Thermal Simulation Setup Confusion

[Anonymous]

Hi, I'm trying to set up a basic thermal simulation to give me a temperature gradient across an assembly I've created.  Ultimately I want to use the thermal stress simulation tool, but I thought I'd try and understand this first. 

I was hoping for some advice as the model doesn't behave as I expected with temperatures in excess and less than my initial conditions.  My model is essentially a thin rectangular section of copper, with 3 concentric thin layers of plastic insulation, surrounded by a thick layer layer of steel, attempting to be a basic representation of a rectangular copper winding in a motor stator slot with it's various insulation layers - image included below, the red is the copper, technicolour layers are the insulation and the blue is the steel.

I've reduced the model to a 1/8 representation for simplicity, and was trying to assess what happens after a large current has been dumped through the winding, heating it to 200 C, but the steel hasn't had chance to heat up, for which I'm assuming a worst case starting temperature of -40 C.  As I know that the thermal simulation calculates a steady state, I was hoping I could create a representation of this by setting my copper to 200 C and steel to -40 C using the applied temperature load for the whole of each object.  I've added no convection loads or the like as my thinking was this happens very quickly.  

My confusion arises in that when I run the simulation, I'd expect it to produce a temperature gradient between 200 C and -40 C, but I actually see maximum temperatures of 207C and minimum at -60 C.  

Am I misunderstanding how the simulation should be configured, or am I just failing to understand the results that are displayed?

Thanks in advance,

Jake

[John_Holtz]

Hi Jake,

Your expectation is correct. The problem is the accuracy of the math.

You have probably constrained the problem too much ("using the applied temperature load for the whole of each object"). This technically leaves the entire 200 to -40 gradient through the thin insulation, which exposes the next problem. The elements through the thickness of the thin insulation are distorted (unless you used a very, very small elements in the insulation).

Here are some things to try:

1. apply the temperature to the centerline of the copper.  This leaves conduction to distribute the temperature through the rest of the copper.
2. apply the temperature to the "outside" or farthest faces of the steel. This leaves conduction to distribute the temperature through the rest of the steel.
3. You can try modeling the insulation, but since the dimension is so much smaller than the rest of the model, you may always have some problems with the mesh quality. Instead of modeling the insulation, you could eliminate those parts and use the thermal resistance (or thermal conductance) between the copper and steel to simulate the insulation.

By the way, I do not think you will get very good answers to a transient problem by using a steady state solution.

Also, where do the maximum and minimum temperatures occur? My guess is that they occur near a "corner" or a "boundary" where the heat flow is making a sharp turn. Such conditions make it difficult to prevent the temperature from going out-of-bounds. If the out-of-limits occur at one or two points, in some cases it is better to ignore those point than it is to try to fix them.

[John_Holtz]

Hi @Anonymous

I wanted to make sure that you saw my post and ask if you had any questions.

[Anonymous]

Thanks for your advice on this John, and apologies for my delay in replying.  Looking at the results further it looks like you're wholly correct and the max and min temperatures occurred at the boundaries between the large bodies with their solid applied temperatures and the thin insulation layers, so that makes a lot of sense.  

My reasoning for modelling each layer of insulation was that I ultimately wanted to model the thermal stress caused by this scenario on each layer and ascertain if it exceeded the yield stress for any of the individual layers, allowing me to configure different layers to use different materials to create a more representative situation, so I guess that leaves the option of a very fine mesh to minimise distortion and accept that the results aren't going to be brilliantly accurate.

Thanks again,

Jake