Initial condition affects coupled and decoupled temperature-displacement models

I found something weird. I don’t understand it yet but I can share the PMX if you would like to try.

The general setup is this:

a long rectangular prismatic has heat input on one of the larger faces and is clamped at some temperature on the opposite face. The initial condition temperature is set to T-20C at the heat input face and the clamp temp is set to T=23C. In other words the top of the rod should expand and the bottom should contract initially making an unhappy parabola/bend. But then, heat input happens from the bottom and the rod is supposed to expand more at the bottom than at the top. I mean you can see the temperature at the bottom is higher than at the top. Yet the bend is still unhappy. However, if you make the initial condition be the entire part or if you increase the heat input, then the part bends in the correct direction.

here is with the initial condition set to all nodes:

Could someone with more experience explain why this happens? I thought the thermal step would be performed first, then the displacement step once the temperature had reached equilibrium. It seems maybe it gets closer to equilibrium if you add more power.

here is the PMX file if you want to play with it. Set the power to 150mW/mm^2 or higher and the strip will bend correctly 50mW/mm^2 and it will bend incorrectly.

temp_bug - Copy.pmx (267.0 KB)

Sorry I had to remove the results to bring the PMX file size down. But yeah, interesting no?

sure enough, if you set the power input to 200mW/mm^2 and run it with a initial time increment of 0.01s and a max of 0.05s you can clearly see that the initial condition is immediate, and that as the bar heats up it changes from that state to something reduced and then to the opposite state/bend. If you apply the initial temp to all nodes, then it just starts from no bend and moves in the correct direction.

at least for low delta T and low power input this feels wrong or counterintuitive because I thought the part would just bend incorrectly and then continue to bend correctly and stop once that final shape was reached but that doesn’t seem to be the case.

What’s the point of applying the initial temperature condition only to the bottom face ? Then the remaining nodes have zero initial temperature.

The same effect can be observed in Abaqus.

1 Like