A major innovation for the digital simulation of industrial fluids

When it comes to numerical fluid simulation, until now it has been necessary to choose between accuracy and speed of calculation. Why is this? Let's take the example of turbulence around an Airbus A380 in flight: an infinite number of vortices form and interact, some of which are no larger than ten microns. To simulate them all accurately would require 10,000 times more memory than the world's most powerful computers... And even if such supercomputers existed, they would take years to solve the equations. This scenario is unthinkable today.

Faced with this industrial conundrum of precision versus speed, the Cagire project team (a joint venture between Inria, the University of Pau and Pays de l'Adour, and the CNRS) has decided to tackle it through the CELTIC project. “Our research focuses on the numerical simulation of turbulent flows, with the aim of accurately representing reality in industrial and environmental flows,” explains Rémi Manceau, research director at the CNRS and head of the team.    

There are currently two main techniques used to model turbulent flows. On the one hand, the Reynolds-Averaged Navier-Stokes (RANS) method does not aim to simulate eddies, but rather to evaluate their average flow. Its advantage is speed, with results obtained in a matter of hours. Although it is currently used in industry, it has one weakness: its lack of accuracy in areas where the flow is turbulent or complex.

On the other hand, the LES (Large Eddy Simulation) method accurately describes large eddies and is also capable of modeling smaller ones. The catch is that it is extremely costly in terms of time and energy, with calculations taking weeks to complete.

Manufacturers are therefore faced with a dilemma: choose fast but approximate calculations, or very accurate calculations that are impossible to use in everyday industrial design. This has led to the idea, explored in recent years, of reserving LES only for areas where it is truly indispensable, and using RANS everywhere else. This principle had already been considered in the early 2000s at Boeing. 

“The concept known as ‘Embedded LES’ was proposed a few years ago,” recalls Rémi Manceau. The calculation domain is divided into zones, each of which is processed using the most appropriate method. But at the interface between a RANS zone and an LES zone, the information is not always compatible. One works with average flow, the other with turbulent flow. This concept therefore never really caught on in the industry."

This is where the Cagire project team and its CELTIC partners come in, proposing to eliminate this interface. “Our approach, called CELES (for Continuous Embedded LES), gradually shifts the calculation model from RANS to LES, without any breaks, using a single system of equations that slides naturally from one zone to another depending on the local flow,” explains the researcher. 

The key? Introducing volume forces into the equations to balance turbulent energy, as these forces act at every point in the fluid. The CELES method calibrates them according to the principle of energy conservation, whereby energy can be transformed from one form to another in an isolated system. In this way, all interface configurations can be managed: at the inlet, outlet, or sides of the LES zone surrounded by RANS.

Rémi Manceau notes: "What the project revealed was that all the hybrid RANS-LES techniques developed over the past 25 years suffered from the same flaw. They did not respect the principle of energy conservation at the transition between the two zones. However, any physical model must apply this principle. This is why RANS/LES hybrid methods had encountered problems until then, distorting the results. They lost energy in the transition between the two zones."  

Aeronautics served as the testing ground, but the scope of application for the CELES approach is much broader. In fact, it applies to all fields dealing with complex flows: automotive, naval, missiles, etc. “Energy too, for simulating turbulence around wind turbines, in nuclear reactors, or hydraulic circuits, a subject on which we are collaborating with EDF,” says Rémi Manceau. The energy company is also interested in modeling atmospheric flows to predict the dispersion of pollutants in the event of an accident in order to minimize its impact. This is a typical situation where LES is only necessary around the power plant, as the rest of the atmosphere can be treated using RANS."

Another promising sector is healthcare, particularly for simulating blood flow in the heart. “The University of Montpellier is working on this. We could imagine a digital twin of the heart that would assist surgeons before operations on the geometry of patients' hearts. The aim would then be to improve turbulent cardiac flow.”

In other words, there is no shortage of prospects for this agile approach... By designing a method that focuses precision solely where it is truly useful, the CELTIC project marks a decisive step toward more powerful simulations that are more energy efficient and better suited to industry.

“The CELES method is ready to be incorporated into energy sector software. However, in the aeronautics sector, it is still a little ahead of its time for companies, many of which are not yet ready to integrate it into their industrial tools,” says Rémi Manceau. Its integration would require a dedicated transfer of technological innovation within a real industrial framework." To be continued...