Bigger, Better, Motor Sport CFD

Alex Read, CD-adapco

In motor sport bigger, is better. Nowhere is this more true than in the world of motor sport CFD, where teams are regularly pushing the envelope, running simulations involving hundreds of millions of cells. A mainstay of F1, and its use rapidly increasing in NASCAR (for aerodynamics, underhood thermal management and engine simulations), this article details how CD-adapco’s new technology solution, STAR-CCM+, is enabling teams in F1 and NASCAR to create, set-up, run and post-process bigger, and of course better, flow and thermal simulations.

Why so big?

Race teams use Computational Fluid Dynamics as an aid and extension to physical testing. It is used to optimise expensive wind tunnel time by identifying the best designs to work with and/or provide information that is difficult to obtain through testing. For NASCAR underhood thermal management, this can be detailed visualisation of the flow and thermal field in the engine compartment. For external aerodynamics, it may be the car’s front and rear down force balance when drafting. 

Race team’s other requirements, common to all simulation engineers, are accurate results and rapid and robust model turn around. This presents a particular challenge in areas like external aerodynamics where components have a strong interdependence. For example, the performance of the rear wing of an F1 car will vary depending on the set-up for the components in front of it and vice-versa.  F1 teams opt for running detailed models of the full car to ensure accurate aerodynamic resolution for all parts of the car.  Similarly in NASCAR the interaction between cars, aerodynamic and even physical contact, is a key part of racing. As a lot of the time the cars are bumper to bumper and door handle to door handle, understanding the effect on front and rear down force and airflow to the engine compartment when in close proximity to other cars can be the difference between success and failure.

Historically, this has presented problems, is my computer big enough store and solve at each of my hundreds of millions of cells? If it is, is my CFD tool sufficiently adept at utilising this enormous computing resource, with hundreds of processors operating in parallel, to enable me to create, set-up, and run and post-process my case within a reasonable timeframe?

CD-adapco has a proud history of solving these problems for race teams: supplying CFD tools to, among others, the double World Championship winning ING-Renault F1 team since its inception in 2001. Their latest offering, STAR-CCM+, has been specifically designed for motor sport CFD, with regular evaluation and specification being carried out by top motor sport teams during its development.

“Around five years ago we starting to develop STAR-CCM+ from a blank sheet of paper,” explains Dr Richard Johns, CD-adapco’s Director for the Automotive industry, “which allowed us to use everything we’d learnt in the previous twenty years of developing and using CFD, as well as the latest computing technology. In addition, at CD-adapco, we see our users as partners and not just clients. As well as using our own know-how, our motor sport partners were integral in defining STAR-CCM+’s specification and reviewing its progress.” The result is a code that’s revolutionizing motor sport CFD, allowing teams to run ever more detailed models, with ever more computational cells, while shortening case turnaround times.

The Technology

Much of the focus on CFD codes has been the time it takes for them to solve the Navier-Stokes equations for a large number of computational cells. Of equal importance is the time required to create, set-up and post-process the case, which can take days or even weeks. A key technology in STAR-CCM+ is the process by which it does this for very large cases. First, all the steps of the simulation process are integrated into one tool: from CAD geometry to post-processing.  This saves a considerable amount of time as there is no requirement to export and import large data sets and there is no requirement to re-specify parameters in different tools or when running iterative design studies. Second, it utilises the latest software technology, a client-server architecture. The server runs on a parallel cluster, distributing the work of processing hundreds of millions of cells, while the light java client only passes the information it needs. What this means is that cases can be set-up, run and post-processed using the light client while the server makes full use of parallel hardware, significantly cutting model turnaround time.

 

Pushing the limits

So far, we’ve got to a simulation of forty cars with a mesh count of one billion polyhedral cells, the equivalent of several billion tetrahedral cells! The purpose of the simulation was twofold, to simulate the complex interaction between multiple NASCARS in close proximity and to see how far we could push STAR-CCM+ on existing hardware. 

The starting point for the calculation was creating a mesh around a single car, providing the baseline drag, lift and yaw force values. A second car was then introduced and analyses were performed with the cars directly in-line, and with an offset as if at beginning an overtaking maneuver.  The drivers’ goal when in the drafting formation is to reduce the drag on both cars, making them collectively faster. If the second car is in the correct position, it has the effect of increasing the pressure at the rear of the lead car, reducing its overall drag. However, the effects are highly dependant on the car positions.  At times, the drag on the second car is reduced as the first car deflects the air over it. In other configurations the drag force on the rear car is actually greater than that on the lead car as it sits in the dirty, highly turbulent wake.  Both cars handling are also affected as the front and rear lift on each varies with the car positions and strong yaw forces occur on the rear car when in the offset position. These simulations provided the aerodynamicists with detailed visualisation of the complex flow patterns around the cars. 

The model was then extended to evaluate what happens in highly complex race conditions, with many cars unevenly spaced and positioned. This was built using a “building block” approach. One car and its close proximity were meshed.  This mesh was then copy and pasted with an offset to produced two cars side-by-side, or nose to tail. To vary the distance between the vehicles a spacer mesh section was used. Using this technique, a model of forty, unevenly distributed cars with a total mesh count of one billion cells was generated. Although this is impressive by today’s standards, according to Dr Richard Johns, multi-billion cell calculations will we common-place in the not too distant future: “As the hardware vendors continue to produce even bigger and faster machines, so the model sizes race teams want to run increases.  It’s our goal to make sure STAR-CCM+ can efficiently handle these enormous calculations. So far we’ve made it to one billion; we don’t see any reason why in the future we can’t go much larger than that.”

More than just flow

Of course, aerodynamics is only one application area for motor sport CFD. NASCAR teams are increasingly using CFD for underhood thermal management simulations to gain detailed insight into the flow and thermal fields. Here, the CFD tool must be able to automatically mesh highly complex geometries and deal with the additional physics required to efficiently model fans and heat exchangers and convective and radiative heat transfer. CD-adapco’s pedigree is strong, having been running underhood simulations for over a decade. “Our codes are the tried and trusted solution for powertrain teams around the world, and when developing STAR-CCM+ it was one of our target applications,” says Johns. “Its combination of the latest generation of our surface wrapper, and dedicated underhood models mean it’s much more than just a flow code.”

At the heart of STAR-CCM+ is an automated process that links a powerful surface wrapper to CD-adapco’s unique meshing technology. The surface wrapper significantly reduces the number of man-hours spent on surface cleanup and, for problems that involve large assemblies of complex geometry parts, reduces the entire meshing process to hours instead of days.  The surface wrapper works by “shrink-wrapping” a high quality triangulated surface mesh onto any geometrical model, closing holes in the geometry and joining disconnected and overlapping surfaces, providing a single manifold surface that can be used to automatically generate a computational mesh without user intervention. STAR-CCM+ also has specialist models for efficient representation of fans and heat exchangers.

Staying ahead

In motor sport CFD from F1 to NASCAR bigger, is better. As teams continue to push the envelope of CFD so the tools they use need to adapt to their requirements.  “By adopting state-of-the-art technology, in combination with years of know-how from us and our partners, STAR-CCM+ is helping race teams run bigger models, faster and ultimately to stay at the front of the grid” says Richard Johns.

[END 1496 words, 7731 characters (no spaces), 9275 (with spaces)]

Images:
Single Car Simulations

Streamlines colored by velocity magnitude

Pressure coefficient on vehicle surface with iso-lines of velocity magnitude

Drafting – two car simulations

Streamlines colored by velocity magnitude

Very large simulations!