8 hours ago, SAS443 said:

The user singlecoil

That they did. I don't recall it being 100% free but the bigger impediment is they also happen to be on the opposite side of a not-so-small country.

What is stopping you from buying these cheap sensors yourself and just head to the closest flying club/flight school and collect the data?

Didn't you also have friends that flew C172s? 

This is quite easy to solve without too much hassle.

Be part of the solution instead of the problem.

 

 

Edited December 3, 20241 yr by SAS443

21 minutes ago, SAS443 said:

What is stopping you from buying these cheap sensors yourself

Indeed I already have one of them. Has this really never been attempted before?

Also, ditto. Get over it: learn to code. Repent of your self-proclaimed lack of engineering skills. You clearly have the best access of anyone. What's your problem here?

20 hours ago, jcomm said:

It is a very interesting subject and one I've tried to learn more about, specially after FS 2020's release, but also when Totorito released the CL 650, because he did use CFD to model some of the characteristics of the real aircraft trying to approach the XP model the closest possible to RW numbers.

That is interesting, I wasn't aware of that but not surprising.
Using CFD to Develop NASA’s X-57 Maxwell Flight Simulator(https://ntrs.nasa.gov/api/citations/20190034017/downloads/20190034017.pdf)

20 hours ago, jcomm said:

There is no winner IMHO.

If you are comparing xplane 12 and msfs 2024, I can't argue that. 

However, if you are talking exclusively about CFD vs BET and the application to flight sim, then CFD is the clear winner from what I see. As I said, it is much slower computationally at the moment but there is a big push to be able to do a lot of it in real-time while also improving the predictions. One of these efforts is "CFD Vision 2030", which is a study commissioned by NASA in collaboration with Boeing.

https://ntrs.nasa.gov/api/citations/20160006942/downloads/20160006942.pdf 

18 hours ago, blingthinger said:

It absolutely can and does. But going back your car example, you cannot predict wildly different geometries that were not in the training set. You cannot train the AI model on a 1000 Cessna-like airframes and 1000 Airbus-like airframes and then expect it to accurately predict a Gripen.

I understand, but the idea is not to make one model that can predict the characteristics of all geometry. Instead, they are training the models on what they are interested in.  So, a different dataset would be needed for delta wings vs A320,B737,A320. 

1 hour ago, brinx said:

I understand, but the idea is not to make one model that can predict the characteristics of all geometry

No you don't. In the context of a consumer sim this is exactly what would need to happen. By default you need a model that can handle any geometry a developer might toss at it.

Even a specialized, geometry-specific model would also then have to factor in all possible flight control positions and sizes, landing gear types, operating points, on and on, for each and every wing geometry it gets trained on. It's true that over time we might build up a database like this, but it will take decades. And then you better pray that the organization that generated it, also kindly releases it for public consumption at a reasonable price.

 

1 hour ago, brinx said:

there is a big push to be able to do a lot of it in real-time while also improving the predictions

I'm sure there is but it's not referenced in that NASA link. Slide 26 does a great job of listing the computational requirements for truly correction-less CFD: many many petaflops. That's not appearing real-time any time soon. And if anything, that NASA-commissioned committee concluded that investment in CFD advancement has decreased. This is not a warm fuzzy 'ooh this is where we'll be in 2030' pdf file. This is a 'call to arms' to continue pushing in a corner of the industry that they claim has slowed down a bit.

 

20 hours ago, blingthinger said:

Absolutely not. Or at least not by itself. A 6DOF accelerometer and GPS is nearly dirt cheap these days. I am very very surprised that of the many people floating these and other sim forums, with all manner of official-looking license lettering and certification in their account Signature below, no one has once done a detailed data acquisition study with any airframe. Put it on github: data and all the details needed to recreate it in the sim. How has this not happened yet? Not even with a good-sized R/C aircraft even? Maybe it does exist and has been lost in the wind?

You completely misunderstood the intention of the review. The instructors are the independent QA group who ensure that neither of the C172 flight model teams pulled any funnies that sacrifice flyability and handling for nudging just one more datapoint closer to 99%. And even in flight simulators, there's a certain subjective element about flight handling and behavior, so human review still makes sense. If an instructor on the actual aircraft says that it slightly shudders twice before stalling on a hot and humid day and a flight model replicates that without some dirty plugin tricks, it's a definite plus for said modeling method.

Edited December 3, 20241 yr by Bjoern

30 minutes ago, Bjoern said:

subjective element about flight handling and behavior

No, no I got you. It sounds like exactly the perfect type of test that somehow still seems to have gone untried. I'm not strictly opposed to getting their opinions, but if that's the only judging metric then there are still gaping holes in the conclusions. There's already opinions a-plenty from well (and not so well) intentioned seasoned pilots out there.

6 hours ago, brinx said:

However, if you are talking exclusively about CFD vs BET and the application to flight sim, then CFD is the clear winner from what I see. As I said, it is much slower computationally at the moment but there is a big push to be able to do a lot of it in real-time while also improving the predictions. One of these efforts is "CFD Vision 2030", which is a study commissioned by NASA in collaboration with Boeing.

You keep on making improper comparisons. Any mention of "CFD" you will find on the web, is extremely different (literally, in terms of order of magnitudes) from the "CFD" which MS uses as a term to indicate their flight model. So you can't extrapolate the characteristics and advantages of one to the other.

In other words, the fact that full-blown CFD produces, in much longer time frames, much more accurate results than what BET can produce in a few milliseconds, does not imply that an extremely simplified form of CFD will produce, in a few milliseconds, better results than BET. Indeed, it may very well be the opposite, we don't know.

So forget most of what you read on the web about CFD and BET, mainly because the use cases of professional/academic CFD and BET mentioned anywhere on the web, are NOT real time flight simulators. CFD there, is mostly used to produce aerodynamic coefficients which are then input into look-up tables flight models for real time applications.

Now, I understand the point you made in the original post, i.e. that with increased processing power, more complex (and accurate) CFD might be used for real time flight simulators. Indeed, that could eventually happen, but it's certainly not something which is coming soon.

And of course, nothing is set in stone: maybe, with increasing processing power available, Laminar Research will progressively integrate its flight model with CFD elements in future releases. And in any case, remember that the present flight models of XP and MSFS also contain look-up tables (airfoil profiles, etc.). So it's not all black and white: CFD vs BET vs Lookup tables, etc. That's why I think this discussion is mostly a moot point.

Edited December 3, 20241 yr by Murmur

1 hour ago, Murmur said:

CFD there, is mostly used to produce aerodynamic coefficients which are then input into look-up tables flight models for real time applications.

Heaping more onto this thought: extracting all those coefficients from that batch of high quality CFD runs will also therefore mean that all of those 3D corrections that sas443 mentioned previously are baked in by default. Dynamics, tip losses, induced AoA changes from control surface deflections, compressibility, etc. etc... All that will be included in those coefficients without having to do any more effort. No corrections needed (because BET theory doesn't specify where the coefficients must come from). Voila! Fully "isolated" 😉 BET!

That said, specifically in that case for XP you'd also then want to bypass Austin's special sprinklings, (but seriously he's in pretty good company with Prandtl, Glauert, Oswald, and Friends) and just get down to that raw, brutish, "crude" BET.

This example is also precisely where 2020/4's algorithm loses fidelity. From all those high quality CFD runs, you'll also be able to extract the trusty FSX curves. And specifically because of  the blending function that merges the out-to-lunch forces coming from asoboCFD with those FSX curves, you lose information about the airframe state that came from that pristine, perfect CFD. BET does not suffer that fate.

afaik

CFD only gives you pressures around the object.

For flight simulation

You still need a way to convert those into net forces on the object.

And then convert those net forces into accelerations.

while handling the fact most users dont have real time operating systems, so the acceleration/velocities they see is unlikely to match the calculated acceleration/velocities.

Xplane is what it is simply because Laminar have put the work in to fix most if not all of the issues associated with that process, CFD isn't a shortcut to putting in the hard work, it's just a good starting point for a flight simulation system that doesn't use lookup tables.

Edited December 4, 20241 yr by mSparks

Asobo : Video game developer 

Laminar Research - dedicated to providing software that accurately reflects the laws of physics (how accurate is debatable" 

 "from Google baba" 

17 hours ago, blingthinger said:

Indeed I already have one of them. Has this really never been attempted before?

Also, ditto. Get over it: learn to code. Repent of your self-proclaimed lack of engineering skills. You clearly have the best access of anyone. What's your problem here?

You have been on record, on and off that you are interested in this data. you even have the device needed for this scientific exercise. So if we stay within the scientific context (which you feel very comfortable in), the task of gathering data typically falls under the responsibility of the requester (that would be you), no?

Instead you went off deflecting everything onto others. And not even bothering to acknowledge the solutions I offered you?

Just call the nearest flight school. You will not be denied, so "access" is not an issue. And you can set-up all gizmos as you wish, this helps to ensure the data is relevant and tailored to your specific needs. You have yet to provide why this is not a suitable option. In fact there are so many benefits to it.

If you are unwilling to put in the effort, this particular discussion will not progress any further. 

 

4 hours ago, SAS443 said:

this particular discussion will not progress any further. 

hahahaha Duly noted. Welcome to being "part of the problem", as you say.

 

7 hours ago, mSparks said:

CFD only gives you pressures around the object.

It's not exactly 'only'. That's all you technically need on the surface of the airframe itself to get forces and rotational moments. Velocity is also there but happens to be zero because of the no-slip wall boundary condition.

You do also get the velocity and pressure field in the surrounding 3D volume including boundary layer formation and growth (a huge deal and also something asoboCFD is missing entirely). Don't get one without the other. That's what the Navier-Stokes eqns do: couple those 2 together.

That's also where e.g. wakes appear on the scene. You'll see the influence of wing on empennage. BET doesn't inherently account for that unless you either actively monitor a trajectory and tweak any downstream coefficients (what Austin does) or use a full CFD model to create the coefficients.

Temperature is there too if you've got the energy equation in the mix which is needed for compressible flows (Mach > 0.3ish)

Edited December 4, 20241 yr by blingthinger

10 hours ago, mSparks said:

You still need a way to convert those into net forces on the object.

Definitely. Even without being an expert in this field, I was easily able to discover how this is done(e.g lift, drag) using the viscous forces and the pressure forces that are being exerted by the fluid on the surfaces of the object, and how to extract the desired coefficients. 

It is quite clear to me, why in the past, BET was the best approach to take when building the foundation of a sim. It is also quite clear to me, with rapid advancement in technology, computational fluid dynamics is now the best approach to take in 2024 if building a new sim. I'm not going to re-argue these points again.

There is one user, for some reason, is bending himself into pretzel trying to find a reason why this cannot be. The facts are laid out, and everyone can decide if they were starting fresh in 2024 which approach they'd invest money in.

10 hours ago, mSparks said:

Xplane is what it is simply because Laminar have put the work in to fix most if not all of the issues associated with that process, CFD isn't a shortcut to putting in the hard work, it's just a good starting point for a flight simulation system that doesn't use lookup tables.

Guess what? I agree with you. CFD isn't a shortcut from putting in hard work. It is just a modern approach that is only now becoming possible with the advancement in technology such as GPUs.

 

As it relates to Xplane, if Laminar does add CFD in the future, that would be great. If they don't that is fine too, Austin has spent decades fine tuning what he can do with BET. 

I've reached my conclusion, and I see no further reason for me to continue discussing it.

 

Edited December 4, 20241 yr by brinx

2 hours ago, brinx said:

computational fluid dynamics is now the best approach to take in 2024 if building a new sim

It will be interesting to watch Flightgear over the coming years. Maybe someone over there will toss in a wild experiment to see what happens. Your link to that NASA pdf was pretty clear in showing why it's still a fool's errand. Plus there's lots of proof of that failure out there right now. But hey, for once I do agree with PMDG's Randazzo in that CFD in a commercial sim is little more than chasing a 'buzzword'.

Regardless, when it does appear for real in coming decades, we'll all be able to look back at the Flight Unlimited devs and thank them for breaking this ground. Will also be interesting to see how long it takes the Level-D crowd to switch over from their current FSX-on-steriods system. They certainly have access to more computer horsepower than us at home. If anyone can use the fidelity it's them.

 

2 hours ago, brinx said:

viscous forces and the pressure forces

Believe it or not, this is actually the easiest part of the whole process. Once you've got the 3D pressure and velocity fields, these basically fall right out of the process! Just sum up all the contributions on each itty-bitty element face after factoring in the face normal and tangential directions. Of course, you need accurately-modeled boundary layers to get those viscous effects. Boundary layer resolution is where the 3D point cloud count really ramps up. There's not even a snowball's-chance-in-the-desert of doing that correctly with only a few thousand points around the airframe.