Tutorial 16 Flow through a hole (in MEMS microperforation)

[kevinarden]

I am thinking that the pressure builds high at the boundaries where there is little flow to relieve it, and the pressure is the lowest where the most flow occurs, because the flow relieves the pressure. I will check my results for pressure to see if it matches.

[JJSz]

Thank you. I have also checked everything again and I have discovered that in my .grd file scaling did not work properly. Aparently keyword "scaling" at the end of file did not enable it. So from the beginning all calculations have being made for a hole model which had bathroom dimensions...
I know...
I have fixed both .grd file and .sif file for time-variant simulation. At least it seems so... I have one ongoing simulation at this moment. Still have 25 minutes to go. I will send results and files after complementation.
I really appreciate how much you are helping me. I know that I am asking for much and I waste your time. I apologise for this. All I can say in my defense is that I am still learning this environment functionality.
Now please excuse me... I have to do some running. This research really got on my nerves XD

[JJSz]

All right, results are similar to these that you send me recently. So i still have the same trouble with pressure distribution. Also I have tried to add saving some other scalars to be able to calculate impedance of the hole. Tried to do it according to this part of Elmer Models Manual.

example.png (48.77 KiB) Viewed 3113 times

I had to significantly decrease amplitude of parabolic velocity function at the inlet. I am wondering if v = 10^(-6) is anyhow reliable but it is the only value that worked for me.
I am attaching new .grd and .sif:

_holeBestMini.grd

(1.3 KiB) Downloaded 174 times

case_slip_time.sif

(6.17 KiB) Downloaded 171 times

Could you please take look at them? Do values that are being returned in .dat file look well for you?

[kevinarden]

I will check tomorrow morning (EST)

[kevinarden]

I based the parabolic equation on the original size of the model so the equation should be adjusted for the new size. However the new model is very small. Originally the base was 4x5 meters, now it is 1.E-6 meters. What is the intended size of the base?

[JJSz]

Hi,
dimensions should remain the same, just scaled to be micro (thus 5 x 4 x 9 micrometers). I have already changed velocity equation to have 0 value at points where x = 0 or x = 5 * 10^(-6) [m] and y = 0 or y = 4 * 10^(-6) [m]. But the amplitude is equal to 1*10*(-6) at the center of the boundary 9. I have doubts as for this value. But only with that small velocity I could again reach convergence in time.

[kevinarden]

What is the desired velocity to reach?

[JJSz]

Small enough to fulfill conditions for Stokes equation.

requirements.png (87.59 KiB) Viewed 2672 times

As for the minimum value... I do not quite know myself. I am still looking for information about that. Unfortunately, this is only one part of MEMS acoustic sensor. I cannot base this value on desired value of sensitivity or SNR because I do not have the rest of this complex structure.
Until now I have refered BCs to several documents including this, where research is conducted with assumption of v=|0.025| [m/s] and no-slip on walls
https://asa.scitation.org/doi/10.1121/1.5098785

requirements2.png (24.53 KiB) Viewed 2672 times

In other work it was already higher: v = |1| [m/s] but with present slip coefficient equal to 1
https://hal.archives-ouvertes.fr/hal-00189253/document

requirements3.png (185.02 KiB) Viewed 2672 times

I have tried all of them previously, but now after doing proper scaling they seem too large

[raback]

Hi

What is the Reynolds number for your problem?

Was the failing to converge the reason for introducing turbulence? I could see that there is some turbulence around the corners. However, these old results of Veijola certainly were computed without turbulence as the Reynolds number was deemed to be rather small.

If the Reynolds number is small that will also effectively make system steady state. Of course the enforcing may be transient but the flow resistance will be instantaneously reflecting the forcing. So if you define impedance for the system it will be real valued. Having imcompressibility effects would introduce history effects which would introduce complex valued impedance (force and velocity in different phase). Compressibility effects are also rather improbable in MEMS.

So in the end if Stokes equation suffices you should only need steady state equation + setting for FlowSolver:

Code: Select all

  Flow Model = string "stokes"

also you should only need one nonlinear iteration.

The test case "FlowResSlip" basically has these strategies.

-Peter

[kevinarden]

https://github.com/ElmerCSC/elmerfem/tr ... lowResSlip

I believe the original model Reynolds number was high due to the dimensions of the part (several meters), now that it has changed to micrometers it is likely different. The link to the mentioned test case is above.