ELi5 Why are CFD and wind tunnels used together for flight simulation

160 views

I am wondering why there aren’t full simulations using CFD for flight.

In: 4

8 Answers

Anonymous 0 Comments

Because fluids are notoriously difficult to predict.

So difficult that there is an ongoing $1,000,000 bounty for anyone who can make *any* headway in understanding the fundamentals of turbulent fluid flow.

So while we can make some simulations that are very good, and don’t take too much processing power, there is no replacing a physical model.

Anonymous 0 Comments

Ever seen computers with just ludicrous stats? Like what could you possibly be doing that needs 2x 60 core CPUs with 4 TB of RAM?

The answer is Computational Fluid Dynamics, and it’d like more please

Accurate simulations require more points, but more points require more processing power. A balance has to be struck between accuracy of the simulation and having it complete within your lifetime.

Computing power and storage has improved so we are able to do more and more complex simulations, but so far its impossible to simulate on a scale that accurately reflects the real world so we still build small scale test models *just to be sure*

Anonymous 0 Comments

Because all models are wrong, but some models are useful.

Wind tunnels can relatively quickly simulate complex flows with no concerns about mesh refinement, and cover continuously changing sweeps of the vehicle configuration, but the walls of the wind tunnel can interfere with the flow, and the small scale means some details can’t be included.

CFD can model the full-sized vehicle without the effects of surrounding walls, and these days is likely cheaper and easier than a wind tunnel, but modeling uncertainties and mesh refinement makes it difficult to confirm that the solution is correct.

You’ve actually missed one: [some simulators] don’t just use CFD and wind tunnels – full sized flight test is a large and important component as well, and often find slightly different performance than both the wind tunnel and CFD. However it’s the highest cost, especially if it’s a vehicle that requires a pilot.

[edited to make it more clearly relevant to the original question being about simulators]

Anonymous 0 Comments

Because fluids are notoriously difficult to predict.

So difficult that there is an ongoing $1,000,000 bounty for anyone who can make *any* headway in understanding the fundamentals of turbulent fluid flow.

So while we can make some simulations that are very good, and don’t take too much processing power, there is no replacing a physical model.

Anonymous 0 Comments

Ever seen computers with just ludicrous stats? Like what could you possibly be doing that needs 2x 60 core CPUs with 4 TB of RAM?

The answer is Computational Fluid Dynamics, and it’d like more please

Accurate simulations require more points, but more points require more processing power. A balance has to be struck between accuracy of the simulation and having it complete within your lifetime.

Computing power and storage has improved so we are able to do more and more complex simulations, but so far its impossible to simulate on a scale that accurately reflects the real world so we still build small scale test models *just to be sure*

Anonymous 0 Comments

Because all models are wrong, but some models are useful.

Wind tunnels can relatively quickly simulate complex flows with no concerns about mesh refinement, and cover continuously changing sweeps of the vehicle configuration, but the walls of the wind tunnel can interfere with the flow, and the small scale means some details can’t be included.

CFD can model the full-sized vehicle without the effects of surrounding walls, and these days is likely cheaper and easier than a wind tunnel, but modeling uncertainties and mesh refinement makes it difficult to confirm that the solution is correct.

You’ve actually missed one: [some simulators] don’t just use CFD and wind tunnels – full sized flight test is a large and important component as well, and often find slightly different performance than both the wind tunnel and CFD. However it’s the highest cost, especially if it’s a vehicle that requires a pilot.

[edited to make it more clearly relevant to the original question being about simulators]

Anonymous 0 Comments

NASA Ames Research center, which has had massive wind tunnel facilities dating Back to the 50s, was also an early center for computational aerodynamics. They acquired the Illiac 4, a 70s era supercomputer, for that program. But these problems are hugely difficult.(I recall my father claiming that most airplane CFD problems could be handled ok at the second or third order, but that rotorcraft ( like the Osprey) went out to the twelfth order.)

So my point, which I failed to make, is that you can’t, yet, solve these problems computationally. That’s why Ames, in its heyday, had numerous wind tunnels ranging from hypersonic down to highway speed, as well as an active flight test program, AND high-end computers.

Anonymous 0 Comments

NASA Ames Research center, which has had massive wind tunnel facilities dating Back to the 50s, was also an early center for computational aerodynamics. They acquired the Illiac 4, a 70s era supercomputer, for that program. But these problems are hugely difficult.(I recall my father claiming that most airplane CFD problems could be handled ok at the second or third order, but that rotorcraft ( like the Osprey) went out to the twelfth order.)

So my point, which I failed to make, is that you can’t, yet, solve these problems computationally. That’s why Ames, in its heyday, had numerous wind tunnels ranging from hypersonic down to highway speed, as well as an active flight test program, AND high-end computers.