Computational Fluid Dynamics Simulation

CFD simulates the dynamics of fluids such as liquids and gases, as well as suspended particles and mixtures of these. The behaviour of fluids is described by models such as the Navier-Stokes equations and Lattice Boltzmann method, with variables such as pressure, density, bulk viscosity, dynamic viscosity, velocity and acceleration. For many scenarios, the equations of fluid dynamics often have no known analytic solution, especially for complex behaviour such as turbulence. Therefore, CFD simulation is needed to solve them.

Resolving the fluid flow equations is not trivial – there is no known general analytical solution for the Navier-Stokes equations, meaning that numeric techniques must be used. Boundary Element, Finite Element or Finite Difference methods were among the most common CFD simulation methods initially, but Finite Volume Methods have gained prominence in the past 20 years and are now standard. More recently, increased hardware performance mean that Lattice Boltzmann Methods have also become viable.

As there is no best CFD simulation method for all applications, engineers must choose the right tool for each industrial workflow.

Coupled Eulerian-LagrangianCoupled Eulerian-Lagrangian and smoothed-particle hydrodynamics (SPH) is ideal for highly coupled fluid-structure interaction (FSI) problems such as non-compressible hydraulics. These methods are implemented in SIMULIA Abaqus/Explicit.

The Finite Volume Solver is better suited for steady or moderately transient flows, such as those found in pipe flows, heat exchangers, pumps and HVAC applications. This method is used in SIMULIA Fluid Dynamics Engineer (FMK).

As a transient technique, the Lattice Boltzmann Method (LBM) is more applicable to highly transient flows such as aerodynamics and aeroacoustics. It can handle highly demanding models both in terms of geometric complexity and detail. LBM technology can also be effective for multiphase applications handling complex models that can even include arbitrary moving parts. Dassault Systèmes SIMULIA offers two LBM products. SIMULIA PowerFLOW is ideal for the aerodynamic, acoustic and soiling scenarios common in aerospace and automotive industries. SIMULIA XFlow meanwhile is typically used for complex moving, multiphase problems such as lubrication, sloshing and some Life Science applications.

Both methods are powerful CFD solvers, but they use different approaches and have different benefits. Navier-Stokes methods treat the fluid as a continuum, while Lattice Boltzmann method treats it as discrete particles.

To solve the Navier-Stokes equations computationally, the physical space to be simulated is divided into many small subdomains called control volumes or cells. The equations are discretized across the cells, and the resulting set of algebraic equations is solved iteratively to obtain the pressure, velocity, temperature (and other physical quantities) in each cell for steady or unsteady flows. Additional discretized transport equations can be solved in the same way to represent other physical phenomena like turbulence and chemical species.

The Lattice Boltzmann method of CFD simulation tracks the microscopic motion of fluid particles through discrete space and time, to simulate the flow of gases and liquids. The fluid space is automatically discretized into cubic voxels, and boundaries into surfels, eliminating the need for conventional surface and volume grid generation. The Very Large Eddy Scale (VLES) turbulence modeling approach ensures that anisotropic fluid structures are captured with high fidelity, which is critical for aerodynamics and aeroacoustics workflows.

Yes, several high-performance computing (HPC) techniques can accelerate CFD simulation. GPU acceleration speeds up the simulation process and allows larger and more complex models to be simulated on a single workstation. One GPU can have the power of over 1000 CPU cores – GPUs reduce hardware cost and enable desktop supercomputing. Multi-GPU acceleration provides even more speed-up, solving extremely large or complex scenarios that cannot be simulated otherwise. GPU benefits all CFD codes, but is especially effective for Lattice Boltzmann Method (LBM) simulation.

Cluster computing can simulate even larger scenarios. Cloud offers greater control of batch simulation processes. Both on-premises and online cloud are available. On-premises cloud uses hardware operated by the user’s company, while online cloud sends the data to secure servers for processing. Online cloud is a good option for users with irregular or periodic simulation needs, avoiding the need to invest in hardware that would only be used occasionally.