I am reading the paper, http://math.mit.edu/~gs/cse/codes/mit18086_navierstokes.pdf

In the paper, the nonlinear term is treated as mix of central central difference and upwind difference using a transition parameter $\gamma \in [0,1]$, and it's defined as $$\frac{U^*-U}{\Delta t} = ((\bar{U}^h)^2 - \gamma \lvert \bar{U}^h \rvert \tilde{U}^h )_x - ((\bar{U}^v \bar{V}^h) - \gamma \lvert \bar{V}^h \rvert \tilde{U}^v )_y $$ The variables $\bar{U}^h$ and $\tilde{U}^h$ are defiend as, $$ (\bar{U}^h)_{i+\frac{1}{2},j} = \frac{U_{i+1,j}+U_{i,j}}{2}\\ (\bar{U}^v)_{i,j+\frac{1}{2}} = \frac{U_{i,j+1}+U_{i,j}}{2}\\ (\tilde{U}^h)_{i+\frac{1}{2},j} = \frac{U_{i+1,j}-U_{i,j}}{2}\\ (\tilde{U}^v)_{i,j+\frac{1}{2}} = \frac{U_{i,j+1}-U_{i,j}}{2}\\ $$

$U^2$ term is $$ ((\bar{U}^h)^2 - \gamma \lvert \bar{U}^h \rvert \tilde{U}^h )_{i+\frac{1}{2}, j} = \bar{U}^h \begin{cases} \left(\frac{1-\gamma}{2}\right) U_{i+1,j} + \left(\frac{1+\gamma}{2}\right) U_{i,j} \;\;\; \text{if} \;\; \bar{U}^h\ge0\\ \left(\frac{1+\gamma}{2}\right) U_{i+1,j} + \left(\frac{1-\gamma}{2}\right) U_{i,j} \;\;\; \text{if} \;\; \bar{U}^h\lt 0\\ \end{cases} $$

And it says, "One can easily see that this becomes averaged central differencing for $\gamma=0$ and conservative upwinding for $\gamma=1$."

But I cannot see this becomes the conservative upwind difference when $\gamma =1$. I thought the conservative upwinding is $$ \begin{cases} (U_{i,j}^2-U_{i-1,j}^2)/\Delta x \;\;\; \text{if}\;\;U_{i,j}\ge0 \\ (U_{i+1,j}^2-U_{i,j}^2)/\Delta x \;\;\; \text{if}\;\;U_{i,j}\lt0 \\ \end{cases} $$

If I plug $U_{i,j}$ in $\bar{U}$ and $\tilde{U}$ with the condition $\bar{U}^h \ge 0$ and $\gamma=1$, the first term becomes $$ ((\bar{U}^h_{i+\frac{1}{2}} U_{i,j}) - (\bar{U}^h_{i-\frac{1}{2}} U_{i-1,j}))/{\Delta x}\\ =((\frac{U_{i+1,j}+U_{i,j}}{2}U_{i,j}) - (\frac{U_{i,j}+U_{i-1,j}}{2} U_{i-1,j}))/{\Delta x} $$

This is not the conservative upwinding difference I thought. I am confused because it depends on $\bar{U}^h$, not $U_{i,j}$. Can someone explain and show the right conservative upwinding difference form?


1 Answer 1


Your last term $$ ((\frac{U_{i+1,j}+U_{i,j}}{2}U_{i,j}) - (\frac{U_{i,j}+U_{i-1,j}}{2} U_{i-1,j}))/{\Delta x} $$ is a conservative approximation.

The idea of conservative upwind scheme for $(u^2)_x$ is to consider it as a difference of fluxes $u u$ at $i+1/2$ and $i-1/2$. The arithmetic averages in this formula are velocities at these points, say $U_{i+1/2,j}$ and $U_{i-1/2,j}$. The values $U_{i,j}$ and $U_{i-1,j}$ are upwinded values of $u$ at these points. In fact this formula shall be used when the averaged velocities are positive, i.e. $U_{i+1/2,j}>0$ and $U_{i-1/2,j}>0$.

P.S. here I add the derivation of such conservative upwind approximation (where I skip the index $j$): $$ (u u)_x(x_i) \approx \frac{(u u)(x_{i+1/2})-(u u)(x_{i-1/2})}{\Delta x} \approx \frac{U_{i+1/2} U_i -U_{i-1/2}U_{i-1}}{\Delta x} $$ if $$ U_{i+1/2}=0.5 (U_i+U_{i+1}) > 0 \hbox{ and } U_{i-1/2}=0.5 (U_{i-1}+U_{i}) > 0 $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.