Analysing and understanding the first simulation results to pinpoint the exact flow features we want to validate at the track
The main focus when it came to visualising and interpreting the results from the first set of simulations on EV23 was to identify key areas to validate at track. The intention of such validation is to increase the accuracy of the CFD setup in areas that will affect total forces and moments on the car, as well as guide decisions on optimising surface design. There's not much point spending days making small adjustments to gain an additional 5% downforce when the target area is located in a particularly sensitive flow field that may differ significantly from reality depending on the settings used, or the scenario being simulated.
With no aerodynamic devices and our radiator mounted at the rear of the car, the flow field around EV23 is relatively simple with most of the notable features being bluff wake structures behind the wheels, driver, and lower rear chassis/driveline area. This applies to both the straight line case and the representative corner case, with the main difference in the cornering case being a power-up of many of the vortical structures, and a skew of the overall vehicle wake as it interacts with itself downstream. Our Kiel probes and differential pressure sensors naturally capture total pressure, so this was the main metric of interest to display and sample values from for off-body flow. The coefficient form of total pressure rescales with respect to local freestream velocity. This enables direct comparisons between different simulated vehicle speeds, but is also more appropriate to use for visualisations in cornering cases, where the freestream velocity is a function of the local radius. Figure 1 shows an animated plane of the total pressure coefficient moving front to rear for both the straight and cornering cases. No numerical data is gained from such animations, but they are an intuitive way to visualise and understand some of the larger and/or dominant flow structures around the vehicle and how the change under different key driving scenarios.
Figure 1: Total pressure coefficient sweeps past the car for the (top) straight and (bottom) skidpan driving scenarios, with values cropped to just below free-stream to provide a more intuitive and less obtrusive "cloud" shape
Before going into detail, I'll point out some interesting differences between the two scenarios. It's very important to note that these simulations are completely un-validated at this point.
The upper wake of the outboard front wheel is much larger in the cornering case, and persists downstream as an almost completely separate structure
Tyre squirt is much more powerful in the cornering case on the sides closest to the centre of the corner, but almost completely suppressed on the sides away from the corner, from a combination of the wheel angle, and the cross-flow induced by the vehicle's slip angle
The mid wake of the inboard front wheel gets sucked in behind the tyre much more aggressively in the cornering case, powering up the rotating upper wake structure
The opposite happens on the outboard front wheel, resulting in almost no rotation of the upper wake structure
The wake from the more exposed upper sections of the car (driver, headrest, roll hoop) are more affected by the cross flow, resulting in them following the radius of the corner while the tyre and bodywork wake is deflected relatively further outboard, resulting in a skew effect that appears to be enhanced by the aforementioned strong rotating upper wake structure on the inboard side
The airflow through the front wheels (inboard to outboard, as a result of the negative toe) pushes the mid wake off the tyre sidewall on the outboard side, however the effect alone is not strong enough to prevent the suction behind the tyre from pulling it back in
I'm slightly dubious of how large the squirt and lower wake structures are for the front wheels, particularly in the straight airflow scenario
For the first validation tests the front wheel wake is of more interest than the rear wheels or driver/headrest wake, as it lies in the target area for aerodynamic surface development and so will have the most influence on simulated performance. Some critical cross sections are shown below in Figure 2 for the straight airflow scenario.
Figure 2: Cross section planes of total pressure for the front wheel wake in the straight airflow scenario, where Z=0 is the midpoint of the wheelbase
The first three images show the progression from the centre of the front axle to just aft of the rearmost surface of the front tyre. These aren't necessarily important for selection and justification of a particular plane to sample for validation, but show the formation of the wake to provide a better understanding of why the structures look like they do. I'll get to the last two images in a minute. The intention for the straight airflow scenario is to have two validation planes to set up rakes for, so that we can validate not only magnitudes and instantaneous locations, but also how the wake shifts and interacts with itself downstream. We also planned for one rake plane for the representative cornering case, time permitting. The criteria for selecting the two straight airflow plane locations were as follows:
Must be a sufficient distance apart in Z such that any rotational or translational flow structures are able to cause a significant change in wake "cloud" shape between the two planes
Forwards-most plane must be as close to the front tyre as possible (to allow sampling of low total pressure magnitudes), but still rearward enough that distinct flow structures have formed
Must be in locations that allow easy, stable mounting for a Kiel probe rake configuration
The last two images in Figure two show the best two candidates for these criteria, and are shown again in isolation with the context of the vehicle in Figure 3.
Figure 3: The two cross section planes selected for validation of the front wheel wake in the straight airflow scenario
Figure 3 shows more clearly the location of these planes on the car, which fall conveniently at the middle and rear chassis nodes of the side impact structure, which will allow easy mounting and constraint and all axes without needed any permanent adjustments to the chassis. Rake design and mounting will be covered in the next post.
Next we decided on what points within the wake we wanted to sample (i.e., choosing the location of probes on a rake). The construction method for the rake was still unknown at this point, but was likely to be made of either sheet metal, or plastic 3D printed. It might not come as a surprise to know we had neither access to nor the budget for metal 3D printing or sintering for something as large as an aero rake, particularly if we go the optimal route of making a unique rake design for each validation test. While plastic printing would allow complete freedom of design, we limited ourselves to positioning the majority of probes in a grid layout in case we were forced to go with a sheet metal design if the plastic were not strong enough, or too flexible. Looking back, this was a strange decision as I can't really imagine a scenario where anything less than 2mm sheet metal (by which point weight would become an issue of its own) would be more rigid than an appropriately reinforced PLA structure, but that's the choice we made at the time.
We have twenty sensors to use, and three of these are dedicated to the pitot-static and yaw probes, leaving seventeen for the rakes. We wanted to have some probes sampling inside the wake "cloud" to get total pressure magnitude readings, while the rest were dedicate to tracking the position of the perimeter of the wake (of course these would also sample magnitudes, but their location inside or outside of the wake was less certain). By analysing how the wake develops and changes as it moves downstream (relative to the car), we got an idea of which parts of the wake were likely to be more sensitive than others, and positioned a probe on the inside and outside of the perimeter at each of these locations. This would allow assessment of if this part of the perimeter lies further inboard/outboard/higher/lower than the current simulations suggest. The rest of the probes were used to fill in the intermediate areas in an approximate grid shape. Ground clearance under expected bump conditions during testing limited how low probes could be places, but the lower wake structure did not contain any particular details of interest, so this wasn't considered a limitation. Figure 4 shows the chosen probe locations for each of the two planes.
Figure 4: Rake probe positions chosen for the two total pressure cross sections
The same was done for the cornering case (not shown), with the outboard side selected for validation. This choice was mostly to avoid potential damage to the rake and Kiel probes in the case of a collision with the cones for the constant radius corner required for this test, which are positioned on the inboard side of the corner. Expected available track time limited the ability to test two rake setups for the scenario. I'll point out that the rakes you see in F1 don't necessarily change for different parts of the track. My guess is that there are two reasons that a single rake will do the job. Firstly, the corner radii are much larger in F1, and their wheelbase is much longer, meaning that both flow curvature and yaw angle are likely less extreme than in FSAE. Secondly, there are typically a lot more probes on an F1 rake, meaning the density is much higher. While we are targeting specific parts of the wake cloud, the sheer number of probes on F1 rakes likely means that a large majority of he wake is covered in good resolution, so even if the wake shape shifts at different points of the lap, it's captured accurately anyway.
By this point we had made the decision to use a 3D printed plastic construction for the rakes, which opened up the potential for 3D surface design. While the Kiel probes we designed exhibited angle insensitivity, it was still preferable to orient the probes tangentially to the expected local velocity vector. This is partly to give the probes the best possible reading (despite wind tunnel tests suggesting this is unnecessary, and with only the lower right probe point at Z=0.35 expecting velocity outside of 35 degrees from Z+), but the main benefit is in aligning the rake structure itself with the local flow to minimise flow blockage and deflection. To this end, the angle of the probes was set based on the velocity vector at the rear of the probe rather than the inlet, with confirmation that the angle did not deviate significantly along any particular probe's length. Total pressure was also extracted from each probe location as the data for direct quantitative validation.
Finally, I also wanted to look at surface flow. We didn't put much emphasis on the importance of this for now, as most of the surface flow on EV23 is either largely unaffected flow, parallel to free-stream, or is completely separated as a result of bluff features (Figure 5).
Figure 5: Surface shear stress magnitude and vector convolution for the main body surface of EV23 as given by the un-validated CFD setup
There is one interesting area with more varied flow behind the from suspension, on the lower bodywork panel. There is a decent amount of vertical deflection of the flow here, being driven by the higher pressure in front of the down-washing floor, and the lower pressure behind the bluff suspension clevises. With no flow separation and a nice new driver suit to protect, it made more sense to use surface tufts here rather than flow-vis paint. I'm not expecting anything profound from this small test, but it's easy and quick to carry out and another chance at picking up any significant discrepancies. The next post will cover the process of designing, manufacturing, and mounting the rakes in preparation for track testing.
Figure 6: Target area for surface flow validation with tufts