In computational fluid dynamics (CFD), simulating incompressible or low-speed fluid flow presents a fundamental mathematical headache: pressure has no explicit evolution equation.
While the Navier-Stokes equations provide a clear prognostic equation for velocity through momentum conservation, the continuity equation ($\nabla \cdot \mathbf{u} = 0$) acts purely as a kinematic constraint. It states that the velocity field must remain divergence-free to preserve mass, but it does not tell us how pressure changes over time. Instead, pressure acts as a mathematical enforcement mechanism—a Lagrange multiplier that instantaneously adjusts to ensure the velocity field satisfies the mass constraint.
If you attempt to solve these equations simultaneously in a naive, coupled fashion, you face massive, poorly conditioned matrix systems. To bypass this computational bottleneck, modern numerical frameworks rely on iterative pressure-velocity decoupling. By breaking the monolithic system into a sequence of segregated steps, researchers can solve for velocity and pressure independently while strictly maintaining physical conservation laws.
The Finite Volume Foundation: Enforcing Conservation
To understand why decoupling works, we must first look at the spatial framework supporting it. The dominant choice for fluid simulations is the Finite Volume Method (FVM).
Unlike finite difference methods that approximate derivatives at discrete points, FVM divides the physical domain into small, contiguous control volumes. As Liu demonstrates, integrating the governing partial differential equations over these control volumes inherently enforces the conservation of physical quantities like mass, momentum, and free energy. The rate of change within a cell is perfectly balanced by the net numerical fluxes passing across its faces.
This strict enforcement is not just a mathematical preference; it is vital for physical consistency. For instance, Song et al. note that maintaining mass conservation across spatial cells is what allows numerical frameworks to remain stable when simulating fluid movement through highly heterogeneous or fractal media. Similarly, Li et al. show that this control-volume approach successfully bridges scales, translating noisy particle-level momentum and energy fluxes into smooth, conservative continuum mechanics.
Iterative Decoupling: SIMPLE, SIMPLEC, and PISO
The most widely used strategy to decouple pressure and velocity within FVM is the segregated iterative algorithm. Instead of solving for velocity ($\mathbf{u}$) and pressure ($p$) simultaneously, these algorithms guess a pressure field, solve for an intermediate velocity, and then apply a correction.
[Guess/Update Pressure p*] ──> [Solve Momentum for Intermediate Velocity u*]
│
▼
[Satisfy Continuity via p' Correction] <── [Solve Pressure-Correction Equation]
The SIMPLE Family
The foundational algorithm in this class is SIMPLE (Semi-Implicit Method for Pressure Linked Equations). In a SIMPLE iteration, the authors of Airflow field simulation model building and application for aerobic compost using CFD outline the core loop:
- An estimated pressure field ($p^*$) is used to solve the momentum equations, yielding an intermediate velocity field ($\mathbf{u}^*$).
- Because $\mathbf{u}^*$ typically violates the divergence-free mass constraint, a pressure-correction equation (derived from the continuity equation) is solved to find a correction factor ($p'$).
- The pressure and velocity fields are updated using $p'$ and under-relaxation factors to ensure stability, and the loop repeats until convergence.
Barbosa Saldaña et al. and Lin et al. leverage this iterative coupling to maintain stability across complex industrial flow regimes and high-temperature combustion chambers. To accelerate this process, Qiu et al. employ SIMPLEC (SIMPLE-Consistent), which modifies the velocity correction step to omit less significant velocity neighbor terms, allowing for faster steady-state convergence without heavy under-relaxation.
The PISO Algorithm
When simulations demand high temporal accuracy—such as capturing transient turbulent eddies or viscoelastic fluid resonance—the PISO (Pressure-Implicit with Splitting of Operators) algorithm takes over.
As utilized by Zhang et al. and Chen et al., PISO extends the SIMPLE philosophy by performing an additional explicit correction step within each time increment. It solves the momentum equation once but performs multiple pressure-correction loops. As Sun et al. highlight, this iterative pressure decoupling drastically reduces the residual velocity error at each time step, making it highly efficient for unsteady, time-dependent flows.
Fractional Steps and the Suppression of Oscillations
Another elegant way to decouple the system is the fractional step projection method. Rather than iterating within a time step to achieve convergence, the projection method splits the time-advancement step into distinct physical mechanisms.
Huang et al. illustrate this process clearly:
- Advection-Diffusion Step: Compute an intermediate velocity field ($\mathbf{u}^*$) by advancing the momentum equations forward in time while ignoring the pressure gradient entirely.
- Projection Step: Because $\mathbf{u}^*$ is not divergence-free, project it onto a divergence-free space. This requires solving a three-dimensional Poisson equation for pressure: $$\nabla^2 p = \frac{\rho}{\Delta t} \nabla \cdot \mathbf{u}^*$$
- Correction Step: Use the gradient of this calculated pressure to correct $\mathbf{u}^*$ into a final, divergence-free velocity field ($\mathbf{u}^{n+1}$).
Suppressing Checkerboard Oscillations
A classic trap in pressure-velocity decoupling occurs when variables are arranged on a standard grid where pressure and velocity are evaluated at the exact same spatial locations (collocated grids). In this setup, a highly oscillatory, non-physical pressure field (e.g., alternating between $+100$ and $-100$ at adjacent nodes) can produce a central-difference pressure gradient of zero. The velocity solver sees no gradient, completely missing these catastrophic "checkerboard" oscillations.
To combat this, Shetty et al. adapt iterative decoupled solutions specifically for collocated arrangements by interpolating face velocities using momentum coefficients, ensuring that adjacent pressure differences are tightly coupled to face fluxes.
From a finite element perspective, Liao et al. demonstrate that decomposing the momentum equations into intermediate and corrected velocity components naturally suppresses these non-physical pressure oscillations. This splitting mechanism bypasses the strict Ladyshenskaya-Babuška-Brezzi (LBB) condition, ensuring numerical stability and spatial accuracy even when modeling highly turbulent flows around irregular geometries.
Alternative Decoupling Paradigms
Beyond standard segregated solvers, researchers have engineered alternative mathematical formulations to break the pressure-velocity bottleneck.
- Pseudo-Compressibility: Instead of forcing the fluid to be perfectly incompressible at every iteration, Tamura et al. introduce an artificial time-derivative term for pressure into the continuity equation. This creates a pseudo-compressible relationship where pressure waves propagate through the domain during the numerical iteration process, coupling the fields dynamically until the simulation converges to a steady, incompressible state.
- Pressure Elimination: Rather than managing the pressure variable iteratively, Colmenares et al. eliminate pressure entirely from the primary system of equations by formulating the fluid problem within a specific functional subspace. The pressure field is then reconstructed through an isolated post-processing step, simplifying the coupled system while preserving the integrity of the velocity field.
- Boundary Decoupling in Particle Methods: The challenge of pressure coupling also extends to particle-based fluid methods like Smoothed Particle Hydrodynamics (SPH). Kolb et al. address this by introducing an alternating iteration scheme that separates boundary-induced and fluid-induced pressure forces. Decoupling these forces prevents the overestimation of repulsive boundary interactions, significantly reducing residual velocity jitter and eliminating the ghost pressure oscillations common near solid surfaces.
Key Foundational Work
Forced Convection Over a Three-Dimensional Horizontal Backward Facing Step (2005) — contributed 1 concept elements to this direction.
20322 Numerical Analysis of Flow around Objects with Human Body (2006) — contributed 1 concept elements to this direction.