DC current and joule heating

DaveStupple

Hi all,

Happy new year.

I am running some simulations of heating a coiled filament. I want to incorporate temperature dependencies for conductivity, both electrical and thermal. As the DC current analysis does not take into account temperature, is it feasible to iterate a few times between DC current and thermal steady state, feeding back in temperatures each time? Or is there a better way? I will try it, but Im not quite at that stage yet.

Victor

If the temperature is fairly uniform, that seems OK, but otherwise it's going to be a nightmare picking out groups of elements of similar temperatures to assign the same material to.

It should be possible to write a Python script to automate that but it'd be pretty complicated too, so might not be worth it.

Victor

On second thought, I forgot that we have temperature dependent electrical conductivity so you wouldn't have to assign materials element by element. Should be pretty simple as long as there aren't any severe nonlinearities that make convergence slow.

DaveStupple

I was thinking of using temperature-dependent electrical conductivity along with temperature-dependent thermal conductivity. After the DC current (maybe with a best-guess uniform elevated temperature) I would feed internal power into the therrmal steady state, then feed the temperatures from the thermal back into a new DC current. I would hope that after a couple of iterations it would converge. I'll let you know what happens.

DaveStupple

Quick update: I improved the mesh as I was getting patchy temperature distributions, so smaller elements with better aspect ratio and made them quadratic. So we have a helical filament with two extended legs. Each leg finishes in a circular 'cut end'. I applied a radiation constraint over the whole surface except the cut ends (so much easier to select with quadratic elements!); emissivity was 0.15 from Matweb - blind faith, I'm afraid. To give some kind of heat sinking from the mounts, I applied to the cut ends a convection of 10^4 W/(m^2.K) - wild guess, roughly equivalent to forced water cooling over this small area. All elements were given temperature-dependent electrical conductivity and thermal conductivity using data harvested from old papers. The electrical one should reduce temperatures by throttling current at high temperatures; the thermal one should increas temperatures by hindering heat transport.

I started with a DC current analysis with a best-guess uniform temperature of 3000K. I then suppressed the temperature and fed the current densities into a steady-state thermal analysis. From this, temperatures are fed back into another DC current... and so on. Seemed to converge quite well. The DC current analyses were very fast but the thermals were getting on for an hour each, so a bit tedious, but I just got on with other stuff. (Stop press! Just tried ccx_static_i8 and it took about 12 minutes with similar results. Lesson learned.)

I'll attach a plot of how the iterations converge. I think there are a couple of dud runs in here where I probably forgot to suppress temperatures or whatever. Wine may have been involved. I'll post some pretty results from Mecway later as I'm away from home without my Space Mouse.

It seems the electrical conductivity triumphed over the thermal as the temperatures were much lower than with no temperature dependence. Pretty pictures later.

Victor

Thanks for the update. Good to hear it worked. Just one question though, are you inventing the light bulb? Because I think there's some prior art in that area.

MikeMcMullen

I am always careful not to post due to potential prior art. Lurk instead so I can pick up bits here and there. May get back to running some simpler FEM problems after a while. Buckling always interested me (especially plastic post buckling behavior) due to its potential for energy absorbtion. Too hard to get a good handle on though.

DaveStupple

Victor, that is quite possibly true. Tungsten filaments are incredibly well studied. I have got hold of a few older papers (and for something like this, the old ones are often the best), but it takes a finite amount of time to read and understand them. I was asked to do this in a hurry before a meeting during the week. It's in part a PR exercise to show we are throwing our resources at a problem. I can tell people I've looked into it and found x, y and z (and I do), but people take more notice when they've got a contour plot backing it up. I take this stuff incredibly seriously and bore everyone to death pointing out the assumptions and shortcomings of the analysis. Two other factors at play here: I like trying new stuff and getting more out of the software, and I don't have any legal means of accessing much of the literature (other than paying for it!).

Victor

Excuse me, I'm normally all serious on here but that inventing the light bulb thing was meant to be a joke :P I assumed you were working on some sciency heater for x-rays or something.

DaveStupple

Don't worry Victor, I didn't take your comments too seriously - if my response gave the opposite impression then apologies! Your assumption is a pretty good one, by the way. We are trying to reverse engineer a pretty old technology - some success but also lots of problems. I can't be any more specific.

With the current analysis, I've confused myself now. I was all ready to do another iterative run using CCX for the thermal steady state (which I thought I had tried and found to be quicker). But it does not like the radiation constraint, so I'm back to using the internal solver. I didn't save the previous CCX attempt - did I manage to omit the radiation, I wonder?

DaveStupple

Pretty picture

Victor

Nice picture.

You can use radiation with CCX in thermal transient. There was some reason I couldn't get it to work right in steady state so I disabled that option.