I've been working on this for months and can't find a good answer.

I'm looking at the incompressible Euler equations in cylindrical coordinates ($r$, $\theta$, $z$), and I am looking at the non linear terms' impact on the momentum conservation equation, notably on the regularity on the axis.

I am using the Fourier transform to describe the $\theta$ dependence of the velocity:

$\overrightarrow{u} = \sum\limits_{m=-N}^{N} \overrightarrow{u_m} e^{im\theta}$

Classically, the regularity of the solution in velocity is expressed with derivatives up to a certain order being null, or explicitly giving the $r$ dependence. If one considers the change of variable $u^+ = u^r + i u^\theta$ and $u^- = u^r - i u^\theta$, one can find some requirements for the regularity of the velocity. Different sources agree on the following properties to be imposed on each terms ($u^+_m$,$u^-_m$, $u^z_m$) of the velocity, for $m>1$:

$u^+_m = r^{m+1} \hat{u}^+_m$

$u^-_m = r^{m-1} \hat{u}^-_m$

$u^z_m = r^{m} \hat{u}^z_m$

However, if one injects this form of the velocity in the momentum conservation equation, one get, for the $z$ equation for example, dividing by $r^m$:

$\frac{\partial \hat{u}^z_m}{\partial t} + \sum\limits_{m_1,m_2\\ m_1 + m_2=m}\left(\frac{r \hat{u}^+_{m_1}}{2} \frac{\partial \hat{u}^z_{m_2}}{\partial r} + \frac{\hat{u}^-_{m_1}}{2r}\frac{\partial \hat{u}^z_{m_2}}{\partial r} + m_2 \hat{u}^+_{m_1}\hat{u}^z_{m_2} + \hat{u}^z_{m_1}\frac{\partial \hat{u}^z_{m_2}}{\partial z} \right)= -\frac{1}{r^m}\frac{\partial p}{\partial z} $

(The non linear terms look different normally with the negative $m_1$ or $m_2$ values, but the point I'm trying to make still holds.)

This equation is the problem I have. One can see that the terms imposed by the non linear terms are not as regular as $u^z$ should be. More precisely all the terms $\frac{\hat{u}^-_{m_1}}{2r}\frac{\partial \hat{u}^z_{m_2}}{\partial r}$ are singular on the axis. This problem is also present in the other equations of the momentum conservation equation. I have tried to improve this by adding some properties to the velocity, but since those properties also have to be verified, it does not help.

My questions are therefore:

-Is there any way to impose the regularity of the non linear terms on the axis ?

-Are there other conditions for the velocity on the axis that could improve this behavior ?

-Have I made any false assumptions or mistakes ?

Any thoughts and ideas are welcomed !

Edit: I have come across a paper (https://www.semanticscholar.org/paper/Formulation-of-a-Galerkin-spectral-element-fourier-Blackburn-Sherwin/b8fcada2f591251942b9ebc8f46f68b5855e7046) that advocates to put the problematic terms to zero. But the argumentation is not clear to me...



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.