Not read this yet, but there was a pretty cool chat with Chung on using the Chung method in the Performance Process podcast the other day. Talk about a nerdy dad-joke bunch. Quite possibly locked down to users only, not sure. Ha ha, just checked, 27 mins is free, the other hour you have to subscribe to. His method is pretty achievable, as long as you have a power meter and a good speed sensor… and a gps. I did it once on an outside velodrome but it was typically windy as hell.
basically “learn how to use OpenFOAM” (open source, free CFD), or “pay a small amount of money for a web service that’s a wrapper around OpenFOAM” (Airshaper, etc). Alternatively, “Pirate ANSYS CFX or ANSYS Fluent”.
(I have a fair amount of experience with CFX from back in the day, unsure how OpenFOAM compares but because it’s free, it’s what everyone is building off of nowadays)
1 Like
No?
there are a lot of non-linear aspects to aerodynamics. you can’t easily extrapolate from a small number of samples, even with machine learning or AI.
I thought that’s what your emoji were indicating, but wasn’t sure.
At this point, l begrudgingly admit that machine learning approaches to weather forecasting and atmospheric chemistry modeling have worked surprisingly well. The physics-based models are solving stiff systems of differential equations with some spectacular nonlinearity. I’m curious if there’s something that makes small-scale CFD even more heinous.
Maybe I’m behind the times with state of the art ML for weather forecasting but my understanding is that they work well because A) we have a truly massive volume of real data to train them on, and B) the landscape under the weather can relatively safely assumed to be static. So a statistical approach to analyzing past information can result in a model that can successfully extrapolate to predict future conditions.
but if we plopped Mt. Everest down in Iowa, the model would suddenly be unable to reliably predict the weather at all, because it’s dependent on real-world conditions resembling its training data, right?
If you trained an ML model on CFD results of a Cessna 172 in cruise, I don’t think it would be able to extrapolate out to good results for the flaps at 10, 20, and 30 degrees. If you trained it on a full suite of CFD results for the entire flight regime of a high-wing single engine like a 172, is the model going to give good results for a low-wing twin engine like a Cessna 310? I’m not sure if you could generalize an ML airflow model; it feels like you’d have to give it so much training data on your specific case that you might as well just do the classic CFD on the discrete cases that you care about.
4 Likes
ML doesn’t do very well with non-linear functions, in my experience. We have some control programs for power flow, but all of the ML experiments didn’t really pan out.
this is interesting, and also pricey.
1 Like
I have a deep distrust for that type of rack strut where the fixing is a clamped tube rather than a bolt through.
I’m assuming it is probably fine given the number of ppl making them and I know it is for better adjustment but I still am uncomfortable with it.
IDK about the one pictured, but the Salsa ones are quite solid! blocks bite into the aluminum struts a bit and seem to hold quite well. I’ve taken mine loaded on some super rough stuff and never had anything wiggle loose.
I have a bit of distrust as well. So far, the Radlawn rack I have with that setup has been fine. I have thought about buying some 8mm ID split collars to put on the struts just below the rack as a backup.
I’ve had great luck with that kind of rack strut clamping, never any issues.
HOWEVER, I hate the parallelogram arrangement of support struts and it seems like that rack does not come with a tang, which is borderline criminal in my mind.
These little basket hold downs are neat.
2 Likes
Bicycles are made out of triangles
1 Like
Yeah how is there not a fork crown attachment
1 Like
it has the ability to add one, just not included for ?? reasons.
Odd that there’s no drop-down for “include a tang”, and not even any “related products” list across the bottom of the page for the tang or p-clamps or basket.
Edit holy cow, try to find a rack tang to buy from them. Searching “tang” get’s you a zillion Tange products like tubes and BBs. You need to go through 4 levels of drop-down menu to get to rack accessories, THEN click to page 2, and THEN you find three rack tangs…except zero of them even have “tang” in the product name!
All three are on this page:
3 Likes
Their web store is the worst I’ve used (so far) and their customer service isn’t any better. My last experience with them was so bad that I’m off buying any Soma and related brands completely.
that’s a bummer, i’ve only ever had great customer service from them
I bought one of those extended RDs to make GRX work with huge cassettes… and they sat on my order for 6 weeks. It took four phone calls and emails to work out that yes, I paid enough on the web shop for shipping to EU, and yes, you can send it now.
The place is like a time capsule of 70’s mom and pop shops. I wonder how it all works out.
1 Like
My first order with them was fine. The second order was not and it took 2.5 months and multiple emails to get an RMA number so it could be returned, then they didn’t offer to send the correct part or refund the original shipping cost after I asked. Eventually, I gave up.
1 Like