CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   current validation technologies for cfd (

allan morrison July 20, 1999 17:19

current validation technologies for cfd
From reading posts from cfd-online there seems to be some problems comfirming actual field measurements with what the cfd software programs say should be there. I would like to see some discussion on very inexpensive measurement techniques - especally those based on very fast measurements using state of art electronic sensors, hot wire, lasers etc. What can be done today that couldn't be done twenty years age even at 100 times the cost? Comments please. Tyia.

John C. Chien July 20, 1999 17:59

Re: current validation technologies for cfd
(1). I think the problem is not within the professional CFD researchers, but rather among the CFD users of commercial CFD codes. (2). This is quite similar to the recent tragedy of flying over the ocean in a dark night without adequate training, experience and license. (3). I would say, if the user knows how to take good measurement of the flow field and knows how to interpret the result, he is also likely to get good result from a commercial CFD code. (4). It is possible to get good and useful results from commercial CFD codes, but it requires experience. My feeling is that most problems discussed here are experience related. One really has to obtain reliable mesh independent CFD solution first before he can compare the results with test data. ( we are not there yet to get 3-D mesh independent solutions!)

clifford bradford July 21, 1999 14:33

Re: current validation technologies for cfd
two of the favorite techniques now that were not available 20 yrs ago are PIV and LDV. i'm not an experimental expert but the results are usually of the quality to be able to validate fine scale cfd results. there is also a technique which does not require seeding the flow with light scattering particles but the name escapes me right know. all these methods have the drawback that they aren't usually able to take readings near a solid surface so the boundary layer cannot be measured directly. but they allow line or sheet measurements (as compared to point measurements using probes) and can make unsteady measurements. another advantage is that the results are of the same format (numerical field data) as cfd results so they can be incorporated easily with cfd results if a mixed experimental/computational design is being done. there are other good techniques for obtaining fluid dynamic results like pressure sensitive paint and (i think) liquid crystal heat transfer measuring coatings (which allows some validation of boundary layer effects)

Patrick Godon July 21, 1999 14:36

Re: current validation technologies for cfd
Tyia, Like John said,

The problem is not with the experiments, but rather with the ability to interpret them using numerical simulations or theoretical analysis. Usually the best way to make sure one understands a flow problem is to make use of (e.g.) 'wind tunnels', etc... So the semi-empirical approach has been very succesfull. 3D time-dependent simulations of flows around objects, transition to turbulence, etc.., are still badly needed. While experiments seems in many cases to be already at hand to explain many physical process in flows. See for example:

Streamwise vortices and transition to turbulence, Hamilton and Abernathy, 1994, J. Fluid Mech., 264, 185.

Daviaud, Hegseth and Berge, 1992, Physical Review Letters, 69, 2511 (subcritical transition to turbulence in plane Couetter flow).

These are examples where the experiments is well in advance of the theory and simulations.

Usually a code will give you good results for a given problem and may give bad results for another problem. In other words each code can solve the NS eqs. with some IC and BC for a given region of the parameter space (where the parameter space is defined e.g. by the Reynolds number, and other free parameters). The code has to be implemented and adapted to solve a given problem. In addition each different numerical method (e.g. Finite Difference, PPM, SPH, Spectral Method) performs better for a given kind of flow (smooth flow, turbulence, shocks, ..). The titles of some books speak for themselves: Numerical Recipes; The Art of Scientific Computing; etc.. To simulate an airfoil (NACA 0012 or whatever) is an art in itself and it takes years to become an artist and to find the right approach and the right hydro code.

The best example is the treament of the boundary conditions: it can be done in many different ways and can lead (accordingly) to many different results (good and bad) for exactly the same problem. So to implement boundary conditions is an art in itself and many trials have to be carried out in order to find the appropriate manner to impose them (on the primitive variables, on their derivatives, on the characteristics of the flow?.. etc..).

All times are GMT -4. The time now is 08:50.