[claudia.h]

I tried to simulate the 3D natural convection flow in pipe with 45° angle to the horizontal. The ambient temperature is 30°C, so I set the Operating Temperature to 303 K. For operating density I used the density of the lowest occurring temperature of the fluid (1020 kg/m³ at 40°C).

The fluid is a polypropylene-glycol mixture and so I use a polynomial function for density, cp, lambda and viscosity.

I use the k-omega SST Model with the following boundary conditions:

Pressure Inlet:
Gauge Total Pressure: 0 Pa,

Specification Method: Intensity and Hydraulic Diameter
Thermal: Total Temperature 313 K

Pressure Outlet:
Gauge Pressure: 0 Pa

Specification Method: Intensity and Hydraulic Diameter
Thermal: Backflow Total Temperature 363 K

On the front side I use a Thermal BC with a Heat Flux of 250 W/m² and on the back side I use a thermal BC (convection) with a Heat Transfer Coefficient of 3 W/m²K and a Free Stream Temperature of 353 K.

When I visualize velocity vectors after simulation (after 1500 Iterations) then the velocity-vectors have the wrong direction (from outlet to the inlet) and that is impossible. It would be a great help for me if anybody has an idea could cause this phenomena. Thanks!

[eng_s_sadeghi]

The problem is that you should just simulate a simple 2D model at first. After finding all the restrictions of natural convection, turn to 3D model. Natural convection is so sensible of mesh size.

[flotus1]

For natural convection simulations, it is quite helpful to initialize the flow field with a simulation of forced convection.
Just set the inlet velocity to a small non-zero value. Run till convergence, then start your natural convection simulation from this initial flow field.

"Typical" mistakes in a natural convection simulation could be that the gravity points in the wrong direction and that the heat flux at the surface has the wrong sign (positive -> entering the surface, negative-> leaving the surface)
Sorry if these two possible mistakes are too simple, I just wanted to be sure .