"Explotion" inside a metal box

[maguirre]

Hi, Elmer community!
I've been using Elmer for a year or less and it seems to be a quite competitive tool for FEM simulations. Until now I've just used ("correctly") the Heat and Stat Current Solvers for a metal bar which cools through natural convection (as a boundary condition) and is heated with the Joule Effect with a voltage source, again as boundary conditions. This is one of the type test applied to HV/MV switchgear.

Now I want to simulate something more difficult in terms of equations. There is a type test for metal-enclosed HV/MV switchgear and controlgear called arc fault test. In this test, an electric arc is produced with a high current applied between two conductors. This arc of about 1.8 MJ of heat (and may reach each 10 kK) heats the surrounding air of the enclosed air in 20-30 ms, creating a big temperature gradient, and increasing its pressure and velocity. An example of this test can be seen here https://www.youtube.com/watch?v=FFRgNHZWstY. From this description, I assumed that the Navier-Stokes and Heat Solvers are needed.

First I tried to simulate the air with the incompressible model (Mach number approx. 0.5 for air at 700 K, the mean temperature not near the arc) with a mesh and time step small enough to recreate the turbulence (Re approx. 10^6 at the same temperature as before). The arc heat produced is introduced as a body force in part of the confined air. As for the buoyancy, I used the Bussinesq approximation. With this model, I could simulate the air heated by the hot zone (https://drive.google.com/file/d/1L-XbjT ... drive_link) and its movement. But the velocities and pressure were much less than the ones observed in experiments (velocities magnitude and pressure 1 or 2 orders below experimental ones) (Velocity: https://drive.google.com/file/d/1A4Sh5P ... drive_link)

I thought that the Bussinesq approach for such big gradients of temperature (and therefore density) was a bit naive because of its explosive nature. Now I'm trying with a volumetric force 2 as -9.81 and the perfect gas model. I also created a "wall" as a solid and not just a boundary condition and increased the size of the hot zone for a better convergence as I read somewhere. But here is the thing, those changes did not modify the air surrounding the hot zone, I expected a rapid temperature increase but nothing happened. I share the mesh and sif files for you to check.

All solid external boundaries are set as conduction to ambient temperature with its heat cond. coeff., all internal as no-slip, and the floor as adiabatic. The boundary between the exterior and the interior gas is set with ambient pressure. Initial pressure and temperature are set as ambient, and 0 velocity for the air.

At this point (a few weeks of trial and error) I don't know if my physics, my practice with ELMER, or both are wrong. Any help would be appreciated. Please, feel free to ask anything vague or badly explained.

-Matias.

[kevinarden]

Your flow appears to be very turbulent which can be a challenge in Elmer.

viewtopic.php?t=6995&hilit=turbulent+flow

[maguirre]

Hello Kevin! Indeed, I used the k-epsilon viscosity model when I simulated with the Boussinesq approach, and it certainly helped.