Over past decade, we have seen more and more development programmes go straight from CFD to production without wind tunnel testing. Reduced development costs are a principle driver in this trend, but increasing computer power has played a massive role too as it enables more detailed models and the ability to easily test over multiple flow conditions.
In a world that bypasses a layer of physical testing, it is important that we create designs that are robust to changes in on-coming flow and recognise that ‘real world’ conditions don’t always match our models.
Although there has been a move towards running transient CFD simulations (where we model flow oscillating with time), a popular method remains the Reynolds Averaged Navier-Stokes (RANS) approach where a ‘steady-state’ or time-averaged approximation of the flow field is calculated. RANS solvers can offer good accuracy at much reduced calculation times.
In a RANS simulation with a straightline (no yaw) flow field and a symmetrical car, we’ll end up with a symmetrical flow field. That is to say, the flow on the left side of the car will be a mirror of that on the right. As a result, we can precisely tune our aerodynamic designs to the on-coming flow. Safe in the knowledge that what we do on the left side of the car will work on the right.
The problem with this is that the car will rarely see a true straightline (symmetrical) flow. There might be a crosswind, or the car is drifting across the track, or perhaps it is overtaking another car. Whatever the reason, our design will never see the conditions it was optimised for.
Consider this simplified example. We’re looking down on two vanes, one being on the left of the car and one on the right. These vanes need to be a mirror of each other so they are symmetrical left to right on the car. We want to minimise drag in our straightline flow field. So the simplest thing to do is make them thin and aligned to the flow.
However, in reality, our vanes will experience a small amount of yaw. Being thin, the flow easily separates from their leading edges creating an unwanted loss in performance. The gains achieved in CFD don’t carry through to the real world and our hard work is rubbished!
If we had started development with a small amount of yaw, we would have introduced some asymmetry into the flow. With the left-hand vane seeing air coming from inboard to outboard, but the right vane seeing it coming outboard to inboard. To handle this, we would have perhaps made the vanes thicker with a more rounded leading edge.
A small change to the model setup has nudged us into making a more robust design. Something that is more likely to work when transferred to the real world.