Hey there,
It’s Robin from CFD Engine & I hope your week has been jam-packed with CFD goodness? I wanted to take this issue to give you a sneak peak at what a typical project looks like for me.
If you’ve seen my CFD toolkit email you might’ve noticed that there aren’t any optimisation (or design-space exploration) tools on the list. Mainly because I don’t do any optimisation or systematic design-space exploration (we all explore design spaces).
I’ve used these approaches, they just don’t feature in a typical project. In fact, I wonder how widespread they are in our space? Are they a part of your typical project?
So, for the record, I thought I’d outline what a typical project looks like for me. It’s just the two of us, so I’ll give you a peak behind the scenes at where the magic happens 🤣
Behind The Scenes
I’ll use a motorsport example (my comfy place) to illustrate this, but let me just say that…
Projects often get messy, they’re always time-poor, constraint-bound and rarely do they end up with something that an optimizer might call an “optimum.”
Projects are always collaborative, so I’ll use “we” not “I” in the description.
That said, projects tend to go something like this…
The long bit
We’ll start by building a project baseline – our line in the sand. It’s either what the car looks like now (or a best-guess, if it’s not been built yet). Based on the best, most truthful representation of the geometry available (scan or CAD).
Mesh resolutions, boundary conditions & model setup are based on previous experience. They’re developed/iterated/customised at this stage, before being locked-off for the remainder of the project.
With the project goals in mind, we’ll trawl the results of the baseline model looking for a handful of areas that might be improved within the project constraints – a mix of technical regulations, manufacturing capability & budgets (time & money).
Perhaps something is stalled or separated? Maybe the cooling ducts aren’t working properly? Maybe the engine bay isn’t venting? Are the rear wing mounts killing the wing? Maybe the engine intake is in a bad spot – that sort of thing.
The fun bit
Next, we’ll focus on these areas in parallel and throw several crude ideas at each one to see if they’re actually sensitive to change.
During this phase, the baseline model will sprout a set of development branches each looking at a particular area or idea.
Whilst much of this goes on in parallel, there is some logic to it – working front-to-back, for example.
We’ll then start polishing – making smaller, more refined changes – including legalising surfaces and making any changes necessary for manufacture.
Eventually we’ll home in on the winning ideas and begin combining them to see how/if they work together.
The development branches begin to coalesce. Some are left for dead, some are merged, until we have a single concept that we’re proposing for the next stage.
Sometimes a couple of concepts will go into a run-off, but it’s usually a cost-to-manufacture, lead time or styling run-off, as opposed to a performance one. “Yes, we know that’s a better bumper, but is it £80,000 better?” That kind of thing.
The end bit
The previously mentioned “next stage” is often manufacture & then straight to the track. Sometimes things make it to the wind tunnel, but not very often. The consequences of which are twofold:
- we rarely have any experimental data with which to confirm our predictions.
- it better work.
The first one would be nice to have, but it’s the second one that leads to future projects & repeat business.
Not a science project
You might have noticed that this is pretty far from a science project. There’s very little scope for validation and heavy reliance on best-practice & method verification.
Things aren’t slapdash though – every care is taken to ensure that good engineering decisions are being made throughout & that the simulations are of high quality.
For my own benefit
This is what a typical CFD project looks like, chez CFD Engine. It’s not perfect, it’s not science, it’s not fancy & it’s not “right” – but it is useful.
Some of you will recognise your projects in my outline – we are CFD kin 🙏 Some of you will be closer to the multi-objective optimisation / design-space exploration end of the spectrum. I’m getting there, slowly. Will I make it before the robots get us?
I just wanted to put this out there, in case there is anyone else looking at a more complex setup & thinking theirs is not enough. Don’t worry – we can still have some fun with our CFD & help make the next design a good bit better than the last.
Let me know your thoughts, always keen to hear back from you on this or anything else.
I’ll be back next week with something a bit more OpenFOAMy
Until then, stay safe,
