Flawless floors

The two concepts on this page are both results of spontaneous thoughts and ideas that I've had over the last few months regarding floors, diffuser, and ground effect. This post is mostly me just putting these ideas to (digital) paper. As such, they are mostly untested concepts for the time being. I might test various elements of them in the future though, which would warrant a "part 2" post.  

1: Ground-effect-enhanced ground effect

The basic idea of this first concept comes from the Gordon Murray T.50, which uses a high power fan to create a low pressure area immediately aft and above an aggressive diffuser kick point. The resulting pressure gradient helps the air through this rapid change of direction without separating from the diffuser roof, leading to all the benefits of an aggressive and big expansion with no downsides (other than the power to drive the fan). 

I like lightweight and simplistic cars though, possibly a mindset brought about by FSAE, with 160 kg no-aero cars that can sustain 2g cornering. A fan is going to be pretty heavy, and also take up a fair bit of space (as is evident just by looking at the rear of the T.50. It's also loud, especially if your car doesn't have a high-revving V12 in it. Would the same effect that the fan achieved be possible purely by using the energy of the car moving through the air? It wouldn't be as strong or as efficient, but it would be lighter, simpler, and quieter. So, I started coming up with ideas for what is effectively "ground-effect-enhanced ground effect", using a Venturi contraction to provide the low pressure source (not really "ground" effect but it sounds cooler) instead of a fan. The image below shows the layout of the design, which at this point is identical to the implementation on the T.50, except the duct is routed to the rear deck area instead of the rear of the car. 

This does introduce complexities in optimisation of the Venturi area. Where a fan would be optimised for maximum static pressure and/or mass flow, the focus now shifts to effective flow entrainment. 

I see two main concepts for mixing the flow:

1. The duct enters the channel from the base. This is the simplest option and provides no physical flow blockage (excluding momentum effects), but may be less efficient depending on the state of the boundary layer and how it's managed. A turbulent boundary layer would help flow mixing, but the effect of different flow momentums may alter the effectiveness. 

2. The duct extends into the centre of the channel. This would likely improve mixing and thus mass flow, but introduces issues relating to shaping both the duct and the Venturi cross-sectional area profile to avoid significant flow blockage or separation/turbulence/swirl that would decrease the pressure recovery.

There is an obvious limitation of this fan-less method though, being the reliance on vehicle velocity for the amount of suction provided to drive the duct. This undoubtedly means performance potential is reduced at lower speeds. Given the required momentum change of under-floor flow is lower at low speeds though, having the design still achieve its main goal of allowing aggressive expansion with no persisting separation is plausible. Figuring out to what extent, and identifying the presence of any remaining speed sensitivity would need wind tunnel testing.

There is potential for an adjustable, flexible floor profile that could reduce the curvature of the main kick, but this introduces significant mechanical complexities (a sliding overlap section would be needed), and undesirable aero characteristics such as imperfect seals and balance shift. Overall, I can't see any reality where this would be worth it for this particular purpose. 

An alternative, if speed sensitivity is an issue and performance drops significantly at lower speeds, is to keep a more "normal", less aggressive roof profile, and use the Venturi suction for boundary layer removal rather than momentum control, by aligning the duct with the flow in a scoop-like configuration. Boundary layer removal would allow use of a slightly more aggressive diffuser without separation, or would otherwise improve the efficiency of an existing diffuser by increasing the effective cross section. 

It would be particularly useful for double step diffusers, where the concave curvature between the kick points introduces a more significant adverse pressure gradient, which at the very least will thicken any boundary layer. Removing this thickened layer would again increase the efficiency, and also allow the rear kick to be more aggressive. Whether extracting the boundary layer forward or aft of the first kick  (see image below) is most effective would be down to testing, but I expect a more rearwards position may have the additional benefit of creating lower static pressure aft of the main kick point, thus driving under-floor flow harder. Desirable aero balance would also be a factor here though. 

While I haven't conducted any proper testing on these concepts, I did run a quick and dirty 2D CFD analysis to see if it had any credit at all. 

Two geometries were compared, one with a gentle single-kick diffuser and no duct, and one with a more aggressive forward kick with convex aft curvature, tested with and without a duct (the diffuser is still relatively gentle as a result of the test being 2D). Both diffusers had the same overall expansion ratio. The tests were run at 150 kph with standard air properties, and some minor optimisation was done to ensure close to maximum performance was achieved with the single-kick design. 

Unsurprisingly, the more aggressive kick with no duct resulted in significant separation and much worse performance, both overall and in terms of lower peak suction magnitude. Some minor separation was present on the single-kick design close to the exit plane. Results for the aggressive diffuser with duct were promising: Total downforce and drag across the whole floor and diffuser was on par with the more gentle single-kick design (not a particularly good metric as this will differ significantly between 2D and 3D implementations), but peak suction increased by 20%. Further, there was no separation at any point right up to the exit plane, suggesting the profile could be pushed even harder. This suggests an optimised 3D implementation could provide a significant performance improvement.

Total downforce on the top surface didn't change much when both setups had the Venturi present (about 3% lower when the duct was included), although the presence of the duct shifted the balance in this area forwards. The presence of the Venturi compared to just a smooth top surface was to reduce local downforce by again around 3%, and increase drag by a similar amount. It's pretty obvious that the Venturi shape and positioning used below could be vastly improved to reduce drag and limit impact on the surrounding flow. 

These results were promising for the potential of a full 3D implementation. Further improvements would be possible in 3D by incorporating both vertical and lateral expansion, increasing the drag efficiency of the Venturi tunnel. Care would need to be taken not to over-expand the corners though. Positioning is also key, both with respect to the available clean airflow, and the local body curvature. For a road-style car, positioning at the sides of the rear of the cockpit would likely be ideal (where many high-performance cars have rear cooling and/or engine intake inlets. For a prototype body car, the flat area between the rear of the monocoque and the rear wheel housings is prime real estate. 

For aesthetics, the tunnels could be incorporated into the body itself rather than as an appendage on top, which would leave only an entry and exit vent on an otherwise regular-looking body surface. The available internal space may be less forgiving on a prototype car though. 

The image below shows an example location, as well as highlighting the options for the Venturi diffuser. A flat-topped diffuser would be harder to implement as it takes up more internal space, but would improve downstream flow quality by down-washing the expanded flow. 

The main consideration for downstream flow quality is the rear wing. The 2D simulation shows static pressure is decreased slightly downstream of the Venturi, but by the rear of the car, it recovers to within 5% of ambient/gage. Upwashing the Venturi outlet could impose two negative effects, being the flow alignment for the rear wing, and the lifting of losses from the Venturi expansion and the floor duct up to the height of the wing. If the outlet flow is managed well, the rear wing will be negligibly affected and in turn improve the Venturi performance by reducing downstream pressure, thus driving it harder. 

There are a couple other potential sources of improvement that could be implemented to enhance Venturi suction. First is the introduction of vortices closely upstream, providing local suction. Expansion of the Venturi would need to be reduced to avoid over-expanding and blowing up the vortices, but the overall result may be worth it. 

The other option, limited to ICE vehicles, is to use the exhaust flow (and wastegate, if present), which could improve performance particularly at lower velocities. Managing turbulence from the velocity shear layers between exhaust flow and external flow could be difficult without compromising performance significantly. Alternatively, the Venturi could be sectioned to keep the exhaust flow separate from the external airspeed-dependant flow. The effect of exhaust pulsing would mean the pressure magnitude varies significantly, but if there is sufficient mass flow through the duct then the momentum of the flow may be sufficient to dampen this effect somewhat. 

The image below shows two different potential implementations of exhaust flow, depending on car body shape. The first option is cleaner and imposes less flow blockage, but relies on the rear of the cockpit being appropriately shaped and the flow over/around it still being relatively high-energy.

2: Un-sprung floor

This is a completely separate concept, but there's no reason it couldn't be combined with the above ideas to improve performance even further. 

If there's any part on a car that would reap the benefits of un-sprung mounting, it would be the floor and diffuser. They are notorious for being sensitive to vehicle platform motions in and about all axis, so being able to hold it at the perfect pitch and ride height with no roll would offer big benefits over a lap. It's been done by a few FSAE teams, but it's not seen on full-size cars, and that comes down to strength and stiffness. It's not going to take much downforce to start bending a 4m unsupported length of carbon fibre. There's also the issue of what happens when one wheel hits a bump. The challenge here then was to figure out how to get around both of these issues. 

All the solutions I came up with still requires the floor to have significant structural stiffness, but the goal is to keep the impact of added weight and sacrificed space smaller than the benefit obtained from a static floor.

Option 1 is the simplest, and involves simply making the floor exceptionally stiff, probably requiring a couple inches of thickness and a decent core material. Strength of the core material isn't too critical, as the force isn't ever going to be excessive in one concentrated area. This option adds a fair chunk of weight in core material, and eats significantly into available chassis travel distance between the chassis and floor (potentially also limiting how low the chassis can sit, which moves the centre of gravity up). Even then, I doubt an aggressive floor would be possible, limiting the overall performance, making the effort of installing such a system questionable as to whether it's worth it. This method is probably only suited to low speed cars such as those competing in autocross style events, but at that point the focus is better put on suspension optimisation instead. 

Option 2 is to link the floor to the chassis to support some of the vertical load via the suspension. This could be simple and passive through preloaded springs (coil or air), or a hydraulic/pneumatic system could be used to rigidly but adaptably couple the floor and chassis and allow precise response to roll, pitch, and bump chassis displacements. This would negate some of the benefits of an un-sprung setup in that the main suspension still supports some of the aero load, but it retains the benefit of insensitive ground effect. If using a controllable hydraulic/pneumatic system, the option of keeping the floor entirely disconnected from the suspension arises, relying solely on support of a handful of actuators. This would come down to the added weight of the additional actuators and having sufficient space. By linking each corner to the suspension/wheels, support for the centre of the floor may only require 1 actuator, whereas an actuator-only setup would required at least 5 if the same level of support and adjustability is assumed. 

Option 3 is to use forces from air pressure to support the floor.

The simplest and most reliable option I can think of is to seal the floor to the chassis via a flexible material, and draw a suction in the cavity. This could be achieved using a fan, or by the Venturi method proposed above, both of which would allow/provide speed-based control of the suction to support the downforce. This is demonstrated in the image below.

Lastly, the issue of independent wheel travel. If one or more wheels hit a bump such that all four wheels no longer sit on the same plane, there needs to be enough compliance in the system to allow for it. For low-speed cars such as FSAE, the whole floor could be made flexible enough to conform to these motions, given the expected downforce and thus required stiffness is low. For a floor with significant downforce, the options are to either:

• Incorporate an active mount system to adapt the length of the mounting arms according to the travel of that particular wheel relative to the plane of the other three, or to rely on the remaining limited flex in the floor, or

• Rely on the car being setup as a high-downforce car with stiff suspension and short travel, and make do with the remaining flexibility in the floor to conform to bumps. To help reduce the flex required, the rotation axis of the wheel-side mount point (e.g. the upper or lower wishbone/control arm) should be kept as close as possible to the rotation axis of the mount arms on the floor (as seen in the above image). Obviously the relative positions of both axes will change with chassis displacement, but as long as they are reasonably close, the flex required in the floor to accommodate the offset will be minimised. 

Finally, I haven't mentioned sliding skirts here at all. They have big potential for increasing aero efficiency and maximum downforce and would be particularly well utilised on an un-sprung floor, although implementing them on such a design would be very difficult. They would need to be mounted to the floor for best use of space and to keep overall weight low, but this would add more weight to the floor that needs to be supported. The advantage of an un-sprung floor here though is that it can be set very close to the ground, so a skirt would not have to be very tall.