Maximising the validation value we get within our time and equipment limits for performance of the completed side-floors
The final floor design performed reasonably well in CFD, but not only is CFD typically over-optimistic (often by around 20%, and that's with the assumption of an "ideal" CFD model), the design also relies on a handful of specific flow features to drive that performance (see Surface design). Ensuring these features are behaving as intended was a big focus of the validation process for this year. By this point we had moved on from our prototype pressure sensor boards to a single more dense PCB design (Figure 1), which allowed easier mounting and management of pneumatic tube routing. The original prototypes can be seen in Test equipment (part 1). The new board sits in a housing with permanently connected soft tubing to route from the sensor ports to an external connector outside the housing, onto which stiffer tubes attach. The stiffer tube is preferred for the majority of the length as it holds its shape better and avoids unwanted interference with the driver or sagging onto the ground.
Figure 1: Final design for the PCB that holds the 20 differential pressure sensors, manages the voltage -> serial -> CAN data handling, and outputs to the sensor CANbus loop.
These vortices (Figure 2) are critical to pressure-side performance as they continually pull clean air down onto the top surface of the floor, maximising high pressure and ensuring high energy air is available to bleed through the slot gap for the suction side of the diffuser flap. The outboard vortex also acts as a barrier against the front wheel wake. Both the inboard and outboard vortices are slowly pushed further outboard by the high pressure build-up, which is a potential improvement for next year, but for now the two key areas to validate are:
That the vortices are well-formed and persist downstream of the brow wing, keeping front wheel wake out and effectively supplying clean air to the floor surface
That their outboard "migration" is not premature, which would compromise performance of the rear half of the pressure side
The ideal method of validating point 1 is a total pressure rake, using the Kiel probes first introduced in Test equipment (part 1). Enough resolution will reveal how circular these vortices are, if they have a nice pressure gradient in towards the core, and if we can see a column of higher pressure (ideally within 20% of freestream, as CFD suggests) down the centre of the two onto the floor surface. Satisfaction of each of these key points will provide confidence in the formation and effectiveness of these vortices. The rake cross section was made just large enough to capture the cores of the two vortices while also extending down towards the floor surface sufficiently to see any down-washing and spreading of higher energy air. Fore/aft positioning was a balance between being close to the brow wing to sample the vortices while they are relatively strong (to reduce the influence of measurement uncertainty), while being rearward enough to maximise the value of the data for CFD correlation (and confidence if the correlation is good) by allowing more time for any unexpected changes in vortex behaviour to arise, if present. We decided on a Z coordinate of 0.2 (Figure 3).
Given we were expecting all potential rake locations to be densely sampling small flow features (compared to the much larger wheel wakes that were tested earlier in the year, requiring situation-specific rakes), we wanted to design a "universal" rake that would work in all locations of interest. The most obvious solution was a simple, evenly-spaced grid of parallel Kiel probes. With the three sensors required for the yaw and pitot-static probes, this would leave two spare sensors if we made a 5x3 grid. These were eventually placed in locations that would benefit resolution in key areas for both planned rake locations, specifically inside the existing 5x3 envelope to keep its simple and compact rectangular shape to avoid potentially compromising future use cases.
I ended up deciding to test the second point of validation first, and for this we used flow-vis paint. This would show us where the two vortices were interacting with the surface flow, and thus their approximate path until they either get too weak or too far from the surface. By running this test first, it would allow us to adjust the planned position of the brow wing rake from Z0.2 if we saw the vortices moving outboard much sooner than expected, for example. The seeding area for the flow vis paint was to be aft of the brow wing); having it start from the leading edge may mean it is forced to the sides of the floor too quickly to be useful, and the small area of separation predicted by CFD behind the LE may prevent the paint from being effectively pushed downstream. We ran some tests in the wind tunnel to find the ideal mix of kerosene to chalk powder for use on a slightly imperfect epoxy resin surface, and decided on a 6:4 weight ratio.
Figure 2: Brow wing vortices represented by streamlines and coloured by total pressure (I'm not a fan of the use of streamlines in any decision-making process given their extremely high sensitivity to seed-point positions, but they can provide a nice visualisation to help with context when discussing specific/localised flow features)
Figure 3: Total pressure at Z0.2 above the floor and behind the front left tyre showing the two brow wing vortices and rake probe locations. Note the coarse shading in the high total pressure regions as a result of reduced automated mesh refinement (AMR) activity in these less-critical areas.
The justification behind wanting to validate the diffuser vortices (Figure 4) is very similar to that for the brow wing vortices; the performance of the floor directly relies on these vortices being where they were designed to be. In this case though, it's not about making sure they are providing clean air to a surface, but instead about making sure they:
Are nice and strong with a low pressure core, as this contributes directly to local downforce
Are not expanding, blowing up, or becoming "detached" from the walls and causing blockage effects and/or separation in the diffuser
These vortices are of a similar scale and spacing to the brow wing vortices, hence the suitability of using the same rake. For the diffuser, we wanted to capture the location and size of each vortex as we did with the brow wing, but we also want to look at boundary layer thickness. The top row of probes was therefore positioned to sample just a few millimetres below where CFD predicted the diffuser roof boundary layer extended down to. Lateral position was pretty obvious for this one as it was constrained by the diffuser roof and outboard wall. Fore/aft position was dictated by how easy it would be to mount the rake right on the exit plane, where the rear of the diffuser strake would provide bracing to hold the rake exactly where we wanted it. This meant we were sampling 70mm forward of the exit plane (Z-0.31), being the length of the Kiel probes, which seemed like it would provide the data we wanted (Figure 5).
At this location, the data from the rake alone will make it pretty obvious whether the vortices are in their intended location or not, so I wasn't particularly interested in using flow vis paint to track their entire path. What I was interested in was the shedding behaviour over the central skate and any separation at the leading edge of the strake. The notch in the skate was specifically designed to shed a neat vortex exactly where we wanted to maximise its strength while preventing it from being pulled into the middle of the diffuser. I'd like to see this behaviour regardless, but it will be especially valuable if the rake data has poor correlation, as observing the upstream shedding behaviour of the vortices will help explain any such discrepancies. The seed areas will be immediately forwards of the skate "notch", and the diffuser strake leading edge.
Figure 4: Diffuser vortices represented by streamlines and coloured by total pressure
Figure 5: Total pressure at Z-0.31 showing the two diffuser vortices and rake probe locations
The main purpose of the squirt suppression element was to create downwash and a slightly increased pressure next to the contact patch to minimise tyre squirt. The down-washing element creates two counter-rotating vortices (Figure 6) as a consequence, which are intended to be directed under the skirt for a small amount of local downforce. Prior to raising a section of the skirt inlet, these vortices were split by the leading edge of the skirt which left low-energy air both above and below the skirt, hurting local static pressures quite a bit. We therefore want to make sure that these vortices are ending up underneath the skirt as intended.
It's pretty clear that flow-vis is again the best method to confirm this behaviour. By seeding the whole leading edge of the skate, we'll be able to see where the vortices end up, and if we should be making the lifted inlet section either higher or wider (that is if we decide to keep this squirt suppression concept next year).
Figure 6: Squirt suppression vortices represented by streamlines and coloured by total pressure
Obviously this is the most important aspect to validate, as its directly linked with downforce and drag figures. These were the first tests to be conducted, with this being one of the reasons why. While it is important, it doesn't provide any means of helping to explain poor correlation, and there is a lot of potential for coincidental correlation where static pressure predictions may be accurate, but the flow behaviour causing these pressures may differ. Placing too much trust in a single validation source could mean that future iterations suffer from incorrect assumptions that are carried forward. These are the reasons why we also use flow vis, rakes, tufts, etc.
For static pressure testing, we want to be able to gage an approximate overall downforce from the floor. Rigorously, this would required hundreds of tap locations, but we don't have the track test time available to run all our test plans, then stop and move our 17 available sensors/tubes to new locations, and then repeat the test plans several times. What we want is to be confident in the cause of the observed pressures (from the rake and flow-vis results), so that we can assume the static pressure distribution will be similar to that predicted by CFD. With this assumption, we can use our 17 taps to sample across the "shape" of the distribution. These real data values can then be applied to the CFD distribution to interpolate between them, then the overall downforce estimate can be integrated across the surface and compared to the CFD estimate for the same surface/s. Figure 7 shows the selected tap locations for the suction-side. Only 6 locations were chosen for the pressure side (Figure 8) due to both the expected lower sensitivity (and thus more likely to agree with CFD if the rake and flow-vis correlation is good), and the difficulty of attaching pneumatic tubing to taps to sample the pressure side due to the "2D" design of the side floors (i.e. no internal volume to run tubes through). This won't allow a surface map for determine pressure-side downforce, but will still act as individual data points for direct correlation of some positive pressure areas.
Figure 7: Suction-side static pressure and tap locations (17) shown on the median corner case with its more complex pressure distribution (static pressure is shown here as a magnitude, hence the apparently better performance of the outboard floor due to the increased local velocity)
Figure 8: Pressure-side static pressure and tap locations (6)
This is something we have no predictions on due to the expensive nature of running transient CFD simulations. All we want to achieve here is obtaining data on static pressure changes when the vehicle has a non-zero rotational velocity in the Z (roll) axis, and any increase in boundary layer separation towards the rear of the diffuser that may reduce performance, particularly if re-attachment is delayed. We'll decide if we want to investigate such behaviour more once we see the results. These tests will use the same static pressure taps on the suction-side, but the test plans will differ from focussing on steady-state cornering to testing different roll rates and steering/cornering frequencies. To look for transient separation, wool tufts will be mounted in the rear section of the diffuser and the underside of the diffuser flap and recorded with a camera linked to GPS time, allowing correlation with the transient static pressure data. We're not interested in surface flow vectors here as we have nothing to correlate them with, so once again there won't be any application of optical flow analysis using these tufts.
All steady-state tests will be run for each of the four design cases used throughout 2024 (straight/ride-height, straight/braking, median corner (13.2 m), skidpan (9 m). Placement of the taps and rake probes as shown above were primarily based on post-processed results of the straight/ride-height cases (also as shown above) due to the ease of getting large amounts of steady-state data from track testing for this case. Small adjustments were then made to placement where flow features changed significantly in one or more other cases. One example of a significant difference is the in-washing of the vortices on the outside brow wing during cornering due to yaw flow induced by the vehicle's slip angle. Our full-car CFD setup currently predicts the inboard of the two vortices to merge with the suspension rocker "wake", causing premature blow-up of the vortex (Figure 9) and compromised pressure side performance as a result. This was a focal point during analysis of track data from cornering scenarios, as the confirmation of this undesirable behaviour would result in it being set as a key improvement area for the updated 2025 side floor design.
Figure 9: Total pressure at Z0.1 showing the large area of low energy air sitting above the floor; a consequence of the inboard brow wake merging with the wake from the exposed suspension rocker
Figure 10: CAD model of the new multi-use rake geometry to hold the Kiel probes (see Figures 3 and 5)