CFD better than Blade Element for flight simulation?

[brinx]

I'm starting to think CFD is the better approach.

I know XPlane uses Blade Element Theory(BET) for simulating aerodynamics. I always thought this was probably the best approach. I also knew that MSFS chose to use Computational Fluid Dynamics(CFD), and I wondered why they did not choose BET which was the selling point for better flight dynamics in XPlane. I saw an interview with SEB of Asobo where he was asked about it and he said something along the lines of, "there are more than one way to calculate flight dynamics, we chose CFD"(not his exact quote). He also mentioned that Asobo was able to significantly bump up the resolution for the computation being used on surfaces without affecting performance in 2024 due to multithreading improvements.

I recently became curious about the subject again and began looking into CFD vs BET. I've become convinced that CFD is a superior method. From what I see, CFD is considered better for both fixed wing and rotary wing aircraft aerodynamics. The drawback being that CFD is computationally expensive. My theory here is that when Austin chose to use Blade Element, CPUs were slow single core chip in the Megahertz not Gigahertz. It would make sense that he chose BET as it is less computationally demanding even though it is less accurate. Xplane using BET was revolutionary and more flexible than FSX, which was using lookup tables for performance calculations. I suppose this allowed xplane to be used to test rough designs of aircrafts. However, with the more powerful multicore CPUs of today and multithreading, isn't CFD now the better option given it can provide more accurate simulations? And yes, I know you can't do full realtime CFD computation.

If you believe Blade Element theory is still better, why?

The below is the case for both fixed wings and rotary wings:

Quote

When it comes to modeling rotary wings, Computational Fluid Dynamics (CFD) is generally considered the better option. Here's why:

CFD for Rotary Wings

• Pros:

  • Provides highly accurate and detailed simulations of airflow around the rotary wings.

  • Can capture complex flow phenomena like turbulence, wake vortices, and dynamic stall.

  • Useful for optimizing aerodynamic performance and understanding intricate flow behaviors.

• Cons:

  • Computationally intensive and time-consuming.

  • Requires significant computational resources and expertise to set up and interpret.

Blade Element Theory (BET)

• Pros:

  • Less computationally demanding compared to CFD.

  • Good for analyzing basic performance and preliminary design.

• Cons:

  • Less accurate for complex flow conditions and highly three-dimensional effects.

  • May not capture all the intricate details of airflow around the rotary wings.

 

Edited November 29, 2024 by brinx

[Aglos77]

I don't know which is better, I do know that most of the outstanding addons usually have many of their features outside of the simulator.

[Franz007]

I don‘t think one is better than the other, both have pros and cons. So I would expect about similar results with both methods. Although in theory CFD should be better. But just in theory.

I can guarantee you that what is being run on your computer when you are flying with MSFS is not the industrial/professional type of CFD (where the Navier-Stokes equations are being solved). MSFS would have to solve this at every instance for multiple airframe components and then integrate all of the forces in real time. I think MSFS runs something called "panel lattice method" which is a very quick linear calculation. I think X-Plane also runs something similar. I'm glad everyone is excited about the use of CFD in their prefered sim but I feel like it is not being explained very well by Asobo and it's mostly hype/marketing. Not trying to burst anyone's bubble as i myself would love to see technology get to point where we can simulate that in real time but it takes 2-3 hours on 250+ solver cores to run a true 3d static (single point) solution.

Edited November 29, 2024 by Franz007

[mSparks]

Blade element theory is the name of the theory used to calculate the forces on a surface moving in a fluid.

Generally involving calculating reynolds numbers

https://en.wikipedia.org/wiki/Reynolds_number

Hence the name "Laminar Research"

Computation Fluid Dynamics is what you call any simulation that calculates the forces involved in a moving fluid.

So imho the question doesn't really make sense, sounds to me like asking if astrophysics is better than the big bang theory for space travel.

Edited November 29, 2024 by mSparks

[Bjoern]

If BET delivers plausible, reproducible results on a code base that somebody can actually understand, it hands down wins over CFD, to which these aspects do not apply. Only like twenty people on the planet truly understand CFD (the rest just pretend and do not) to know all the caveats, pitfalls and limitations and I would be extremely surprised if Microsoft managed to hire one of them and even more so if that person managed to reduce computational demand from a server cluster to a gaming console.

[Murmur]

3 hours ago, brinx said:

If you believe Blade Element theory is still better, why?

The below is the case for both fixed wings and rotary wings:

That's a chatGPT response, isn't it?

Under the umbrella term "CFD" are included a very extensive range of calculation techniques, from the more simplified ones, to the more accurate (and unfeasible in real time) ones.

Needless to say, ChatGPT does not know the specifics of the CFD and Blade Element Theory implementations in either sim. One is not necessarily better or worse than the other, it mostly depends on how they have been specifically implemented in each sim. There could be "good" CFD/BET and "bad" CFD/BET.

TL;DR: ChatGPT is not a reliable source on such specific matters like these.

Edited November 29, 2024 by Murmur

[Murmur]

2 hours ago, Franz007 said:

I think MSFS runs something called "panel lattice method" which is a very quick linear calculation.

I'm pretty sure MSFS2020 wasn't running such a sophisticated method like the panel or vortex lattice. It was running a sort of hybrid method that actually also used the old legacy table-based aerodynamic coefficients of FSX.

Don't know about MSFS2024 though.

[brinx]

3 hours ago, mSparks said:

So imho the question doesn't really make sense, sounds to me like asking if astrophysics is better than the big bang theory for space travel.

I'm by no means an expert on CFD vs BET, I only recently started looking into it. From what I've learned thus far, they can both be used to predict flight characteristics of aircrafts. They take very different approaches though. CFD seems to be a more advanced and accurate method of calculating how the object(e.g, a plane) behaves in the fluid(air). 

"In aerospace engineering, CFD is used to predict airflow over an aircraft, or the forces acting on it. This allows engineers to design more efficient and quieter aircraft, as well as improve the safety of flight. CFD simulations can be used to study a wide range of problems, including lift, drag, and stability."

CFD appears to be better at capturing "complex flow phenomena like turbulence, wake vortices, and dynamic stall"

The application of CFD to aircraft design and development is well established and its use is growing in both the aerospace industry and academia. The benefits of using CFD are many and varied

https://www.ascend-tech.com/blog/improving-aircraft-aerodynamics-with-cfd-simulation

 

Edited November 29, 2024 by brinx

[Franz007]

@brinx Yes in aerospace engineering, not within a desktop-sim in realtime. That‘s why it doesn‘t apply for desktop sims and may not be better than BET at the end, when we are talking about a realtime simulation on a desktop PC.

[mSparks]

35 minutes ago, brinx said:

I'm by no means an expert on CFD vs BET, I only recently started looking into it. From what I've learned thus, they can both be used to predict flight characteristics of aircrafts. They take very different approaches though. CFD seems to be a more advanced and accurate method of calculating how the object(e.g, a plane) behaves in the fluid(air). 

"In aerospace engineering, CFD is used to predict airflow over an aircraft, or the forces acting on it. This allows engineers to design more efficient and quieter aircraft, as well as improve the safety of flight. CFD simulations can be used to study a wide range of problems, including lift, drag, and stability."

CFD appears to be better at capturing "complex flow phenomena like turbulence, wake vortices, and dynamic stall"

https://www.ascend-tech.com/blog/improving-aircraft-aerodynamics-with-cfd-simulation

 

Off the shelf CFD software generally doesnt care about computation time - doesn't matter if it takes 2 days to compute a few seconds of fluid flow, thats still cheaper than spending $100,000 building a real one and all you need. BET and other theories are still part of that, but generally you want millimeter accuracy - the effect of every crease and corner and curve is important.

1 hour ago, Murmur said:

Don't know about MSFS2024 though

Whats funny, is although it did kinda occur to me, I didn't properly realise this was just an msfs v Xplane comparison thread in a different guise. I care so little about what they are doing these days I didn't even think to think about then.

1 hour ago, Murmur said:

There could be "good" CFD/BET and "bad" CFD/BET.

I've now had 3 different VR flight sim experiences. 

DCS, Xplane and most recently warplanes for the meta quest 3.

The worst of those in terms of simulation is hands down warplanes, its also the one Ive currently been playing more because its so easy to access, especially when travelling. Doesnt replace xplane tho, which is bordering on perfect.

The "hard part" for Laminar Research has been getting the CFD simple enough that it runs in real time, but not so simple that key behaviors fall over.

Thats taken Austin Mayer and his team 2 decades and more working/researching in a significant commercial setting (that requires it to work the same or better than perf tables) to get to where we are today, no one seems to know much about what MS is doing with CFD to try to catch up with that, 100% agree its the right direction for them to take, but tbh I'm not going to care unless they bring it to the quest 3 before Laminar.

Edited November 29, 2024 by mSparks

[brinx]

4 hours ago, Bjoern said:

Only like twenty people on the planet truly understand CFD (the rest just pretend and do not) to know all the caveats, pitfalls and limitations and I would be extremely surprised if Microsoft managed to hire one of them and even more so if that person managed to reduce computational demand from a server cluster to a gaming console.

Where did you get that ridiculous number? With all the software out their capable of CFD simulation, you believe only 20 people understand the subject? In that case, the entire world should unite to protect these 20 people at all costs.

58 minutes ago, Franz007 said:

Yes in aerospace engineering, not within a desktop-sim in realtime. That‘s why it doesn‘t apply for desktop sims and may not be better than BET at the end, when we are talking about a realtime simulation on a desktop PC.

@Franz007 I know it is not possible to do a full-on simulation in real time currently.  However, I now believe that our PCs are good enough to do a version of CFD that is accurate enough for the purpose of home simulation. I understand that NASA would need to be able to compute flight characteristics that are not possible in real-time. For the purpose of home flight simulation, we want to get as close as possible within the limits of the hardware that we have today.

I was just wondering if anyone could make a compelling argument as to why BET is still better than CFD today. I have not seen one. I think BET was probably the best option to use when we had slow single core processors, before fast multi-core processors became commonplace.  Today, if I were building a new simulator, I think I would choose to implement CFD based on what I'm reading. 

Also, not saying BET is bad, I just think that CFD is better.

Edited November 30, 2024 by brinx

[Murmur]

21 minutes ago, brinx said:

I was just wondering if anyone could make a compelling argument as to why BET is still better than CFD today. I have not seen one. I think BET was probably the best option to use when we had slow single core processors, before fast multi-core processors became commonplace.  Today, if I were building a new simulator, I think I would choose to implement CFD based on what I'm reading. 

Also, not saying BET is bad, I just think that CFD is better.

You do not know the details of how MSFS has implemented the so-called CFD, nor the details of how XP has implemented BET, so you can't judge whether one implementation is better than the other. The only judgment could be the final results possible in either platform in terms of flight model.

Also, do not conflate the generic term "CFD" (which, as I explained, encompasses a lot of different techniques in accuracy and complexity and AFAIK none of which is currently used for real-time applications in the engineering field) with a specific implementation in a real-time flight simulator.

Edited November 30, 2024 by Murmur

[blingthinger]

7 hours ago, Murmur said:

That's a chatGPT response

Sheesh. I'm still not used to AI being a source for this stuff.

Didn't you once say that Flight Unlimited used a CFD model yrs ago?

 

6 hours ago, brinx said:

they can both be used to predict flight characteristics of aircrafts.

Depends on what you want out of it and what resources you have access to.

CFD is a virtual wind tunnel. It demands fully accurate knowledge of the geometry and a very high density cloud of points in the 3D space around it. Think on the order of 10s of millions of points for a Cessna at the lowest fidelity and slowest speed. A 747 at cruise moves up to 100s of millions of points. High-resolution studies use billions of points.

Asobo's 2020 CFD model defaulted to 8000 points around the boxy "balsa-glider" geometry. 2024 maybe triples that but not much more, with balsa-glider-like wings and a curvy-permitted fuselage. 

So here's what you'd do IRL: make a very high resolution 3D model and run it on your supercomputer at all sorts of angles and velocities. At each operating point you'd sum up the forces at all the points on the airframe surface and create one list (table) each for lift, drag, and moments at all those operating points. You'd then put those tables into your FSX/P3D/2020/2024 model. Why the latter 2 in that list even with CFD? Because the CFD result are still totally out-to-lunch. The geometry is wrong and the point cloud is very very coarse. You still need a realistic anchor point. That anchor is either legit CFD, wind tunnel, or flight test data. Or a blended mixture of all 3. This is why FSX is still at the core of 2024.

Another route is to slice up the many CFD solutions sitting on your supercomputer into, say... 5-10 slices per wing/lifting surface and sum up the forces along that slice. You'd create one force/moment table for each slice. You could also sum all points on the fuselage and create a single table for that as well. Same for any engine nacelles, etc. These tables would get fed into your XP BET model.

Both flavors of table model (single-point or multi-point (BET) ) are very much alive and well in the "enginerding" world. If you had a good-enough set of tables in your BET, you could also use that as the anchor for 2020/4's CFD.

We're at least 10-15 yrs away from a legit real-time CFD model and it will run on the GPU when that does happen. BET is orders of magnitude faster and less resource intensive than CFD and for the next decade or 2 offers more real-time fidelity than CFD can.

Edited November 30, 2024 by blingthinger

[Aglos77]

35 minutes ago, blingthinger said:

Sheesh. I'm still not used to AI being a source for this stuff.

Didn't you once say that Flight Unlimited used a CFD model yrs ago?

 

Depends on what you want out of it and what resources you have access to.

CFD is a virtual wind tunnel. It demands fully accurate knowledge of the geometry and a very high density cloud of points in the 3D space around it. Think on the order of 10s of millions of points for a Cessna at the lowest fidelity and slowest speed. A 747 at cruise moves up to 100s of millions of points. High-resolution studies use billions of points.

Asobo's 2020 CFD model defaulted to 8000 points around the boxy "balsa-glider" geometry. 2024 maybe triples that but not much more, with balsa-glider-like wings and a curvy-permitted fuselage. 

So here's what you'd do IRL: make a very high resolution 3D model and run it on your supercomputer at all sorts of angles and velocities. At each operating point you'd sum up the forces at all the points on the airframe surface and create one list (table) each for lift, drag, and moments at all those operating points. You'd then put those tables into your FSX/P3D/2020/2024 model. Why the latter 2 in that list even with CFD? Because the CFD result are still totally out-to-lunch. The geometry is wrong and the point cloud is very very coarse. You still need a realistic anchor point. That anchor is either legit CFD, wind tunnel, or flight test data. Or a blended mixture of all 3. This is why FSX is still at the core of 2024.

Another route is to slice up the many CFD solutions sitting on your supercomputer into, say... 5-10 slices per wing/lifting surface and sum up the forces along that slice. You'd create one force/moment table for each slice. You could also sum all points on the fuselage and create a single table for that as well. Same for any engine nacelles, etc. These tables would get fed into your XP BET model.

Both flavors of table model (single-point or multi-point (BET) ) are very much alive and well in the "enginerding" world. If you had a good-enough set of tables in your BET, you could also use that as the anchor for 2020/4's CFD.

We're at least 10-15 yrs away from a legit real-time CFD model and it will run on the GPU when that does happen. BET is orders of magnitude faster and less resource intensive than CFD and for the next decade or 2 offers more real-time fidelity than CFD can.

We will not see half of here if Microsoft does not abandon the product first 😉

[GoranM]

Put it this way.  Austin and the staff at Laminar could change their model to take advantage of CFD.  There's a reason they won't.  And it's not because they can't.  

Here's why.

1 hour ago, blingthinger said:

BET is orders of magnitude faster and less resource intensive than CFD and for the next decade or 2 offers more real-time fidelity than CFD can.