[MVRC] Mantium Virtual Racecar Challenge 2016

[CAEdevice]

Hi etsmc, it is difficult to understand the results of your test: maybe you should choose the one with higher downfoce or higher df/dr ratio.

Generally speaking, if you want to optimize the diffuser design, my advice is to use the whole car (not only the rear wheels influence the diffuser design, even the frot wheels have a significant effect), but with only few parameters to control the diffuser (you should start with a very simple geometry in order to reduce the number of cfd tests).

Here is some examples:

1) Diffuser section is a straight line = only one parameter, the exit height.

2) You can add an intermediate point (single curvature diffuser) = two more degrees of freedom, three parameters.

3) another intermediate point (double curvature): five parameters.

Keep in mind that the angle of attack of the diffuser is the most important parameters. With the actual rules and ride heigh, most of cars use a concave diffuser.

Other details that affect the diffuser performance are:

a) Internal strakes: you should start with no strakes or 2 strakes, small advantages come from an higher number of strakes, think about it only after having optimized everything else.

b) Flow behaviour around the gearbox/impact structure geometry

c) Presence of a small gurney flap above the diffuser end (it depends on the general car geometry: with my layout it can reduce the cooling efficiency, so I can't use a gurnay taller than 15mm)

During 2015 I used a complex double curvature diffuser and I obtained worse results than in 2016 with a much more simple one. At the moment I discovered interesting things just optimizing the "straight line diffuser" with only one parameter.

Another advice is to have and archive with geometry and cfd data, in order to not repeat a development path twice.

PS: I considered a constant section when the car is viewed from the side, but last year I saw at least one astonishing "3d" diffuser (I think it was JJRs car). To develop it would require a big number of parameters to be optimised. You could try to guess that kind of geomwtry without cfd, but only if you are aure about the basics (angle of attack and exit height).

[rjsa]

Can you post pictures? Like the #2 from both CP and isosurface?

BUT, just so you don't waste too much time, the performance of the various elements in your car do not perform linearly, the interfere with each other. It means 2 + 2 could be 3, 4 or 5.

So you have a rear wing that adds 3 points of downforce, a diffuser that adds 2. When you combine both you'll likely get 6 points, because the wing usually enhances the diffuser's performance. Or you have a new front wing that gives you 1 point of downforce and a new fender that takes out one point of drag. You join both and still lose the point in drag but the wings loses 2 points instead of gaining 1, since the new fender blocks the flow downstream.

So the best approach would be to design a complete basic car and then start developing the elements.

[rjsa]

Still on car development, I don't know how you guys organise development, here's my take:

I studied Naval architecture, so the concept of "Design Spiral" was hammered into my head:

http://naval-architecture.blogspot.com. ... piral.html

The idea is to repeatedly refine each element of the model, dealing with one aspect at a time, trying always to maximise it's performance while avoiding degrading the performance of other elements, or even better, also improve those element's performance along the way.

I start at the front and do rounds at the front diffuser. Find a best solution then move into the HX. Trying to improve flow without damaging the rest. Then to the rear diffuser and rear wing. Now a good look at the rest of the body for reduced lift and drag. Then start over from the front.

[CAEdevice]

Interesting approach, it is important to test a single detail per time once the general design has been defined.

[LVDH]

Using a simple shape and doing simulations on a diffuser was how I finally understood how a diffuser on a car works. This was many years ago because I was frustrated by what you can read (or hear) about how they work. These simple simulations gave me the knowledge in understanding what happens. So I think it is a good idea to do this.

Unfortunately I have still not worked out a bullet proof way to simulate you cars with second order accuracy. I will try a bit more next week. If it fails we will simply have to go back to where we were last year.

As a bonus I have a special image to the current discussion about car development:

You can see the predicted lap times vs. the design iteration of my car for the 2015 Sepang race. Yes I ran through many design variations. I had a setup that allowed me to hammer out many CAD designs and run them very quickly in a simplified CFD process.
You can see that I was good at creating new parts that improved lap time bit by bit. Then you see the spikes where it goes way up. These are not crazy designs but a design where all the good parts were added to the previous fast car. They almost never worked together as a team! Only at the end of the design process I took smaller and more careful tweaks that then gradually improved the lap time.

[CAEdevice]

Another question:

I'm reviewing my design against the rulebook.

The last paragraph from K3.3: "there must be no gaps in any of the sections formed"

Now this:
http://i.cubeupload.com/62IzN4.jpg
My doubt: A Y normal plane crossing the rounded internal diffuser surface would create more than one separated section, being each on of them separate but closed curves. Is this valid?
http://i.cubeupload.com/LSeqJI.jpg

Hi, I retrieved this post by Rjsa to ask André or Chris to confirm that this geometry is allowed (I am designing something similar).

[CAEdevice]

I have the suspect that during 2017 I used a wrong value for the max front overhang (950mm) ... wich is the correct value, 950mm o 1000mm (from the front wheels axle)?

[CAEdevice]

Using a simple shape and doing simulations on a diffuser was how I finally understood how a diffuser on a car works. This was many years ago because I was frustrated by what you can read (or hear) about how they work. These simple simulations gave me the knowledge in understanding what happens. So I think it is a good idea to do this.

Unfortunately I have still not worked out a bullet proof way to simulate you cars with second order accuracy. I will try a bit more next week. If it fails we will simply have to go back to where we were last year.

As a bonus I have a special image to the current discussion about car development:
http://mantiumchallenge.com/wp-content/ ... ng2015.png
You can see the predicted lap times vs. the design iteration of my car for the 2015 Sepang race. Yes I ran through many design variations. I had a setup that allowed me to hammer out many CAD designs and run them very quickly in a simplified CFD process.
You can see that I was good at creating new parts that improved lap time bit by bit. Then you see the spikes where it goes way up. These are not crazy designs but a design where all the good parts were added to the previous fast car. They almost never worked together as a team! Only at the end of the design process I took smaller and more careful tweaks that then gradually improved the lap time.

Impressive effort for a single race!

[rjsa]

I have the suspect that during 2017 I used a wrong value for the max front overhang (950mm) ... wich is the correct value, 950mm o 1000mm (from the front wheels axle)?

I've been using 1000mm and if fits into the templates provided for 2016.

[CAEdevice]

I have the suspect that during 2017 I used a wrong value for the max front overhang (950mm) ... wich is the correct value, 950mm o 1000mm (from the front wheels axle)?

I've been using 1000mm and if fits into the templates provided for 2016.

Thanks! I hope to gain some df from these 50mm...

[variante]

I studied Naval architecture, so the concept of "Design Spiral" was hammered into my head:

http://naval-architecture.blogspot.com. ... piral.html

The idea is to repeatedly refine each element of the model, dealing with one aspect at a time, trying always to maximise it's performance while avoiding degrading the performance of other elements, or even better, also improve those element's performance along the way.

I start at the front and do rounds at the front diffuser. Find a best solution then move into the HX. Trying to improve flow without damaging the rest. Then to the rear diffuser and rear wing. Now a good look at the rest of the body for reduced lift and drag. Then start over from the front.

I don't quite agree with that for the simple reason that it is not an efficient process (time wise).
Yes, it will grant perfectly satisfying results after n iterations, but it will take a big number of iterations...
It is a trial and error method, which hardly matches with a complex and organically connected system, given its incredibly high number of combinations of geometries and shapes.
I prefer, when designing a new part, to keep an eye on every bit of the car at the same time, or even designing a part just to increase the performance of other parts.

BTW That "Design Spiral" sounds similar to what is adopted by computers for automated optimization processes, which indeed take a huge amount of time. Luckily we (humans) still have an advantage over machines (for the moment) as we are able to get pretty much to the point immediatly.

[CAEdevice]

I deal every with structural optimisation and I think that there is not a "best" method: it depends on the available resources (money, hardware, time, experience...) that in real life are always restricted. Some time an optimization strategy is better, some times a pure guess based on experience is the only possible solution.

For the MVRC I have been working only about details for two years (I will use for the next race the almost the same car I designed last year): I change only one small detail for each run. It doesn't matter if the steps are very small.

[rjsa]

I studied Naval architecture, so the concept of "Design Spiral" was hammered into my head:

http://naval-architecture.blogspot.com. ... piral.html

The idea is to repeatedly refine each element of the model, dealing with one aspect at a time, trying always to maximise it's performance while avoiding degrading the performance of other elements, or even better, also improve those element's performance along the way.

I start at the front and do rounds at the front diffuser. Find a best solution then move into the HX. Trying to improve flow without damaging the rest. Then to the rear diffuser and rear wing. Now a good look at the rest of the body for reduced lift and drag. Then start over from the front.

I don't quite agree with that for the simple reason that it is not an efficient process (time wise).
Yes, it will grant perfectly satisfying results after n iterations, but it will take a big number of iterations...
It is a trial and error method, which hardly matches with a complex and organically connected system, given its incredibly high number of combinations of geometries and shapes.
I prefer, when designing a new part, to keep an eye on every bit of the car at the same time, or even designing a part just to increase the performance of other parts.

BTW That "Design Spiral" sounds similar to what is adopted by computers for automated optimization processes, which indeed take a huge amount of time. Luckily we (humans) still have an advantage over machines (for the moment) as we are able to get pretty much to the point immediately.

It's not that it's completely single minded. Mainly during the first interactions I did go changing a lot at a time based on visual CFD feedback.

But it comes a time when evolution stops and you can't tell what's helping and what's hindering. More often than not things do not work as expected, so taking a methodical approach proved to yield faster and more consistent results. It's not that I'm using shape optimization techniques and generating interactions blindly. Most times each interaction is me trying to get to the point. A few times some methodical variations to find an optimal point.

I run the simulations on a macbook pro that I just used during travels up to now. Each run takes approximately 5 hours. So I'll work on stuff, prep four interactions and run them. Sometimes four variations of of the same part. Sometimes mixed parts to explore new avenues of development. Next day I see what's good and what's dead in the water and start again.

Doing one thing at a time keeps it sane. Going over everything pays off, there's always another drop to squeeze. And trying again after you spent time working on the whole rest of the car always brings new ideas.

[CAEdevice]

I was looking at the Le Mans free practice: the best laptime (Neel Jani, Porsche) is about the same that the best cars of our challenge (TF, JJR) did on the same track last year. This shows how good is our rulebook and the simulation (CFD and laptime) in the MVRC (and our designers as well).

[RicME85]

Very interesting. Is it that the cars are highly efficient or does the engine brute force MVRC cars around the track? (Will be honest, I haven't checked to see how the MVRC engine compares to LMP1-H engines). Or is it something else?