[Joystix]

Hey Guys,

I am taking a project course in applied mechanics, where we simulate a wind turbine blade. The blade is around 63 meters long, put into a 120 degree "cake" computational domain with a rotating part (close to the blade) an stationary part around, the frozen rotor principle I believe it is called.

We have made runs on it which work and converge good, the latest with a "fine" mesh of around 15-16 million elements.

We are now in the verification and validation phase. Validation is done by comparing wind tunnel data from a Linear Cascade Rig and a generic NACA 4412 profile, with a similar simulation setup in ANSYS CFX. The experiments and computational simulation has a good concordance, so that is all good. But what about verification?

The only data we have been able to verify is the power output on the windturbine power plant, with some engineering assumptions of course. The power is the only thing specified by the fictive company that has given us this task aswell.

So, finally, my question is the following: What other things can we measure in order to verify that the simulation results are "good". I was thinking doing some simple lift/drag comparison for a random section of the blade and then do the same using aerodynamic theory. Also, someone suggested looking at the tip vorticies, but I am not sure what to "look for" if doing that.

Anyone care to shed some light on the matter? Would very much appreciate it! Thanks in advance!

[ghorrocks]

In CFD, verification is confirming that the correct physical models have been selected, that the numerical approach to the mathematical model is correct.

Validation is the next step where you confirm your simulation is accurate to reality.

ref: http://www.grc.nasa.gov/WWW/wind/val.../overview.html

So in your case, your verification should involve things like:
* Checking the frozen rotor approach is appropriate
* Checking you have not missed some important physics.

A part of this which you have not mentioned is sensitivity checks of convergence tolerance, time step size (if transient) and most importantly mesh size.

It sounds like you do not have high quality experimental data to compare against for validation, so things to consider are:
* Running a similar simulation where high quality experimental data exists (eg a NACA 4412 profile and the linear cascade)
* Comparing against whatever data you have.

Your comments about comparing to aerodynamic theory results at a position on the blade do not sound like a good approach, the NACA 4412 simulation is a much better comaprision for that; and I cannot see how the tip vorticies can be used in V&V unless you have good quality experimental data on them.

This FAQ lists some comments: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

And the following question on "Are my results publishable" has an excellent link to some very useful information on the topic.

The key textbook on V&V is "Computational Fluid Dynamics" by Roache. Highly recommended if you want to have a deep understanding of CFD accuracy.

[Joystix]

Originally Posted by ghorrocks

In CFD, verification is confirming that the correct physical models have been selected, that the numerical approach to the mathematical model is correct.

Validation is the next step where you confirm your simulation is accurate to reality.

ref: http://www.grc.nasa.gov/WWW/wind/val.../overview.html

So in your case, your verification should involve things like:
* Checking the frozen rotor approach is appropriate
* Checking you have not missed some important physics.

A part of this which you have not mentioned is sensitivity checks of convergence tolerance, time step size (if transient) and most importantly mesh size.

It sounds like you do not have high quality experimental data to compare against for validation, so things to consider are:
* Running a similar simulation where high quality experimental data exists (eg a NACA 4412 profile and the linear cascade)
* Comparing against whatever data you have.

Your comments about comparing to aerodynamic theory results at a position on the blade do not sound like a good approach, the NACA 4412 simulation is a much better comaprision for that; and I cannot see how the tip vorticies can be used in V&V unless you have good quality experimental data on them.

This FAQ lists some comments: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

And the following question on "Are my results publishable" has an excellent link to some very useful information on the topic.

The key textbook on V&V is "Computational Fluid Dynamics" by Roache. Highly recommended if you want to have a deep understanding of CFD accuracy.

Thank you very much for the quick and thorough reply! I am still pretty green on the CFD work-flow, even though I have had many courses on the subject. This project course is probably a lot more like the way things are done in industry, and a not very easy for us to know. So your inputs is greatly appreciated!

We have compared thing like mesh size, iteration schemes and such. Just forgot to mention that in my first post!

If anybody else has some input, please share

[ghorrocks]

If you want to do a really professional job of a mesh sensitivity study, look into grid convergence index and Richardson exrapolation. These are techniques to show rigorously your simulation's level of accuracy.