Convergence Problem in k-e model
 

Hi Everybody !

I am simulating flow of liquid steel inside a tundish. Trough shaped vessel of 2910 x 500 x 700 mm dimensions. When I run different cases of it with 500 , 400 or 300 mm height instead of 500mm I find the residuals behaving in a different manner. Sometimes it converges smoothly with reduced relaxation parameters. sometimes it does not. I used PISO for pressure velocity coupling. Why is it that sometimes I get a convergence problem and there is always a mass imbalance of order of 10-3 when the inlet mass is 10 kg/s approximately. Can anyone throw some light where I am making a mistake.

Thanks in advance

Rajeev

Re: Convergence Problem in k-e model
 

Hello,

we don't get much information about your case.

What are your boundary conditions? How do you initialize your solution? How does your grid looks like? What about the quality of your mesh?

A mass imbalance of 1e-2 is not that bad, is it? 0.1% of the mass-inflow. Did you monitor tis imbalance? Is it still changeing when your residuals look stable?

Why the PISO? What results do you obtain by using the default settings?

greetings,

Laika, still orbiting

Re: Convergence Problem in k-e model
 

Hello Laika,

Thanks for showing interest,

Well I am using Fluent to solve flow simulations in steel making. For continuous casting molten steel is poured into a trough shaped vessel called tundish. Usually the tundish has multiple outlets for the liquid steel to come out and fall into the continuous casting mold where it starts freezing into a solid shell.

The liquid steel is poured at a temp of 1550 deg. c. The sides of the tundish is lined with refractory and inside the vessel refractory dams are put to change the flow direction.

I am fitting hexahedral cells using cooper scheme because the outlets are circular in shape.

At the inlet which is an area of 65mm dia on the free surface I am putting boundary condition as mass flow rate 12.05 kg/s, density of the material being 7000 kg/m3.

At the outlet I am putting pressure outlet as the boundary condition with backflow temp of 1750 K. (Normally the outlet flow is higher than this temp). Rest of the surfaces (side walls) I am giving heat flux as the thermal boundary.

The top surface is a free surface which I am specifying as wall with specified zero shear stress.

Normally when I run this solution with scaled residuals as monitoring criteria it hovers in the range of 10e-4 with energy residuals aroung 10e-7.

Mass imbalance still remains as in the order of 10e-2. When I use PISO with second order discretization scheme then the flow residuals goes down appreciably.

Mesh size is roughly hexahedral grids of 20mm. Skewness varying from 0-0.8.

Could you through some light if I can reduce the residuals. I am using the standard k-e model.

If intersted I can send you the requisite journal files which you can run and see for yourself where I am making any mistake.

Re: Convergence Problem in k-e model
 

Hello Laika,

Thanks for showing interest,

Well I am using Fluent to solve flow simulations in steel making. For continuous casting molten steel is poured into a trough shaped vessel called tundish. Usually the tundish has multiple outlets for the liquid steel to come out and fall into the continuous casting mold where it starts freezing into a solid shell.

The liquid steel is poured at a temp of 1550 deg. c. The sides of the tundish is lined with refractory and inside the vessel refractory dams are put to change the flow direction.

I am fitting hexahedral cells using cooper scheme because the outlets are circular in shape.

At the inlet which is an area of 65mm dia on the free surface I am putting boundary condition as mass flow rate 12.05 kg/s, density of the material being 7000 kg/m3.

At the outlet I am putting pressure outlet as the boundary condition with backflow temp of 1750 K. (Normally the outlet flow is higher than this temp). Rest of the surfaces (side walls) I am giving heat flux as the thermal boundary.

The top surface is a free surface which I am specifying as wall with specified zero shear stress.

Normally when I run this solution with scaled residuals as monitoring criteria it hovers in the range of 10e-4 with energy residuals aroung 10e-7.

Mass imbalance still remains as in the order of 10e-2. When I use PISO with second order discretization scheme then the flow residuals goes down appreciably.

Mesh size is roughly hexahedral grids of 20mm. Skewness varying from 0-0.8.

Could you through some light if I can reduce the residuals. I am using the standard k-e model.

If intersted I can send you the requisite journal files which you can run and see for yourself where I am making any mistake.

Thanks Rajeev