4th gen aero

Yep. Easy enough to include extra features in 2D, a little trickier in 3D since the meshes are much bigger and more complex. Maybe it makes sense for us to put together a list of all the things we think we'd want to tweak to see the effects, so it's considered in the modeling from the get-go? I've modeled myself into a corner before, and had to back up quite a bit to include something I didn't know at first that I needed.

 

Also something to keep in mind - anything that will induce non-uniform flow out of plane will be missed in a 2D analysis. I imagine some of that will be present with stuff like lips, especially at the front corners. At some point we'll definitely want to build out a 3D model and check out how different our drag/downforce predictions are. I'm betting things will get signficantly different once we model things like mirrors, the underbody of the car, and even the rotation of the wheels.

Edited May 16, 2020 by awfulwaffle

Oh - 100 mph run shows a 178 pounds of drag force on the car, and about 380 pounds of lift. If someone has a way to figure out the frontal cross-sectional area of the wagon we can back out a drag coefficient. Wouldn't put as much stock in the lift/downforce value since the underside is perfectly smooth here. Edited May 16, 2020 by awfulwaffle

Oh - 100 mph run shows a 178 pounds of drag force on the car, and about 380 pounds of lift. If someone has a way to figure out the frontal cross-sectional area of the wagon we can back out a drag coefficient. Wouldn't put as much stock in the lift/downforce value since the underside is perfectly smooth here.

 

this is really nerdy, geeky, very cool.

 

From my yarn experiment and watching Kip's engine blow chunks (oil mist came out exhaust), the air exiting from underneath the rear bumper curls up sharply to the point where the rear window was completely coated in oil.

 

The smooth underside in 2d is throwing things off it seems. however the first mostly blue image looks pretty close to Kip's wagon's oil trace.

 

the air catch point at the top of the windshield is interesting. more than I expected.

 

The underside is relatively flat except for the rear from axles back. Not sure how to model that in 2d. The 3d modeling Ive seen uses half the car (splitting down the middle). idk how that affects the compute time.

 

how do you read the legend on the side? idk what the numbers mean nor at what mph the plot is?

things im curious about:

how much air dam can you add before its useless?

how much splitter is too much?

which is better sur dam or splitter?

does adding bigger undertray help?

if add ducts for radiator, is it better out top, side or leave it be?

is the stock rear wing any good?

does removing rear wing help?

how much tray/diffuser do you need before the air doesn't curl up the back window?

If i add a rear wing, where's the best place to put it? and much does it help?

 

other randoms:

usdm pre-facelift nose vs post?

jdm nose pre vs post?

what's best flowing nose?

how much does the extended snorkel on the hood scoop help?

can the interaction of scoop and tmic be modeled? if so, can the cfm be improved?

 

What software is needed to do this?

what's the limiting factor? compute time? configuration? making new parts?

how do we contribute?

When something is found to work, how does it get outputted so madrig can make it?

this is really nerdy, geeky, very cool.

 

From my yarn experiment and watching Kip's engine blow chunks (oil mist came out exhaust), the air exiting from underneath the rear bumper curls up sharply to the point where the rear window was completely coated in oil.

 

The smooth underside in 2d is throwing things off it seems. however the first mostly blue image looks pretty close to Kip's wagon's oil trace.

 

I absolutely believe that the smooth underside is throwing things off. Hell, the air coming off the rotating wheels probably helps kick the flow up the back of the wagon some. That said, with a vertical back like the wagon has there it's nearly impossible for the flow to stay attached because the static pressure is rising so fast, so it's going to separate eventually. It's hard to tell what direction the air's actually flowing from those plots - they're just meant to show the absolute magnitude of the velocity.

 

Here's a zoom in on the velocity vectors at the back. You can see that there's a big recirculating zone there because the flow's separated off the roof and the back bumper. EDIT: Also note in the TKE plot in the plot above that there's still a bit of kinetic energy in this region. While the 'average' flow is recirculating somewhat lazily, there's still significant turbulent fluctuations about this average./EDIT This is typical of any 'dump' style diffuser, which is pretty much what the back of our wagons ends up being.

 

 

the air catch point at the top of the windshield is interesting. more than I expected.

 

Agreed. I think what we're seeing here is the combined effect of compressibility and the slope of the windshield. Looks to me like the flow compresses as it climbs most of the way up the windshield, and when the curvature starts to drop back down towards the roof the density drops a little bit and the speed of the flow picks up as a result.

 

The underside is relatively flat except for the rear from axles back. Not sure how to model that in 2d. The 3d modeling Ive seen uses half the car (splitting down the middle). idk how that affects the compute time.

 

I wonder if we can just model major features - ie the 'scoop' on the underside of the bumper, the exhaust, subframe and driveshaft tunnel. Splitting the model in half is absolutely the right way to go - it'll cut your mesh size down in half, which can have a varying effect on compute time depending on what solver options are used but will reduce computational expense no matter what. Either way, it gives you more compute power to send to the details rather than trying to capture the entire car as a whole.

 

 

how do you read the legend on the side? idk what the numbers mean nor at what mph the plot is?

 

Sorry, everything's in metric because I like to divide by 10. Everything in the above post is at 100 mph, or 44.7 meters per second. The pressure plot is in Pascals (101325 Pa = 14.696 psi), and I have no idea what the equivalent turbulence units are in the US system because I've never wanted to hurt my brain that badly The TKE plot pretty much shows the distribution of kinetic energy carried by the air as it spins randomly in little turbulent vortices.

 

EDIT: Oh derp, I left stuff in scientific notation. Pretty much, any time you see e+##, it means take the number before that, and multiply it by 10 raised to the power of the number. So 6.77e+01 in the velocity plot attached to this post would be 6.77x10^1 = 67.7 meters per second.

 

I can include freedom units in future plots, would be good to have both.

 

 

 

Edited May 16, 2020 by awfulwaffle
added some words because i made a connection

metric is fine. I have to learn a whole new language so conversion doesn't help much. don't convert for my benefit.

 

hmm, that last picture suggests a waist spoiler might be effective.

 

with regards windshield I think the 3d model would show a different result. The only time the windshield has that shape is at center line. The rest rolls off to the side. Rain patterns suggested that.

 

can you do a zoom like that for the nose?

That's a good point on the windshield - I also wonder what it means for the flow that comes down the back of the wagon, if it gets turned out a bit by the sideways rounding of the windshield.

 

On the waist spoiler, I've heard that in the ether before. To the point where people say you don't need a rear wiper anymore on a wagon, because it interrupts the recirculation and keeps rain/snow/dirt from shooting up the back of the car. Worth exploring for sure.

 

Here's the nose shot. Next time I work on the model I'll see if there's a way to add a porous boundary where the grille is so the air doesn't just get stopped dead there - there is an option in 3D but I've never tried to do it in 2D.

What software is needed to do this?

what's the limiting factor? compute time? configuration? making new parts?

how do we contribute?

When something is found to work, how does it get outputted so madrig can make it?

 

My take on it is that our biggest time lag will be making up CAD models for whatever modification parts we want to test out. It's not a huge deal in 2D since it's just tweaking curves, but when we get into 3D it'll balloon if we have to remodel a whole car every time.

 

I think we could probably do it in a modular fashion, sort of how Nonamedude suggested back when. I.e. have a 'baseline' model with all the parts we might be interested in tweaking as separate entities in an assembly, and then extra components that we want to test out can be assembled in and their factory counterparts get hidden when we want to check them out.

 

The good thing about this approach is that as long as the 'baseline' body doesn't change, it shouldn't need to be remeshed every time we make a tweak. We'll just have to update the CFD grid for the new components that get tacked on for a given analysis, and make sure everything still plays nice together.

 

Maybe it'd help to identify any CAD users who are willing to pitch in and are confident they can tackle this kind of stuff? I'm OK, but I work mostly analysis and some software development, so I don't know all the intricate tricks the hardcore CAD users learned to make things go faster/easier.

All this conversation is several magnitudes beyond my understanding, I'm really impressed by how knowledgeable our forum members are and I hope to catch up at some point. I had PM'd Boxkita earlier today and I guess that's what prompted him to start this thread. I had been watching YouTube videos on aerodynamics after listening to a Flatirons tuning podcast on the same subject.

 

In any case, the video I watched that is probably most relevant to this discussion has to do with tools for measuring aerodynamics. The ones that intrigue me the most are the magnehelic gauge (measures pressure differential) and a potentiometer for measure suspension travel (more downforce, more squat).

 

It seems to me that one area of the car that seems like a no-brainer to me is, as mentioned before, the area from the rear axles back (am I wrong?). In any case, between the magnehelic gauge and the potentiometer, either of those tools would quickly move us from theory to reality. I'd be down to invest in these tools once I get a firmer grasp on the subject.

 

And before I forget:

 

[ame=http://www.youtube.com/watch?v=D00-wglvpK0]What tools do you need to aerodynamically test your car modifications? - YouTube[/ame]

Shoot, I'd give up my student license to SolidWorks if I still had it. I haven't played around with it as much as I want to. I put my engineering degree on hold for the time being.

 

Has anyone heard from Nonamedude? I sent him a PM with no reply. Hope he's alright.

That's a really informative video. Super surprised how cheap that mechanical gauge is. At the same time, it'd be nice to have the ability to log/store data with the Arduino. I know I have a hard time peeling my eyes away from the road to keep an eye on oil pressure and AFR, not sure how well trying to look at a needle gauge would work.

Shoot, I'd give up my student license to SolidWorks if I still had it. I haven't played around with it as much as I want to. I put my engineering degree on hold for the time being.

 

Has anyone heard from Nonamedude? I sent him a PM with no reply. Hope he's alright.

 

Been wondering that myself, he sort of just dropped off the face of the earth there.

 

Do they take away Solidworks student licenses that fast? I used my student version of ProE/Creo for like 3 years after I graduated, for personal projects.

i have a.college student email. Until a month ago, my computer was useless except for surfing interwebs.

 

my school teaches solidworks, so theoretically I can get a license.

 

creo student license: https://www.ptc.com/en/academic-program/academic-products/free-software/creo-college-download

 

Link says you get onshape too. Isnt there an online modeling system that lets you use it free as long as you open source your design? Would allow shared dev of components? Still have to import into cfd software at some time cost.

 

Kyle.drives yt about onshape:

Edited May 16, 2020 by boxkita

using linear pots for ride works, however, you have to do some post processing due to the jiggle of the road surface. Also the data flow has to be higher enough to show actual data. Using a laser to measure distance to the ground is another solution.

 

Using an arduino could work as long as you can get the ecu output to match up what's going on. Knowing what the car is supposed to do vs actual is critical.

 

Tbf, gathering data on the road is much harder than modeling in cfd. Ive learned/confirmed more in last 6 hours than years of dinking around. The yarn work Gunnar and I did was pretty risky as explaining to a cop why some dude is in the cargo space looking at flapping yarn while doing 100mph on the highway would have been a tough discussion.

That'd be really cool, to have a common working copy that everyone could work on. No idea what software does that though. It's starting to seem like whatever we end up with, would require some effort to get into CFD-land. At least I can't think of a way to not have to re-build the CFD model every time if we have to convert to STEP as an intermediate. The actual flow model setup won't differ much, but the mesh may need to be tweaked case to case. Maybe we just try and see if we run into a lot of meshing issues or not?

 

Got that first free LGT model you put up into Solidworks and it's some of the way there, but would some work before we can run CFD on it.

Edited May 16, 2020 by awfulwaffle

in a previous life, I wrote automated software packaging and compiler tools. If the process can be scripted/automated, we can use open source tools to generate pretty much everything.

 

onshape exports as step, iges and a variety of other formats - https://cad.onshape.com/help/Content/exporting-files.htm

 

Got that first free LGT model you put up into Solidworks and it's some of the way there, but would some work before we can run CFD on it.

 

what work has to be done? clean up? or more specialized wirk?

updated first post with links software etc

Autodesk Fusion 360 has free 1yr licenses for individuals/students.

I’m playing with it trying to design stuff for my 3D printer.

20+ yrs ago I used to 3D model, working on picking it up again, at least to some extent.

I’m teaching a Graphics class in the fall in VR so I’ll need it

Some of the people on clubroadster.net (a Miata forum) have done similar work and one thing they found was the rear bumper acts like a parachute generating 7 pounds of drag at 75 mph. A number of them have cut the rear bumper away to avoid this. This could be an interesting item to study on a Subaru because the car is wider and a Magnehelic gauge would be ideal for this as there would be no need to drill bodywork to get a measuring point.

using linear pots for ride works, however, you have to do some post processing due to the jiggle of the road surface. Also the data flow has to be higher enough to show actual data. Using a laser to measure distance to the ground is another solution.

 

Using an arduino could work as long as you can get the ecu output to match up what's going on. Knowing what the car is supposed to do vs actual is critical.

 

Tbf, gathering data on the road is much harder than modeling in cfd. Ive learned/confirmed more in last 6 hours than years of dinking around. The yarn work Gunnar and I did was pretty risky as explaining to a cop why some dude is in the cargo space looking at flapping yarn while doing 100mph on the highway would have been a tough discussion.

 

That would be a fun conversation to have for sure. On the Arduino note, think you could sync up with an external trigger of sorts so your logged ECU data would be timestamped more or less the same as the suspension travel data? I've had something similar to this on the back burner for a while so definitely interested in the topic.

in a previous life, I wrote automated software packaging and compiler tools. If the process can be scripted/automated, we can use open source tools to generate pretty much everything.

 

onshape exports as step, iges and a variety of other formats - https://cad.onshape.com/help/Content/exporting-files.htm

 

I'd love to script this if we can - there's already a built-in scripting/journaling environment in the ANSYS tools that I'm using. I will say that the journaling is touchy though, in scenarios where the model changes dramatically run to run.

 

What I mean is that I've encountered problems where due to some curve or part being changed, the indexing of all the faces and vertices in the CAD model gets shifted by some fixed amount. The auto-generated journal points to all the features by their assigned index, so it tries to apply operations meant for entirely different faces to the ones in the 'updated' model and blows up.

 

That said, there's an option to script in Python as well. Maybe we can come up with a way to ensure the indices remain consistent that way. Thinking along the lines of - assign an identifier to each face in CAD that is retained in the .STEP. Python automation script reads in these identifiers with the model, and makes sure to tag them with the same index every time before passing them on for meshing and CFD solution.

 

Here's a decent primer on the subject. I'm currently running in Fluent, but all the solvers mentioned are available to me (I think):

 

https://support.ansys.com/staticassets/ANSYS/Conference/Houston/downloads/Customizing%20ANSYS%20Workbench.pdf

 

what work has to be done? clean up? or more specialized wirk?

 

The model's pretty faceted, and has quite a few facets missing so it's not 'watertight'. We'd need to close it up and convert it to a solid, and then preferably smooth. Then we could add in whatever features we wanted. I had to import the .dae into Blender to generate an .stl that then got read into Solidworks, so maybe I just need to do that better.

 

Autodesk Fusion 360 has free 1yr licenses for individuals/students.

I’m playing with it trying to design stuff for my 3D printer.

20+ yrs ago I used to 3D model, working on picking it up again, at least to some extent.

I’m teaching a Graphics class in the fall in VR so I’ll need it

 

 

Sounds like you're volunteering to help as a professional development exercise!

 

 

 

 

 

Time to dig into Onshape - looks promising

Edited May 16, 2020 by awfulwaffle

Oh, and I'll bring this to the group's attention - there are options for open source CFD that we could explore. I can't run that on the compute cluster at work, but for smaller 2D jobs it's something we could use individually too.

 

Best option in my memory is OpenFOAM: https://www.openfoam.com/

 

I poked around with a little a few years back and the learning curve was moderately steep, but I'm sure between all of us we could figure this out if it's a route we wanted to go.

 

 

Here's a free CAD package as well. It was one of the better-featured options when I checked it out a few years ago, but I still found it a little clunky. Maybe it's gotten better since then, though.

 

https://www.freecadweb.org/

Edited May 16, 2020 by awfulwaffle

Just made a simple part in Onshape, and exported it to Solidworks and it stayed parametric! I saw in their list of export formats that ProE is also one, but I didn't see it as an option when I tried. Wonder if that's behind a paywall...would be nice to have to avoid having to convert to .STEP and losing all model parameters every time we made a change.

Just made a simple part in Onshape, and exported it to Solidworks and it stayed parametric! I saw in their list of export formats that ProE is also one, but I didn't see it as an option when I tried. Wonder if that's behind a paywall...would be nice to have to avoid having to convert to .STEP and losing all model parameters every time we made a change.

 

from my limited reading, onshape puts all the professional formats behind a paywall. A hobby license doesn't give access to it, you need a pro license at 1000$/seat/month if I remember right.

 

Having group workflow and shared files and parametric???, should outweigh the frustrations of something. I forgot my point.

freecad looks good. supports all the formats discussed so far, is parametric, free, has rendering with limited cfd, has scripts.

 

maybe start there?