Hi. I am about to generate a CFD modell for simulating the flow behaviour arround a microstructure in a flowfield. this microsructure "sits" on a surface, a flat plate, say. The first microstructure I am about to moddel is basicly a hot wire, with a typical diameter varying from 1 to 5 micrometres. This wire is placed over a cavity in the surface, typical 10*30 micrometres. The aim is to use this to measure the shearstress at the surface. But before any model is to be built a CFD moddeling is to be performed to analyse the disturbance in the flow caused by the structure and the heat-transfre from the hot-wire to the flow. There are several problems (as far as I can see)assosiated with the creation of a mesh, boundary conditions, interpretation and validation of the results. 1) Creation of a mesh arround a 1 to 5 micrometres wire recuires VERY small cell volumes. How will the small cell volumes influence the accuracy of the solution. Will this introduce large numerical errors because of the very small fluxes in each cell, thinking of the limited number of digits stored. 2) No law of the wall functions can be used, so a viscous layer must be built. This is known to be difficult in 3D. A viscous layer consists normally of VERY thin and long cells. But colse to the structure this can not be the case as much shorter cells are required to capture the gradients assosiated with the wire and cavity. How will this influence the results? 3) Boundary conditions. How large should the flow domain be to be able to specify valid boundary conditions? This is an important issue as a very fine mesh is required and a large domain results in a VERY high number of cells. Is there any way this could be avoided? 4) OK, so we have solution. BUT how can we validate it? Flow around a microstructure can not be investigated experimentally. So large scale experiments are required. But this is not what was simulated using CFD. So the results can not be validated. Should the CFD modell also be scaled up?

Any comments resived with thanks.

Sigve

Hi, Sigve

Properly non-dimensionalise your problem. Is is easy. If you do, many of your questions will be answred easily. If you don't, you will fail entirely anyway.

Never say 'very small'. It does not make sense. Say, 'much smaller than... '. Non-dimensionalise.

I am a bit jumpy today, hence the form, sorry, but believe me, this advice is the best you can get at this stage.

For example, whether your problem can be solved or not at the current level of computers depends solely on the value of a non-dimensional number, so called Reynolds number, for your problem, and on nothing else. But this is a long story. Calculate this number, first. Non-dimensionalise.

Yours Sergei

Hi Sigve,

You need to be careful when using CFD to model gas flow over and around structures which are in the micron range or less. When sizes are that small, the assumption of no slip at boundary surfaces becomes less valid since the mean free path of a gas molecule becomes comparable to the dimensions of the flow obstruction. The usual check on whether continuum flow assumptions are valid is to look at the Knudsen number (which is just the ratio of the mean free path of the gas molecule to the cylinder diameter). If the Knudsen number is above 0.1, then you usually start to think about slip flow corrections. You can find more info on slip-flow calculations and experimental measurements for various heated wires (hot-wire anemometer)at: http://naca.larc.nasa.gov/reports/19...ca-tn-4369.pdf. Since most CFD programs utilize normal Navier-Stokes/continuum mechanics approaches, slip flow behavior would be missed. This usually shows up in a higher calculated drag or pressure drop than is the case experimentally.

"Over 1000 measured convective heat-transfer coefficients for normal cylinders in subsonic slip flow have been correlated by using Nusselt number as a function of Reynolds and Knudsen (or Mach) numbers. The experimental range corresponds to the following dimensionless groups: Mach number, M, 0.05 to 0.80; Reynolds number Re, 1 to 75; Knudsen number Kn, 0.009 to 0.077. Air temperatures between 0 and 280 degrees F and cylinder temperatures between 34 and 620 degrees F were used."

Thanks, Phil.

I knew there was something that I should be aware of, but I could not put my finger on it. Because using non-dimensional parameters is in most cases the best way to analyse and compare flow behaviour. But when the mechanics deciding the flow behaviour changes the this can be lead you on to comparing two cases witch is quite different. So more information is needed. So the Nusselt and Knudsen number will give me this additional information.

Thanks again.

Sigve