awfulwaffle

Murray clamps, I bet

.stl and .obj are good for me, and I can do any of the formats that comprise a file with a set of points defined. The former is easier to work with, but none of them are particularly easy from the standpoint of making changes down the line. We'll have to take steps to parameterize certain features as we go and see the need, I think.

I literally have a tote full of OEM headers and cross-pipes. Also have a guy in my area who ports the headers.

Don't make me get the paddle! Cuz I'm one to talk when it comes to letting things lag lol

Ooooo and we have access to pre/post facelift wagons and a facelifted sedan!

Count me in, I'd put up some cash if we can get a nice detailed scan of the whole car, including underside and engine bay. Unfortunately not. Restricted to an on-site license server. Helps keep the Russians off our intranet Sounds like a cool opportunity to learn as long as we all stay motivated etc. All I've ever used is commercial flow solvers, or very specific custom-written codes for particular applications like Ajerezcl was talking about. Getting us some familiarity with a heftier open source code like FOAM would be rad. I remember doing a comparison of available codes during one of the CFD courses I took in grad school, and openFOAM was pretty well featured relatively speaking. I'm sure it's only improved since then. Sounds like the hypothetical task list is as follows, correct me if I'm wrong: 1. Put out a call to see if there are enough interested parties to crowdfund having an LGT 3D scanned. Can we do a full car (whatever's available) and then patch in partial scans for pre/post facelift differences if we want to later? 2. Clean scanned geometry, defeature as needed and generate grid in some common format. 3. Get started learning the openFOAM interface in preparation for running the full-size sim. Maybe whoever does this can try and duplicate some of the 2D results we have, just to get the workflow down? 4. Figure out what conditions we're interested to start with, and build the openFOAM cases. 5. Try submitting jobs to Ajerezcl's cluster 6. .....? 7. Profit

ANSYS does offer student licenses, but I believe they limit the size of the models you can run which would make it a non-starter in this case. I'm game for openFOAM, has anyone here used it before? Regarding model - where did we leave off? I remember we found some decent 3D models that you had to pay for. Could buy one and spend some time adding features that are missing? Engine bay, underside cavities, etc? I'd gladly spend a weekend or two building a quality grid off of the model once we have it. Maybe we can award the first ever Legacygt.com PhD if someone wants to write us a 3D viscous compressible flow solver just for this thread

So if I understand correctly, the out-of-production environment consists of two 10 core/20 thread nodes, with 64 GB of total memory? If so, I think we could absolutely work with that if we manage grid size carefully. It'll be a little on the slow side, but I'd think we could turn around a reasonably-resolved RANS run in a day or two on that setup. Ideally, we'd use ANSYS Fluent, is that available? If not, I'm sure we can adapt to whatever is available based on the talent pool in this thread/on the forum. Btw, your English is just fine from my perspective

Fantastic. How would we handle credentials/access? Worst case, I can handle grid generation and case setup on my end and pass files over ftp or similar.

That seems very low on a VF52 based on what I've seen as well. In order for that to be the case, your compressor would have to be way off-island on the flow/efficiency map, meaning you're making target pressure at a much lower density (higher temperature). If the turbo itself is fine, this would imply a significant post-turbo air leak in my book or some other parasitic loss that's causing the compressor to work way harder than it should be. The other option is a pretty nasty aero issue somewhere in the compression section of the turbo. Meaning, bad flowpaths on either the wheel or compressor cover, whether due to damage or manufacturing issues. This would drive up losses and effectively mean you're operating on a different (and crappier) map than the VF52 should be. Were quality parts used in the rebuild, and is there any visible damage anywhere?

I would datalog the car and proceed from there. Without that, we'd only be guessing and you'd just be trying things until something worked. Do you have means to log the car? For what it's worth, if the car is tuned for 100 RON, US gas won't suffice and you'd be knocking and pulling timing.

We gonna fix the saggy butt. That thing would be positively tucked if it had the Eibachs front and back.

I had Mahle turbos recommended to me by a local Subaru master tech, the ones available on RockAuto. So far, I have 2 customers running them without issue, older one for about a year. Seems like a reasonable cheaper alternative to OEM. Their VF52 is like 600 bucks and the quality looks top notch. EDIT: Scratch that, looks like prices have gone up. Still, 700-800 bucks. If you need a working, used VF46 to replace the stock turbo on your car, I have one around somewhere.

Could be something, could be nothing. Fuel trims look good, IAM pull obviously not great. How's the car otherwise? Any new noises, oil burn etc? I used to get random knock events from a chattering puck clutch, trans dipstick rattling on TMIC, etc

Lost every gear but 3rd driving in to work, which was suddenly accessible with the shifter in the neutral position. Turns out the plate that clamps the big end shift linkage bushing to the actuator rod on the trans opted for early retirement - completely vanished but the retaining nut was still torqued in place. Upgraded to a random conical washer I found on the floor. The trash pile giveth.

Shoot, I think I owe you buckets. Moving my shop right now and completely spaced. I'll dig em out of a tote next time I'm at the new place.

No parents no rules!!

Used it to get parts for my shop truck.......seems backwards

Consider a cocaine addiction as a cheaper alternative.

Interesting datapoints. I've noted that the FMIC won't heatsoak at all, but will consistently sit 5-6 degF over what the dash reads as ambient temperature. Wonder if the hot side piping picking up heat from the bay is responsible, would explain the difference vs the Grimmspeed TMIC.

Seconded. Note that the 340 lph rating is attainable only at the rated voltage, which our cars definitely don't supply. I ended up following Underdog's rewire and picked up a good margin (IIRC the voltage at the pump went from ~8.5V to ~11V at 100% DC). Also learned through that process that the FPCM from an R34 GTR is plug and play with our cars, and provides more chooch up top than the Subaru-spec FPCM. Ran my car with one for a bit and everything, as a shopmate had one sitting around. It's about 2-3x the size.

Finished re-rolling my e85 map for the 16G, as I wait to make sure the car's not gonna blow up before getting the x500 on. Same timing, same AFR targets, same boost targets, just massaged boost control and turbo dynamics for the FMIC and ELH. Verdict is: lost about 400 RPM of spool, lost about 20 whp at peak but picked up about 40 ft-lbs according to VirtualDyno runs on the same stretch of road. Power stays nearly flat up to redline (maybe 20 hp drop from peak), versus the much larger drop with the crappy Perrin-style TMIC and factory UELs. Pretty neat.

Mine's currently full of a quart each of cheapo 75w90, 75w140 and Lucas Oil Stabilizer.... Shifts like butter, but still sounds like a truck.

Yep, 2D sketches - CAD surfaces generated from FSM side views. Last I looked at it, OpenFOAM implemented many of the same spatial discretization and turbulence models available in Fluent, though maybe to a less sophisiticated degree. Can't recall the options for pressure-velocity coupling and other solver algorithms offhand. I'd love to get it running on my home box as well, and I think it should be plenty usable for 2D runs as well as limited domain studies (ie portions of the engine bay, underbody, etc). Even if we can get 3D CAD geometry, I think we'll have a tough time running full 3D models of a car on any typical local PC. To get meaningful resolution on even a halved/symmetric model, I think we'd be looking at grids on the order of 20+ million tetrahedral elements - maybe as low as 10 million if we do some selective defeaturing and grid coarsening. From experience, grids of this size run pretty dang slow on a local machine, and are memory intensive. For reference - last analysis of this scale I ran locally on my workstation took about a week to run. That machine is a 16 core Xeon box with 128 GB of RAM, and memory load was near 50%. PC sounds like a jet engine and is borderline useless while the solve is running. We could look into polyhedral grid generators, if OpenFOAM supports that type of grid. That will certainly help with element count and solution time, though I don't think it'll really make much of a difference in terms of memory usage if I recall correctly. I'm out of school now and my car hasn't re-exploded again (yet), so I'd love to help get this going again. When I'm overloaded I tend to spend what little spare time I have drinking beers and staring at a wall, which is mostly why I went quiet last time around. Having folks working together will make me feel like a POS for doing that, great for personal motivation I do also have the Bluetooth anemometer I bought a while back for this project, for getting some test data. I also believe I owe everyone a 2D parametric study, with an input parameter list that we defined somewhere in this thread.

Yeehaw, show that wallet who's boss! Looking forward to following the build