What is a CFD analysis?

OptiFluides, as an engineering firm specializing in numerical simulation, primarily provides Computational Fluid Dynamics (CFD) analyses to assist our industrial clients with their fluid mechanics issues.

Although now widely used, the term “CFD study” sometimes has different definitions depending on the sector of activity. Therefore, we will take a moment here to provide a clear and comprehensive definition of what a CFD study is and can be.

Types of analyses in fluid mechanics

But before we rush into CFD studies, let’s take a step back: what are the different ways to address a fluid mechanics problem?

There are several possible approaches: analytical studies, experimental studies, and numerical studies (or CFD calculations). These approaches differ in terms of their degree of complexity, the time required to complete them, and their cost. Although they are sometimes presented as antagonistic, these approaches are very often complementary and can sometimes be combined to address a complex issue. At OptiFluides, we offer these different approaches to help you meet your industrial challenges.

Definition and challenges of a CFD analysis

A CFD (Computational Fluid Dynamics) study is a technical engineering study that involves using computing resources to simulate the flow of a fluid (liquid or gas) in a product or system in order to predict its behavior.

In a CFD study, computational tools are used to numerically solve the Navier-Stokes equations governing fluid flows, thereby predicting the relevant flow parameters, such as velocity, pressure, temperature, heat flux, shear stress, etc.

Due to the non-linear behavior of fluid flows, a CFD study is an important asset for better understanding how a product or system works and interacts with its environment, optimizing a design, sizing a system, anticipating problems, and thus reducing experimental testing, development time, and ultimately costs.

What are the differences between analytical, experimental, and CFD analyses?

In an analytical study, theoretical equations derived from physics are used to propose an exact or simplified mathematical solution. Examples include calculating pressure loss in a straight pipe or in a simple geometric singularity, calculating leakage flow through a reference orifice, calculating the flow rate on a spillway crest, and even semi-analytical resolution of partial differential equations on heat balances, for example. To propose these solutions, it is generally necessary to make simplifications and strong assumptions that allow to refer to a reference problem whose theoretical solution is known. This is a good first approach, but it requires variable implementation time, careful verification of the applicability limits of the solutions found, and significant simplification of the problems. This is not always possible for industrial problems. We must then move on to the next level of complexity.

There are two possible approaches: experimental study or numerical simulation. In an experimental study, we either equip the subject with “in-situ” sensors, i.e., on the product or system that is actually installed and in operation, or we build a prototype, model, or test bench in order to measure the physical quantities directly. This has advantages: the results are very realistic and all complex phenomena are captured, but there are also many limitations: high costs and long lead times, scaling effects, external disturbances, and point measurements.

Numerical simulation makes it possible to model the product or system under study at actual scale, to test multiple configurations in a short time and within a generally much more reasonable budget, in an “ideal” case without external disturbances and allowing knowledge of all the physical quantities of the flow at any point in time and space. However, just like the experimental approach, the numerical approach has its limitations: numerical error, knowledge of boundary conditions, validity of mathematical models, adequacy of numerical parameters, etc. We can therefore see that the experimental and numerical approaches are not opposed but highly complementary, particularly when it comes to complex issues, whether for recalibrating numerical models or better understanding the origin of a phenomenon identified experimentally.

What are the different types of CFD analyses?

Once the decision has been made to perform computational fluid dynamics (CFD) calculations, it is important to understand that there is a whole universe of models available to provide answers with varying levels of complexity, precision, and cost, depending largely on the problem being studied. The first step in making an initial selection is to refer to the classification of flows:

Steady-state or transient: a flow is said to be steady-state when all physical quantities (velocity, pressure, temperature) are constant over time. For a transient flow, these properties may vary.
One-dimensional, planar, axisymmetric, three-dimensional: depending on the number of directions in which the physical quantities are non-zero, in Cartesian or cylindrical coordinates.
iscous or non-viscous: depending on the importance of viscosity in the diffusion and dissipation of momentum (negligible or zero viscosity).
Incompressible or compressible: depending on the variations in fluid density with pressure for a given flow.
Laminar or turbulent: depending on the relative importance of inertial effects compared to viscous effects.
ingle-phase or multi-phase: depending on the presence of one or more physical states (liquid/gas or even solid with possible phase changes).
Single species or multi-species: depending on the number of chemical species in the same physical state.
Internal or external: depending on whether or not there are walls “containing” the flow.
Isothermal or with heat transfer: depending on whether or not there are thermal gradients within the flow, with all the associated modes of heat transfer.

The first step in a CFD study is therefore to determine the type of flow involved and then select the appropriate models. It should be noted that there may be different modeling approaches for a given type of flow, as is the case with turbulence, for example. Here again, it is a matter of choosing the most appropriate approach to address the issue at hand.

What are the differences between an aerodynamic analysis and a hydraulic analysis?

Another approach to classifying the fluid mechanics problems we solve is to distinguish between those involving mainly gases, liquids, or both. Although somewhat simplistic, this classification distinguishes between:

Aerodynamic analyses, which mainly concern the study of air flows and, by extension, any gas. The first thing that comes to mind here are studies relating to the ventilation of premises (i.e., mainly stationary, three-dimensional, viscous, slightly compressible or even incompressible, turbulent, single-phase, potentially multi-species, internal, and anisothermal, if you’ve been following closely), but we can also include the atmospheric dispersion of polluting gases (transient, external), the modeling of a turbocharger (single species but with a rotating reference frame, compressible, internal), or the simulation of a heat exchanger using metal foams.
Hydraulic analyses, which refers more specifically to the flow of water and, by extension, liquids. A key distinction must be made here between pressurized flow (no air/water or gas/liquid interface) and free surface flow (where an air/water interface must be modeled). Typical examples include numerical modeling calculations for hydraulic structures, whether for dam spillways or lock chamber filling pipes. This category also includes calculations of pressure loss and flow distribution in hydraulic distributors, the mixing of chemical species in a reactor, and oil cooling of nuclear power plant transformers.
Finally, there is a last category of studies known as multiphase studies, in which we focus specifically on physical exchanges between phases. This includes, for example, spray calculations, nasal cleansing, numerical simulation of fluidized bed reactors, and flash vaporization calculations.

Throughout its 15 years of existence, OptiFluides has carried out projects on the different types of flow presented here and has acquired recognized experience in complex modeling.

Do you have a problem related to fluid flow? Take the time to contact us and we will work with you to determine what support we can provide!

CFD analyses

What is the average duration of a CFD study?

The average duration (and budget) of a CFD study depends on the complexity of the model requested, the number of scenarios to be processed, and the elements necessary for the study that are already available (CAD, knowledge of input conditions, etc.).

Some relatively simple studies can be completed within a week, while more complex projects require work programs spanning several years.

A reasonable timeframe for most problems is one to two months, with the possibility of mobilizing more resources for urgent cases.

How accurate can the results be?

The accuracy of CFD studies depends on the complexity of the problem, the quality of the model, and the level of verification and validation performed.

Overall, for simple problems such as calculating pressure loss in pipes or flow distribution, the error can be less than 5%. For more complex problems involving highly turbulent, multiphase flows with rapid and complex phenomena (cavitation, vortex-induced vibrations, etc.), the error will depend on the degree of verification performed—according to the customer’s requirements and budget—the quality of the input data, and the reference results available to quantify it.

Errors can arise from the quality of the mesh, but also from the choice of physical models (turbulence, compressibility, consideration of thermal effects, etc.) and boundary conditions.

To guard against this, the V&V (Verification and Validation) standard developed by ASME, among others, provides a framework for:

Verification, i.e., control of numerical error, in particular through the stages of mesh convergence and numerical convergence of the calculation.
Validation, i.e., controlling modeling error by comparing it to experimental data from manipulations or the literature.

In industry, an error of 10% is generally considered acceptable, although in absolute terms, this value does not really make sense. Indeed, for certain complex flows, a response with a 50% error will be considered excellent (because it is a very complex problem with little initial knowledge), while it is possible to achieve levels of 2-3% on other problems.

Do I need a CFD study for my project?

That will depend on the problem. Generally speaking, if one or more of the following points apply, a CFD study may be relevant:

The problem is physically complex: analytical study impossible
Existing correlations are unsuitable: no empirical or test-based laws available
Experimental testing is too costly, impossible, or insufficient: data from the literature can be used if the models are complex and require validation.
Detailed local information is needed: to identify recirculation zones, hot spots, detailed velocity fields, etc.

As mentioned above, systematically proposing a CFD study is not the right approach: whenever possible, if an analytical and sufficiently predictive answer exists, it will be prioritized. Similarly, if high accuracy is required and the physical models are complex, a CFD study without experimental validation will prove to be of little use. Finally, in some cases, a semi-analytical approach (an analytical model enriched by simulation results) can also be a very good compromise.

The best way to answer this question is to discuss your problem together.

Which software programs are used?

The software used depends on the problem to be solved. There are a wide variety of numerical methods (Finite Volume Method – FVM, Smoothed Particle Hydrodynamics – SPH, Lattice Boltzmann Method – LBM, Finite Difference Method – FDM, spectral methods, mixed formulations, etc.), resulting in just as many possible software programs. We have conducted a non-exhaustive review of some CFD software.

For the same problem, in most cases, different software and different methods will lead to a similar solution. The main point is therefore to have a good command of the software, and particularly the limitations of the models, by implementing good V&V (Verification & Validation) practices.

Some more specific problems can only be addressed with certain numerical methods and software. Examples include fluid/structure coupling, the search for particularly high levels of accuracy, and granular flows.

Learn more about computational fluid dynamics.