Computational Fluid Dynamics and CFD engineer : What is it?

Through my professional activity, I offer companies expertise in Computational Fluid Dynamics, or CFD. “And what is it?” I have already been told. Here is, for those who are not familiar with it, what my job consists of.

Fluid mechanics... With computers

Definition and origin

CFD (Computational Fluid Dynamics) consists of studying the physics of flows by numerically solving the equations governing the behaviour of fluids (using a computer or computing cluster). In this way, it is possible to calculate the temporal and/or spatial evolution of any fluid system, with a degree of approximation to reality that is directly dependent on the modelling assumptions chosen.

Historically, this speciality of numerical physics was primarily developed to study aerodynamics in the automotive and aerospace industries. Then it gradually spread to other sectors. Today, numerical simulation in fluid mechanics has become indispensable in a large number of fields, and concerns almost all sectors of activity. Today, the development of computing resources means that multi-physics simulations can also be envisaged, combining fluid mechanics with other processes (chemistry, structural mechanics, etc.).

Some CFD applications

Any process involving one or more fluids can be the subject of CFD modelling. As a result, and as mentioned above, most sectors of activity have to deal with problems of this type. Examples include, but are not limited to :

  • Building: ventilation, heating/air conditioning, smoke extraction, condensation.
  • Town planning: wind flow in urban areas, resistance of structures to gusts, air quality studies.
  • Energy: combustion in engines, cooling in nuclear power stations, wind farm efficiency.
  • Transport: vehicle profiling and aerodynamics, drag.
  • Hydraulics: free-surface flows, turbine operation, cavitation.
  • Health: blood flows, virus propagation in confined environments.
  • Etc.

To give you a more concrete idea, here are some demonstration videos of CFD simulations carried out at NEMOSFLOW.

Some CFD simulations. Watch all my videos here.

Overview of the business approach

Generally speaking, a CFD problem is often dealt with using an approach consisting of three main stages, which are outlined below.

Modelling the problem

It is during this phase that the digital representation of the actual process is constructed. It is therefore necessary to:

  • Construct the geometry of the problem and the computational domain.
  • Discretize this computational domain so that the desired equations can be solved within it, i.e. construct a mesh. To be completely rigorous, it should be specified that this step is not necessary with certain approaches. However, most of the major commercial and open-source codes on the market are based on the so-called finite volume method, which does require the construction of a mesh.
  • Define the equations to be solved according to the physics of the problem: presence of thermal phenomena, turbulence, phase changes, chemical reactions, etc….
  • Set the boundary conditions of the calculation domain, and the initial conditions of the problem being treated.
  • Fill in the physical properties of the materials present.

In particular, the CFD engineer’s expertise must enable him to find a good compromise between the degree of precision / type of results desired, and the assumptions that it is possible to adopt in the modelling. In particular, ensuring that the right terms are included in the equations, and that the boundary conditions/initial conditions correspond to the actual process, are essential points in obtaining a relevant simulation.

cfd simulation of school - mesh
Geometric model and mesh example

Numerical simulation of the problem

This second stage involves the use of CFD software which, at any point (or volume) in the mesh and for any time step, will solve the evolution of the model constructed previously. Depending on the complexity of the problem, the type of software used and the computing power available, this resolution may take a few seconds or several months.

The CFD engineer must use his numerical skills to ensure that the simulation runs smoothly. To do this, he or she must in particular choose appropriately:

  • The spatial and temporal numerical schemes.
  • The algorithms for solving the linear systems created.
  • The numerical parameters required for good convergence of the calculation towards a physical solution, without leading to an excessively long or unnecessarily fine simulation.

Results post-processing and interpretation

Once this final post-processing stage has been reached, it is first necessary to ensure that the results are consistent. For this, the existence of a reference case where it is possible to compare with experimental results often proves invaluable. Failing that, the evolution of the quantities of interest (taking into account the expected physical behaviour, conservation laws, etc.) should enable the model constructed to be validated (or corrections to be made). Only then can the results be used for the desired purpose (optimisation, understanding, etc.).

cfd simulation of school - Streamline
Results example : streamlines colored by velocity magnitude

Required tools

To carry out and analyse a simulation, the CFD engineer needs the following:

  • A computing cluster (or at least a powerful workstation for less demanding simulations).
  • CAD (Computer Aided Design) software.
  • Meshing software.
  • CFD software.
  • Visualisation software.

There are many solutions on the market, which can be divided into two main categories:

  • Proprietary solutions (Fluent, Star ccm+, etc.).
  • Open source solutions (OpenFOAM, Code_Saturne, Paraview, etc.).

At NEMOSFLOW, I mainly use open source software. For more details, take a look at this page.

The benefits of CFD for industry

Used wisely, CFD proves to be a major asset:

  • To simulate systems before building them, enabling them to be optimised in a pre-project phase.
  • In order to optimise a process via a multitude of parametric simulations, all at a very modest cost and time compared with a test campaign.
  • Access to all the physical quantities, at any point of the mesh constructed and at any time simulated, gives access to a much larger quantity of information than experimentally, which promotes understanding of the physical phenomena involved.
  • Be careful, however, not to lose sight of the essential point: the relevance of the simulation to the real process. To guarantee reliable results, all modelling must be confronted with validations: verification of correct behaviour on a simple academic case, or comparison with experimental tests, for example.

CFD is therefore both a research tool and an industrial tool, which has its rightful place in the world of engineering. Today, CFD is used in a wide range of sectors, from transport and meteorology to chemistry, medicine and construction. The optimisation it brings can also have ecological benefits. A few examples of applications can be found here.

Role of the CFD engineer

Ultimately, working in the world of CFD requires a number of skills:

  • Sound knowledge of fluid mechanics, physics, applied mathematics, programming and scientific computing.
  • Mastery of the tools on which they can base their expertise: CFD software, meshing software, scientific visualisation software, etc. But also scientific programming languages.

The CFD engineer has, to some extent, the role of interface between the physical world and the digital world. This often involves a great deal of interaction with other professions (civil engineering, processes, IT, etc.), and with other areas of physics (electromagnetism, mechanics, chemistry, etc.). Their expertise must enable them to correctly represent the desired physics, using the tools and computing power at their disposal. As the saying goes, they have to make the right calculations, not just calculations!

This is obviously the objective that I am pursuing through my work at NEMOSFLOW. For more details, take a look at the CFD services I offer.