[kbzowski]

Hi,
I am trying to model cooling of steel block (initial 900 oC) through which flows mineral oil. The geometry looks like: http://i.imgur.com/lkIYAqA.png
Oil is pumped (inlet boundary condition) with velocity around 0.2 m/s. Initial temperature of the fluid is 12 oC. The simulations assume that the oil fills the channel already.

Materials properties were verified from few different sources, so I believe they are correct. Only viscosity is provided as the temperature function.

According to the physical experiment after 1500 sec. block should have no more then 400 oC. It was measured by a few thermocouples. In case of my simulations temperatures did not drop below 750.

Moreover exiting oil should have around 200oC but in my case the temperature increase only slightly (max 20 oC on the outlet).

I suppose if temperature of the oil is not increasing the viscosity of the oil does not dropping and this is the major problem with slow cooling.

What I am missing?

Basics of the CFX simulations:
Transient simulation, total time 2000 s.
Inlet: normal velocity 0.2 m/s, temperature 12oC
Outlet: Outlet with static relative pressure 0 atm.
Turbulence: I was trying with laminar as well as k-epsilon turbulence. With laminar - cooling was more rapid (but far from experiment). This is a little strange, because I thought that heat transfer is mainly through turbulence in such a case.
Interface: Conservative interface flux with thermal resistance 0.00001 m^2 K/W
Solid: Initial 900 oC
Fluid: Initial 12oC, 0 velocity

[ghorrocks]

You need to work out if this flow is laminar or turbulent. If laminar use the laminar flow model, if turbulence use a turbulence model.

Have a look at this FAQ: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

Finally: If you cool steel from 900C to 10C over ~30 minutes the steel will undergo several phase changes (BCC to FCC and back again). These phase changes will absorb and release large amounts of heat as you pass through the phase changes. If you do not take these phase changes into account you can be a long way off. Have you taken this into account?

[kbzowski]

Thank you for the reply.
I've already read the FAQ and checked almost every point. I am experimenting with this simulation for a quite a long time already. I can imagine that results can be differ by 50-100 oC in case of applying wrong turbulence model. But in my case the difference is really huge.

I believe that I am not aware of some physical phenomena which occurs during the flow. Or I have not selected something important in CFX. The system is closed and tightly insulated, so I am not able to see what happens inside.

In case of phase transformations. I am aware of it, but currently I am not including such a phenomena in the simulation. I have access to experimental data (cooling 900-400 oC) and I am not able to see there any effect of the phase transformation. Moreover as far as I know phase transformation energy (martensitic transformation in this case) would increase local temperature. So I believe this is not a point here (at least for now).

[ghorrocks]

If you are getting less heat transfer with a turbulence mode then that is very strange as turbulence should only increase heat transfer. Are you sure the flow resistance is correct?

My recommendation here is to do a fluid only model for now. So model the fluid region by itself (so no solid) and apply a constant temperature or convection BC on the walls. Do a careful sensitivity study on mesh, convergence tolerance and boundary proximity. Make sure you can get this simple model accurate and reliable before you add the solid region back in.

Phase transformation: Yes, the temperature will probably rise for a while as you go through a phase transition. Didn't you say that after 1500s the temperature should be about 400C? Most steels will not go Martensitic when cooled that slowly. It is the Austenitic/Bainite/Pearlite transition you are probably seeing there.