic response. The force history must be resolved with adequate temporal resolution — if the time step is too large, the force peaks are missed.

Time-Step Selection from the Physics

A transient CFD time step should be chosen from the frequencies, convection times and moving-boundary events that must be resolved, not from solver convenience alone. Useful checks include cells traversed per step, Courant number where relevant to the numerical scheme, rotor or actuator motion per step, vortex-shedding period and the highest forcing frequency needed by a downstream structural analysis. A solution can remain numerically stable with a time step that is too large to reproduce the amplitude or phase of the physical event. Perform temporal refinement on a decision-driving output such as force amplitude, pressure spectrum, heat flux or phase lag, and separate inner-iteration convergence from time-step adequacy. If an unsteady load will be mapped into structural dynamics, the CFD sampling rate and structural frequency range must be compatible.

Statistical Convergence, Spectra & Phase-Resolved Outputs

Many unsteady flows never repeat exactly. For vortex shedding, turbulent wakes and rotating machinery, the engineering result may therefore be a mean, RMS value, spectrum, probability distribution or phase-averaged field rather than a single snapshot. The simulation should run long enough after initial transients for those statistics to become stable. Check running means, RMS levels and dominant spectral peaks over progressively longer windows. For periodic machinery, phase averaging can separate coherent blade-passing content from broadband turbulence. For broadband forcing, retain enough duration and sample rate to support the required spectral resolution. A visually convincing animation is not evidence of statistical convergence; the evidence is stability of the quantities used for design and substantiation.

Transient CFD is complete only when both time-step resolution and observation-window length are adequate for the engineering quantity being reported.

Key Takeaways

• Transient CFD resolves time-dependent flow — essential for inherently unsteady phenomena
• A steady solver may converge while missing the unsteady physics that actually matters
• The time step must resolve the fastest time scale of interest (CFL condition)
• Statistical averaging requires the initial transient to pass and sufficient averaging time
• Moving mesh and sliding mesh handle moving boundaries and rotating machinery