Discretization of mass conservation equation in a 2D staggered grid

kiwiguigou · April 15, 2016, 16:34

Hi everybody,

As mentioned in the title I would like to solve the mass conservation to simulate the boiling phenomenon into a tube bundle. I am facing some issues to discretize these equations into a staggered grid after to have applied an integration into a finite volume.

Accordingly to the classical staggered grid: the flow properties are content in the control volume whereas the velocity information is on the bondaries of this control volume. Here is a representation of the staggered grid I use (for the notations)

My governing equation is:

$\frac{\delta \left({\propto }_k{\rho }_k\right)}{\delta t}+\frac{\delta \left({\propto }_k{\rho }_ku_k\right)}{\delta x}+\frac{\delta \left({\propto }_k{\rho }_ku_k\right)}{\delta y}={\left(-1\right)}^k\left(\mathrm{\Gamma }e-\mathrm{\Gamma }c\right)+{\dot{M}}_{k\ in}$

Where the subcript k is for the phase type (1 for liquid; 2 for gas)

$\mathrm{\Gamma }c\$ and $\mathrm{\Gamma }e\$ are the evaporation and condensation rate defined as:

$\mathrm{\Gamma }e=\frac{{\propto }_1{\rho }_1}{{\tau }_e}.\frac{\left(h_1-h^{'}\right)}{\left(h^{''}-h^{'}\right)}$
$\mathrm{\Gamma }c=\frac{{\propto }_1{\rho }_1}{{\tau }_c}.\frac{\left(h^{'}-h_1\right)}{\left(h^{''}-h^{'}\right)}$

Where H" is the staturated enthalpy of the gas, H' is the saturated enthalpy of the liquid and $\tau$ is the relaxation time. First question: does someone have a definition to calculate that relaxation time?

After having integrated the mass conservation equation for the control volume I;J, I get:

$\left[{\left({\propto }_1{\rho }_1\right)}^n_{I;J}-{\left({\propto }_1{\rho }_1\right)}^{n-1}_{I;J}\right]\frac{\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}}{\mathrm{\Delta }\mathrm{t}}+\left[{\left({\propto }_1{\rho }_1u_1\right)}^n_{i+1;j}-{\left({\propto }_1{\rho }_1u_1\right)}^n_{i;j}\right]$ $\mathrm{\Delta }\mathrm{y}+\left[{\left({\propto }_1{\rho }_1v_1\right)}^n_{i;j+1}-{\left({\propto }_1{\rho }_1v_1\right)}^n_{i;j}\right]\mathrm{\Delta }\mathrm{x}=$ $\ {\left[\mathrm{-}\mathrm{\Gamma }e+\mathrm{\Gamma }c+{\dot{M}}_{1\ in}\right]}^n_{I;J}\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}$

By making the assumption that the velocity field is known, I can resolve this equation and find the void fraction and the density. My issue arrives now:
I don't know how to approximate the terms of void fraction and density which are at the bondaries of the control volume because the information is not stocked there (ex: ${\left({\propto }_1{\rho }_1\right)}^n_{i+1;j}$ )

If someone can help me on this I will be very grateful.

In advance, thank you.

Kiwi

FMDenaro · April 15, 2016, 18:35

your grid arrangement is not totally clear to me from the picture...

I immagine that rho(i,j), u(i,j), v(i,j) are staggered each other, right?
For example, if rho(i,j) is at the position x(i),y(j), then u(i,j) is at x(i)+dx/2,y(j) and v(i,j) is at x(i),y(j)+dy/2?

If it is true, the integration for rho is for the CV centred at x(i),y(j).

kiwiguigou · April 16, 2016, 08:26

Hi FMdenaro and thank you for your answer.

Yes you are right; if we use the notation (i,j) (i+1/2) and (j+1/2) the equation can consequenlty be written as:

$\left[{\left({\propto }_1{\rho }_1\right)}^n_{i;j}-{\left({\propto }_1{\rho }_1\right)}^{n-1}_{i;j}\right]\frac{\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}}{\mathrm{\Delta }\mathrm{t}}+\left[{\left({\propto }_1{\rho }_1u_1\right)}^n_{i+1/2;j}-{\left({\propto }_1{\rho }_1u_1\right)}^n_{i-1/2;j}\right]$ $\mathrm{\Delta }\mathrm{y}+\left[{\left({\propto }_1{\rho }_1v_1\right)}^n_{i;j+1/2}-{\left({\propto }_1{\rho }_1v_1\right)}^n_{i;j-1/2}\right]\mathrm{\Delta }\mathrm{x}=$ $\ {\left[\mathrm{-}\mathrm{\Gamma }e+\mathrm{\Gamma }c+{\dot{M}}_{1\ in}\right]}^n_{i;j}\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}$

Is it more clear for you with that writing?
So you mean that for the terms ${\left({\propto }_1{\rho }_1\right)}^n_{i+1/2;j}$ I can make the approximation $\frac{{\left({\propto }_1{\rho }_1u_1\right)}^n_{i+1;j}-{\left({\propto }_1{\rho }_1u_1\right)}^n_{i;j}}{\mathrm{\Delta }\mathrm{x}}$ ?

So the equation would become:

$\left[{\left({\propto }_1{\rho }_1\right)}^n_{i;j}-{\left({\propto }_1{\rho }_1\right)}^{n-1}_{i;j}\right]\frac{\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}}{\mathrm{\Delta }\mathrm{t}}+$ $\left[{\frac{{\left({\propto }_1{\rho }_1\right)}^n_{i+1;j}-{\left({\propto }_1{\rho }_1\right)}^n_{i;j}}{\mathrm{\Delta }\mathrm{x}}\left(u_1\right)}^n_{i+{1}/{2};j}-{\frac{{\left({\propto }_1{\rho }_1\right)}^n_{i;j}-{\left({\propto }_1{\rho }_1\right)}^n_{i-1;j}}{\mathrm{\Delta }\mathrm{x}}\left(u_1\right)}^n_{i-{1}/{2};j}\right]\mathrm{\Delta }\mathrm{y}+$ $\left[{\frac{{\left({\propto }_1{\rho }_1\right)}^n_{i;j+1}-{\left({\propto }_1{\rho }_1\right)}^n_{i;j}}{\mathrm{\Delta }\mathrm{y}}\left(v_1\right)}^n_{i;j+{1}/{2}}-{\frac{{\left({\propto }_1{\rho }_1\right)}^n_{i;j}-{\left({\propto }_1{\rho }_1\right)}^n_{i;j-1}}{\mathrm{\Delta }\mathrm{y}}\left(v_1\right)}^n_{i;j-{1}/{2}}\right]\mathrm{\Delta }\mathrm{x}=$ $\ {\left[\mathrm{-}\mathrm{\Gamma }e+\mathrm{\Gamma }c+{\dot{M}}_{1\ in}\right]}^n_{i;j}\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}$

Is that correct ?

I read also that we can do this approximation with the UPWIND scheme but I am not confident to use this scheme because it is based on the value of the velocity; for example:

${\left({\propto }_1{\rho }_1\right)}^n_{i+1/2;j}=$ ${\left({\propto }_1{\rho }_1\right)}^n_{i;j}$ if ${\left(u_1\right)}^n_{i+1/2;j}>0$

OR

${\left({\propto }_1{\rho }_1\right)}^n_{i+1/2;j}=$ ${\left({\propto }_1{\rho }_1\right)}^n_{i+1;j}$ if ${\left(u_1\right)}^n_{i+1/2;j}<0$

But I have also make the test for ${\left(u_1\right)}^n_{i-1/2;j}$ ${\left(v_1\right)}^n_{i;j+1/2}$ and ${\left(v_1\right)}^n_{i;j-1/2}$ so the fill in of my matrix coefficients become very tough ...

What is the best option?

FMDenaro · April 16, 2016, 15:32

Quote:

Originally Posted by kiwiguigou

So you mean that for the terms ${\left({\propto }_1{\rho }_1\right)}^n_{i+1/2;j}$ I can make the approximation $\frac{{\left({\propto }_1{\rho }_1u_1\right)}^n_{i+1;j}-{\left({\propto }_1{\rho }_1u_1\right)}^n_{i;j}}{\mathrm{\Delta }\mathrm{x}}$ ?

not at all ... you have a convective flux at one section it can not be discretized as first derivative!
You can use zeroth-order polynomial in a first order upwind reconstruction or a linear polynomial for a second order reconstruction

kiwiguigou · April 16, 2016, 16:57

Understood I can't use a first derivative solution.
But I am not sure to fully understand what you mean by "linear polynomial for a second order reconstruction". Could you give me an example ?

Thank you.

April 15, 2016, 16:34	Discretization of mass conservation equation in a 2D staggered grid	#1
kiwiguigou New Member kiwi Join Date: Apr 2016 Posts: 3 Rep Power: 10	Hi everybody, As mentioned in the title I would like to solve the mass conservation to simulate the boiling phenomenon into a tube bundle. I am facing some issues to discretize these equations into a staggered grid after to have applied an integration into a finite volume. Accordingly to the classical staggered grid: the flow properties are content in the control volume whereas the velocity information is on the bondaries of this control volume. Here is a representation of the staggered grid I use (for the notations) My governing equation is: $\frac{\delta \left({\propto }_k{\rho }_k\right)}{\delta t}+\frac{\delta \left({\propto }_k{\rho }_ku_k\right)}{\delta x}+\frac{\delta \left({\propto }_k{\rho }_ku_k\right)}{\delta y}={\left(-1\right)}^k\left(\mathrm{\Gamma }e-\mathrm{\Gamma }c\right)+{\dot{M}}_{k\ in}$ Where the subcript k is for the phase type (1 for liquid; 2 for gas) $\mathrm{\Gamma }c\$ and $\mathrm{\Gamma }e\$ are the evaporation and condensation rate defined as: $\mathrm{\Gamma }e=\frac{{\propto }_1{\rho }_1}{{\tau }_e}.\frac{\left(h_1-h^{'}\right)}{\left(h^{''}-h^{'}\right)}$ $\mathrm{\Gamma }c=\frac{{\propto }_1{\rho }_1}{{\tau }_c}.\frac{\left(h^{'}-h_1\right)}{\left(h^{''}-h^{'}\right)}$ Where H" is the staturated enthalpy of the gas, H' is the saturated enthalpy of the liquid and $\tau$ is the relaxation time. First question: does someone have a definition to calculate that relaxation time? After having integrated the mass conservation equation for the control volume I;J, I get: $\left[{\left({\propto }_1{\rho }_1\right)}^n_{I;J}-{\left({\propto }_1{\rho }_1\right)}^{n-1}_{I;J}\right]\frac{\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}}{\mathrm{\Delta }\mathrm{t}}+\left[{\left({\propto }_1{\rho }_1u_1\right)}^n_{i+1;j}-{\left({\propto }_1{\rho }_1u_1\right)}^n_{i;j}\right]$ $\mathrm{\Delta }\mathrm{y}+\left[{\left({\propto }_1{\rho }_1v_1\right)}^n_{i;j+1}-{\left({\propto }_1{\rho }_1v_1\right)}^n_{i;j}\right]\mathrm{\Delta }\mathrm{x}=$ $\ {\left[\mathrm{-}\mathrm{\Gamma }e+\mathrm{\Gamma }c+{\dot{M}}_{1\ in}\right]}^n_{I;J}\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}$ By making the assumption that the velocity field is known, I can resolve this equation and find the void fraction and the density. My issue arrives now: I don't know how to approximate the terms of void fraction and density which are at the bondaries of the control volume because the information is not stocked there (ex: ${\left({\propto }_1{\rho }_1\right)}^n_{i+1;j}$ ) If someone can help me on this I will be very grateful. In advance, thank you. Kiwi

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Setting the height of the stream in the free channel	kevinmccartin	CFX	12	October 13, 2022 22:43
Las term in 2D axisymmetric mass conservation equation	logme	FLUENT	0	June 10, 2015 07:06
error message	cuteapathy	CFX	14	March 20, 2012 07:45
Question on the discretization of momentum equation in icoFoam	MPJ	OpenFOAM	3	October 4, 2011 10:44
Question regarding non-uniform FVM staggered grid	quarkz	Main CFD Forum	2	June 18, 2011 18:18

April 15, 2016, 18:35		#2
FMDenaro Senior Member Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,897 Rep Power: 73	your grid arrangement is not totally clear to me from the picture... I immagine that rho(i,j), u(i,j), v(i,j) are staggered each other, right? For example, if rho(i,j) is at the position x(i),y(j), then u(i,j) is at x(i)+dx/2,y(j) and v(i,j) is at x(i),y(j)+dy/2? If it is true, the integration for rho is for the CV centred at x(i),y(j).

April 16, 2016, 08:26		#3
kiwiguigou New Member kiwi Join Date: Apr 2016 Posts: 3 Rep Power: 10	Hi FMdenaro and thank you for your answer. Yes you are right; if we use the notation (i,j) (i+1/2) and (j+1/2) the equation can consequenlty be written as: $\left[{\left({\propto }_1{\rho }_1\right)}^n_{i;j}-{\left({\propto }_1{\rho }_1\right)}^{n-1}_{i;j}\right]\frac{\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}}{\mathrm{\Delta }\mathrm{t}}+\left[{\left({\propto }_1{\rho }_1u_1\right)}^n_{i+1/2;j}-{\left({\propto }_1{\rho }_1u_1\right)}^n_{i-1/2;j}\right]$ $\mathrm{\Delta }\mathrm{y}+\left[{\left({\propto }_1{\rho }_1v_1\right)}^n_{i;j+1/2}-{\left({\propto }_1{\rho }_1v_1\right)}^n_{i;j-1/2}\right]\mathrm{\Delta }\mathrm{x}=$ $\ {\left[\mathrm{-}\mathrm{\Gamma }e+\mathrm{\Gamma }c+{\dot{M}}_{1\ in}\right]}^n_{i;j}\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}$ Is it more clear for you with that writing? So you mean that for the terms ${\left({\propto }_1{\rho }_1\right)}^n_{i+1/2;j}$ I can make the approximation $\frac{{\left({\propto }_1{\rho }_1u_1\right)}^n_{i+1;j}-{\left({\propto }_1{\rho }_1u_1\right)}^n_{i;j}}{\mathrm{\Delta }\mathrm{x}}$ ? So the equation would become: $\left[{\left({\propto }_1{\rho }_1\right)}^n_{i;j}-{\left({\propto }_1{\rho }_1\right)}^{n-1}_{i;j}\right]\frac{\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}}{\mathrm{\Delta }\mathrm{t}}+$ $\left[{\frac{{\left({\propto }_1{\rho }_1\right)}^n_{i+1;j}-{\left({\propto }_1{\rho }_1\right)}^n_{i;j}}{\mathrm{\Delta }\mathrm{x}}\left(u_1\right)}^n_{i+{1}/{2};j}-{\frac{{\left({\propto }_1{\rho }_1\right)}^n_{i;j}-{\left({\propto }_1{\rho }_1\right)}^n_{i-1;j}}{\mathrm{\Delta }\mathrm{x}}\left(u_1\right)}^n_{i-{1}/{2};j}\right]\mathrm{\Delta }\mathrm{y}+$ $\left[{\frac{{\left({\propto }_1{\rho }_1\right)}^n_{i;j+1}-{\left({\propto }_1{\rho }_1\right)}^n_{i;j}}{\mathrm{\Delta }\mathrm{y}}\left(v_1\right)}^n_{i;j+{1}/{2}}-{\frac{{\left({\propto }_1{\rho }_1\right)}^n_{i;j}-{\left({\propto }_1{\rho }_1\right)}^n_{i;j-1}}{\mathrm{\Delta }\mathrm{y}}\left(v_1\right)}^n_{i;j-{1}/{2}}\right]\mathrm{\Delta }\mathrm{x}=$ $\ {\left[\mathrm{-}\mathrm{\Gamma }e+\mathrm{\Gamma }c+{\dot{M}}_{1\ in}\right]}^n_{i;j}\mathrm{\Delta }\mathrm{x}\mathrm{\Delta }\mathrm{y}$ Is that correct ? I read also that we can do this approximation with the UPWIND scheme but I am not confident to use this scheme because it is based on the value of the velocity; for example: ${\left({\propto }_1{\rho }_1\right)}^n_{i+1/2;j}=$ ${\left({\propto }_1{\rho }_1\right)}^n_{i;j}$ if ${\left(u_1\right)}^n_{i+1/2;j}>0$ OR ${\left({\propto }_1{\rho }_1\right)}^n_{i+1/2;j}=$ ${\left({\propto }_1{\rho }_1\right)}^n_{i+1;j}$ if ${\left(u_1\right)}^n_{i+1/2;j}<0$ But I have also make the test for ${\left(u_1\right)}^n_{i-1/2;j}$ ${\left(v_1\right)}^n_{i;j+1/2}$ and ${\left(v_1\right)}^n_{i;j-1/2}$ so the fill in of my matrix coefficients become very tough ... What is the best option?

April 16, 2016, 16:57		#5
kiwiguigou New Member kiwi Join Date: Apr 2016 Posts: 3 Rep Power: 10	Understood I can't use a first derivative solution. But I am not sure to fully understand what you mean by "linear polynomial for a second order reconstruction". Could you give me an example ? Thank you.