CFD is still not enough?
 

Hi,

I have been watching some documentries and read some articles about things concerning aero and was curious as to why people still use wind tunnels and the smoke wands on scale models.

What is lacking in CFD which makes these physical tests still required? Or are these physical tests still needed as validation for the CFD analysis?

The one that comes to my mind immediately was the one of the boeing's blended wing project: http://www.flightglobal.com/Articles...nded+wing.html

They have built a scale model so that they can test the control systems and other aspects of the plane.

I have done hardware in the loop simulation myself before, so my thinking is that it is purely validation of the results, but I would like to know for certain if and what limitation still exist with CFD.

I am an engineer, but I don't have fluids background, fluids are just an interest of mine.

Thanks

Re: CFD is still not enough?
 

Interesting points, PI Guy.

As I see things at present, we have a reasonable handle on incompressible, very-viscous, very-slow flws & on the other extreme - compresible inviscid high-speed flows, over into hypersonics etc.

The whole batch of flows in between has _no_ solid theoretical foundation & resorts to all kinds of hocus-pocus to model the fluid flows. Lots of questions about accuracy eg. lift, drag. All kinds of numerical kludges, convection-stabilisation, x-times timestepping & wierd-&-wonderful turbulence-models have to be used to produce a solution. Whether the solution produced is vaguely close to reality - most folks would not dare to say they are comfortable.

Would you say that, under this scenarion, that it's safe to throw away the classical windtunnels & resort to purely numeric methods? :)

diaw...

Re: CFD is still not enough?
 

Mainly because we still have to use some simplification in order to be able to solve the equations describing the particular flow situation (i.e. the equations we solve are not exactly those governing the flow of our interest), because we still have to choose a trade-off between accuracy and runtime, because we often have to simplify the geometry (of the flow or of the duct), because.... Actually, sometimes we need an experiment even before a computer simulation is run for very first time - otherwise we'd have to guess a coeff required...

Re: CFD is still not enough?
 

The requirements in aeroacoustics are drastically different from the field of CFD.A number of reviews have been published on this topic.Check these papers when you find time.

Computational Aeroacoustics: A Review SK Lele - AIAA Paper, 1997

Computational aeroacoustics- A review CKW Tam - AIAA Journal, 1995

Re: CFD is still not enough?
 

Part of the answer may be in this question :

How many configurations (flaps, wheels, incidence...) could you measured within a CFD run of 8 hours and within a wind tunnel session of the same time ?

CFD will give you a detail picture of a very particular configuration (something like IRM or X-ray analysis), at the opposite, wind tunnel will give you a most global overview but for a lot of differente configurations.

Re: CFD is still not enough?
 

Modern measurment techniques like PIV or preassure sensitive paint can provide a lot more details than the usual pressure taps.

CFD sometimes plainly gives the wrong results. You might run a case with four different turbulence models and get four different results.

In the other hand, there are certain cases that are either hideously expensive or impossible to test in a windtunnel. Try to get the space shuttle in lift-off configuration into a wind tunnel ...

You could, for example, first have model tests. Than try to rebuild them by CFD (i.e. calibrate your settings and coefficients to be able to reproduce the test results). Finally you use CFD to compute the real flight conditions - and then you cross your fingers ...

Re: CFD is still not enough?
 

If you look at the procesures that solve fluid dynamics equation, there are truncation error (approximating PDE using Algebraic equations), and round off error (computer error). These two errors will always exist.

If you look at fluid dynamics equations, there are turbulent equations always based on experiments. The constants in turbulent equations are purely empirical. This is also the reason that for different application different turbulent equations applied.

Re: CFD is still not enough?
 

It is a good question, but to my mind the answer is quite simple ... fluid dynamic problems can be very challenging.

I think it is completely the wrong mind-set to set CFD, experiments and theoretical fluid mechanics in opposition to one another, they are all just TOOLS and should all be used by the experienced Fluid Dynamicist.

The easiest analogy is that a Fluid Dynamicist is like a carpenter and CFD is like a hammer, experiments like a screwdriver and theoretical methods (asymptotic techniques, decompositions etc ...) are like a chisel ... all techniques are appropiate under different conditions. These conditions are dictated not only by the flow physics, but by the expense of the technique, the level of understanding required of the phenomena, and the time constraints (and I am sure there are more to be added to this list).

People always go on about how bad a result can be obtained through CFD, but it is also possible to get a terrible result through an experiment. What if an unsteady signal is filtered in the wrong way, a pressure probe isn't aligned properly, a pressure tapping blocked, a calibration that is no longer valid is being used ... the quality of the result obtained will ALWAYS be dependent on the quality of the person doing the work, not the tool that he/she is using.

An interesting idea that I heard elsewhere is ... "The only person that believes a CFD result is the person doing the CFD, everyone believes an experimental result, except the person doing the experiment"

Re: CFD is still not enough?
 

Amen!

Re: CFD is still not enough?
 

As you can see from the response, there are many reasons why CFD is "still not enough". I agree with 'Anonymous' and many others who say that it never will be enough, simply because it's not supposed to be. Numerical CFD is a tool that >complements< other tools (analytical, experimental), rather than fully >replacing< them!

If I had to pick one major reason why CFD will never be enough: Even if you could solve the flow equations exactly (Ha ha ...) you will not likely run CFD on the exact design geometry. In initial design steps the geometry is not defined in all details, and even in the final design, the actual machine (airplane, turbo engine, other devices), is far too complex to model in all details, including secondary flow through cavities, control mechanisms, and so on. You'll work your way to the final design step by step, making use of all your tools, getting the best out of analytical, numerical, and experimental methods, for best design efficiency and accuracy.

Re: CFD is still not enough?
 

Forget to mention that also you will never get rea...l boundary condition for CFD simulations.

Also at wall boundary, specially turbulent boundary, we are still using approximate for both momentum and heat transfer.

Re: CFD is still not enough?
 

Even though CFD is not always accurate. But it is a very good tool for preliminary design, and for many cases, it is more accurate than a prediction based on empirical equations. It also shortens the design period by reducing the number of experiments.

For many cases, the current tech can't make it true to study the flow experimentally, such as miniature designs and high speed rotating flows. There is no way to study miniature high speed rotating flow. So there is no way to find design defect based only on the pressure ratio. One has to reply on CFD simulation to identify possible problems.

CFD is very helpful based on how you use it.

Re: CFD is still not enough?
 

I agree that CFD is just another tool. But some thoughts on:

1. You do not get real BCs in CFD In fact you may be more likely to get "real" BCs for some problems from the CFD than from the experiment. For example a road vehicle has a BL on its lower surface. There is no BL on the actual road. Even the most elaborate wind tunnel tweaks (moving belts, BL suction etc) cannot completely remove the effects of the wind tunnels BL. Thus the CFD solution is more "real" than the wind tunnel when looking at that particularly part of the physics.

2. The complexity of geometry favors test models. This is not so. Again using the example of a road vehicle. The CFD can simulate rotating tires and brake rotors. It can simulate detailed flow through the engine compartment. It can even sink mass at the engine intake and source mass at the tail pipe (though those are unlikely to change the solution much). Adding this level of geometric fidelity to a physical model is very rarely done, yet is a routine practice among automotive CFD analysts

I recently was involved with some CFD on an aircraft which was tested in a wind tunnel with no flow passage through the engine. This blockage resulted in an unphysical pitching moment that CFD was used to "calibrate" out of the test solution. Now, it certainly would be impractical to try and model engine flow and thrust in the wind tunnel model. And the Wind tunnel model generated more data points in a week than the CFD could in several months. Neither tool would have been much good without the other.

My to cents -Andy R