Unreasonable temperature values and lack of heat transfer with thin solid walls

    • David Dehler

      During the CFD simulation runs, I am stuck at a few places and would like better clarity on the errors, the concepts I need to improve, and how I can improve upon them. I have taken just a part of my model to explain my issues. The horizontal fluid (air) is entering at 400K, which is supposed to transfer the heat through the common shared wall to the vertical fluid domain at 300K.

      I have designed a model with one inlet on the left side and a corresponding outlet on the right side. TA hollow plate is introduced in the middle, which carries another fluid from an inlet at the bottom towards the outlet at the top. I wish to achieve heat transfer from the high-temperature lateral flow fluid to the cross-flow fluid (flowing vertically in the hollow plate from the bottom to the top) through their shared wall domain.

      Issues faced:

      1) Graph.png - The energy residual of the graph is not converging, which results in unusually high temperatures (in the order of 10^3 K) when the lateral entry fluid temperature is 500K, and the vertical flow bottom entry fluid temperature is 300K. 

      What is the reason for this?

      2) When the same problem is solved by suppressing the solid domain and considering just the fluid domains, the temperature values are reasonable (still incorrect, though as in the absence of solid domains, heat transfer should not take place as there is a gap between the two fluid regions and no contact) within the acceptable limits. However, this won't help our case in establishing heat transfer, as the two fluids are supposed to be separated by a solid domain.

      This problem has attempted to be solved through both laminar and turbulent flow models (k-omega standard, k-omega SST, and k-epsilon - enhanced wall treatment), different solution methods (Least squares cell-based, Green Gauss node-based, and Green Gauss cell-based), different combinations of the first order and second-order upwind for momentum, energy, pressure, specific dissipation rate, and turbulent kinetic energy. I have taken one such simulation setting and attached the corresponding images with this email.

      The velocity flow profile and average pressure drop look fine, but since the energy graph isn't converging, the temperature contour profiles are absurd. The temperature is roughly expected to be the same throughout the fluid domain without the solid domain when no heat transfer is taking place,. Still, I would like to study how the heat transfer from one fluid domain to the other takesplace in presence of the solid domain. Even the mesh looks uniform throughout (Element size 5mm). There is some fundamental issue with the meshing when the solid domain is involved, which I haven't understood as the mesh looks consistent throughout the model with various method sizing, edge sizing, and element sizes assigned.

      Additionally, I tried my best to improve the orthogonal quality > 0.1 of the model after applying various edge sizing and face meshing. During this process, I still managed to get the lowest orthogonal quality of 0.6 and a total of 4 million+ elements, but this mesh isn't getting exported to the setup tab as my PC lacks such a high computational power. Is this common for such a simple model? 

      Furthermore, I aim to have adiabatic walls throughout the model, except the common wall region of both fluids through which the heat transfer will take place. The environment is atmospheric air, and no effect of the outside environment should interfere with the solid and fluid domains. I have tried using "System Coupling" (Shouldn't system coupling be used if the solid and fluid domains are free to interact with each other? In my case, I assigned system coupling so that heat transfer can take place from fluid to solid and to another fluid) as well as "Heat Flux = 0" for adiabatic wall conditions. When the model is complicated, how should I approach selecting the inner walls and assigning separate wall conditions since I have "Added material" to all extrusions in order to have one single solid domain? Should I be extruding solid bodies by "Add Frozen" to access some small difficult-to-access inner walls separately, which isn't possible to assign while extruding using "Add material"?
      Do let me know if I was unclear while explaining something. I am looking forward to improving my results.

      Thanking you

    • Rob
      Ansys Employee

      There are a few problems here. 

      First one is you may not have contact between the two fluid streams. You should have a wall and wall:shadow pair between the zones. As you're talking about add frozen that's DesignModeler so you want to create a multibody part before meshing. To create the geometry go through the tutorials as there are several approaches that are equally valid. 

      I can't work out whether there's a solid casing around the larger channel or not. 

      Next problem is the mesh. You need more resolution in the zones. 

      Finally, for the CFD part, you need a longer inlet & outlet for the larger channel. Have a look at the Learning courses on here, but you really want to move the inlet & outlet to some diameters from the area of interest. 

Viewing 1 reply thread
  • You must be logged in to reply to this topic.