Surface design

Design of the aerodynamic surfaces this year is heavily influenced by the overarching goals relating to gaining experience in experimental data collection, composite manufacture, and CFD setup tuning. As long as we got something generating enough forces to easily measure and compare to CFD, we would satisfy most of our goals for this year and we would be happy enough. With this in mind, I didn't want to spend too much time on optimisation, particularly where it would complicate manufacture (such as 3D airfoils), as these are all "last 10%" sort of things. Even if we wanted maximum performance as a goal, we'd get far more value for time adding simple front and rear wings than maximising the floor's potential. 

In terms of flow field targets, I had nothing specific in mind holistically-speaking as we have no useful flow features from upstream to utilise, nor any aero surfaces or cooling inlets downstream to consider. Generally, I was expecting something like this:

Instead of setting overall flow field goals, I made a list of more local phenomena to focus on:

• Make full use of the clean air channel passing between the body and front tyres

• Keep front wheel wake outboard of this clean area as much as possible

• Keep clean air under the centre of the chassis to maximise downforce in this area where roll sensitivity is at a minimum

• Maximise positive pressure on top of the floor to reduce reliance on the more sensitive suction-side

I had always planned to design a thin surface-based floor rather than aerofoil-based (i.e. no internal volume, just a layup around a single sheet of core material), both for ease of manufacture and for keeping weight minimal. I still wanted to explore an aerofoil-based option in case it turned out to be vastly superior, but I had doubts. The only drawback to this surface-style approach is reduced stiffness, but this didn't end up causing any issues. 

We started at the front to decide where to position the inlet as a first guess. The inlet plane was positioned as far forward as possible to maximise available plan-view area for the floor. The height, width, and position were chosen to utilise the maximum amount of clean mass flow, by superimposing the wheel wake (total pressure) from the four CFD (see below). The inlet shape was constructed to entirely fill the clean window between the wheel, lower wishbone, and rocker wakes common to all four scenarios. 

A side effect of this was the realisation of just how much the size of the lower wake differs between the cases, from different dynamic cambers, steering angle, tyre deformations, and yaw angle. You may also notice the mid wake no longer extends inboard anywhere near as far as it did on EV23; this was caused primarily by moving the brake calliper from the rear to the front of the shell, reducing blockage of the flow enterring and moving outboard through the wheel shell.

Next, the outlet shape was specified. Here, the main limitations were keeping the width inboard of the rear tyre (to avoid having a higher pressure area directly aft of the diffuser), while maximising the height to get the largest expansion ratio (and therefore peak suction) without inducing separation. The rest of the floor bridged the inlet and outlet with a transition down to minimum cross-section area just forward of the midpoint. At the midpoint, width was expanded to maximise the plan-view area near the peak suction location. A "skirt" was extended out to the maximum allowed width to utilise the high pressure build-up in front of the rear tyre, and make it harder for front tyre wake to be sucked down and into the diffuser. 

The first iteration was pretty tame overall, as I would rather start safe and slowly push it harder than start with something messy and have to work backwards. 

Having the front edge of the skirt set back from the main floor inlet was purely rules-based to keep behind the front wheel exclusion box. Before adding any furniture (strakes, winglets, surface discontinuities) or flap elements to the floor, I was targeting a minimum Cl*A of -0.7 for all four test cases, which was just above the cutoff for two 2.5 kg floors costing $1000 to be worth their own weight, cost, and drag. This would justify their use at competition even if we ran out of time to either design or manufacture additional furniture/flaps etc. 

I wasn't expecting much oomph from this first design iteration, and indeed it didn't deliver a blinding performance estimate, with a CLA of -0.33 (although to be fair, it did have to overcome the 5 points of lift that the bare car was making). 

I won't go through the whole design process of getting from here to the final product, but I'll mention a few key things.

At the start, changes were very coarse and almost trial-and-error based as we looked for any promising trends to follow. Once finding a few design features that seemed to provide performance improvements, we started to post-process results more carefully and analyse exactly what the flow was doing, what was causing the trends, what an "ideal" flow would look like for the trend being followed, and ultimately how we could adjust the surface design to achieve it. Generally, the straight airflow scenario was used to identify trends as it was the least computationally expensive to run. The cornering cases were then used to fine-tune and identify sensitivities. Since the lack of a suitable wind tunnel for vehicle testing meant we had to rely so heavily on CFD, this sort of critical and analytical thinking was key to minimising the number of simulations required to keep things moving forward on schedule. I doubled down on this once we started getting close to a final design, and imposed a rule that no simulations were to be run without brief documentation of what the aim was, how any changes to surface design in the simulation were expected to achieve this, and if it was not achieved then what post-processing was to be analysed to determine why. This reduced time wasted on chasing spur-of-the-moment design changes where improvements were more reliant on trial-and-error. 

Eventually we arrived at a design with an expected Cl*A of -0.74 in straight airflow, -0.71 in the median corner case, -0.72 in the skidpan case, and 0.68 in the braking case. Losing less than 5% total downforce in the cornering cases was a promising result. We accepted this and began adding and optimising additional features. The scope for surface design wasn't too limited, but I also didn't want to go overboard not only because of the diminishing returns, but also because we have no validation data for complex flows or strong vortices yet. For this same reason though, I wanted some interesting flow features so we could actually analyse them experimentally and tune the CFD setup accordingly. 

The final design after 4 weeks of continuous cycles between CFD and concept sketches is shown in the two images below. 

CFD performance in an ideal straight airflow (18 m/s) scenario:

Cl*A -1.08
Cd*A 0.61
%Rear 84

Using the uncertainty from the EV23 CFD setup, this puts predicted lift somewhere in the range of -1.04 to -1.12, including a larger factor of safety of 50% as a crude way of accounting for the somewhat questionable verification processes that was used (see Certainly not for a new and much more rigorous verification process).

Maximum CFD downforce efficiency is achieved in the skidpan scenario thanks to some cornering-specific design optimisation, in which Cl*A reaches -1.14, with a minor drag increase and 1% forward aero balance shift. 

The downforce figures don't quite match our downforce goals, and there's probably scope for another 20% downforce with changes to the current design, particularly in better front wheel wake management and more utilisation of the skirt area. I also suspect the current design could have reached the original goal as-is if not for a number of limitations we encountered. I'll get to the limitations later, but first I'll briefly run through the surface design and the functions of each design feature.

A little note on CFD setup too before I move on:

All floor aero surfaces utilised a wall-resolved approach targeting y+ = 1. The choice of a resolved approach came from a combination of two factors:

• Minor changes to separation behaviour were observed in the diffuser compared to a wall-modelled approach

• Enabling of a smoother mesh size transition, considering the floor cell surface size and vortex refinement volume size, which were otherwise both smaller than an ideal wall-modelled height of y+ > 30

Prism layers were initially set to extend to the expected local boundary layer thickness and later adjusted in areas where boundary layer thickening/separation were found to occur. CAD thicknesses of all structural composite surfaces were 3mm (3-4 layers of carbon and 2mm core material), and furniture surfaces were kept to 0.5mm (2 layers of carbon with no core material). Trailing edges were given non-zero thickness to reflect manufacturing limitations, and this also gave an easy way to apply a smaller local mesh size in these areas.

Right, time to get onto the limitations we found.

My downforce goal (Cl*A = -1.2) ended up being a little high, but only just. We would have easily achieved it with a lower ride height giving increased ground effect (aero mapping showed a 7 point downforce increase just by lowering the ride height a uniform 5mm), but we have to maintain quite a high ride height (currently about 55mm) with our car to accommodate the bumps and high compression zones on our hilly test track without the compromise of overly stiff suspension. We also want to be able to clear the very steep top of our workshop driveway, and straddle the concrete curbs at our test track to avoid damage to the chassis and/or accumulator/motor mounts if a driver were to lose control on what is a very narrow (~5m) track. 

A similar issue was the higher than expected drag moment, effectively trying to rotate the car back onto the rear wheels. While a target aero balance of 70% seemed perfectly achievable by focussing downforce near the centre of the car, quite a bit of force is required to overcome the moment applied by drag. With the modest downforce a floor provides, this means the peak downforce needs to be focussed much further forwards than one might initially expect. I had planned for this by pushing the throat of the floor forwards and by positioning the whole structure as far forward as possible, but it simply wasn't enough to overcome the drag and bring the aero balance to the desired 70%. 

The biggest improvement here which will be brought up for next year's chassis design is to bring the front of the chassis down lower, lowering the height of stagnation pressure region at the nose, and bringing more of the chassis into increased ground effect. A quick simulation suggested this alone could bring the aero balance forward from 84 to 76 with negligible change to drag and a 4 point downforce increase. The current nose was designed as it is because of the ergonomics being carried forward from the previous vehicle iteration, where that year's design philosophy saw ergonomics getting priority over smaller chassis optimisations for mass or VCG (and when aerodynamic performance was not being considered). 

Overall I'm reasonably happy with the design. It's got decent CFD performance (particularly for its estimated/target weight of 2.5 kg total), but there's a lot to be improved on performance wise, some of which can be looked into next year. The outboard half of the floor can definitely be worked much harder than it is currently, and I'd love to do away with the brow wing as it's a rather inelegant, fragile, and drag-inducing solution to a problem that may not even have to exist, should the right design be found.