Discussion Forum

hi, guys

first of all, i am a junior at FEM and sorry for my broken English

As in the title, is there any technic or evidence such as solver log, result data, which can evaluate how accurate my result is againt reality? My model is 3D - current meter which is applied joule heating, heat transfer and laminar flow. It works on measuring flow velocity by calculating temperature diffence of resistor for specific time span. Given simulation result and real experiment of prototype, temperature increases very fast from the beginning(heating) for ~0.5s, then slows down and saturates.

Anyway, for single electrode sensor(one resistor), simulation result is quite similar to reality. And now my current meter model is using 3-electrodes which measures temperature difference between two electrodes for the same purpose, flow velocity. For this 3-electrode model, there was several failure during sovling and warnings. By meshing this model very carefully, I eliminated all the warnings and got result. However, temperature, as well as current density, fluctuated with a peak at the first and then saturated. After refining mesh or appling smaller tolerence, I could get smoother result. converged, But still has a little fluctuation. I do not know now whether I am going to right destination or not. So, how do i know how close to reality my result is? without real experiment.

And I have another question. When scaling dependant variable, which number shold I take for p(pressure) in laminar flow? For larminar flow in cilynderical tube, I use twice as much as mean velocity fou u(max. velocity in tube) and p is selected by defult. It works good, I think, reducing computing time and sometimes removing warings. But, sometimes carrying not good result than even no-scale-solving. In this situation, what solud I consider for p?

Thank you guys for reading long long question and thank you in advance in case of my late response good day!