6
$\begingroup$
  1. Can you explain for me what is convection-dominated problems? Definition and examples if possible.

  2. Why don't we can apply standard discretization methods (finite difference, finite element, finite volume methods) for convection-dominated equation?

  3. What is the importance of global Péclet number in this problem?

$\endgroup$
4
  • $\begingroup$ Welcome to MSE, I added a PDE tag to the question. $\endgroup$
    – Shuhao Cao
    Jun 18 '13 at 22:02
  • 1
    $\begingroup$ Convection-dominated equation normally have strongly irregular solutions (large jumps, discontinuities), traditional numerical methods always assume the solution is smooth. That's why many people using Discontinuous Galerkin finite element to numerically solve this type of problems. $\endgroup$
    – Shuhao Cao
    Jun 18 '13 at 22:05
  • $\begingroup$ Thanks @ShuhaoCao. Now, I can have better understanding about it. And what about the third question about Péclet number? $\endgroup$ Jun 18 '13 at 22:14
  • $\begingroup$ I will write up an answer then. $\endgroup$
    – Shuhao Cao
    Jun 18 '13 at 22:17
9
$\begingroup$

Suppose we are modeling a quantity $u$, say the concentration of a chemical, driven by a flow in some fluid in a region $\Omega$ with no source (meaning we are not adding more chemical into the fluid after starting the timer). Then the convection-diffusion pde to describe the phenomenon is: $$ \frac{\partial u}{\partial t} = \nabla \cdot (D \nabla u - \vec{b}u ).\tag{1} $$ This comes from linear hyperbolic conservation law (in the differential form): $$ \frac{\partial u}{\partial t} = -\nabla \cdot \vec{F}, $$ where $\vec{F}$ is the flux vector, and $$ \int_{\partial \Omega}\vec{F}\cdot n\,dS $$ measure the amount of chemical flows out from the domain of interest $\Omega$, and $\vec{F} = -D \nabla u + \vec{b}u$. $D$ is the diffusion constant, it describes how the quantity diffuses, $\vec{b}$ is a flow field, it may carry $u$ around. If we assume $D$ is a constant and $\vec{b}$ is a divergence free flow (incompressible), then (1) is: $$ \frac{\partial u}{\partial t} = \underbrace{D\Delta u}_{\text{Diffusion term}} - \underbrace{\vec{b}\cdot \nabla u}_{\text{Convection term}}. \tag{2} $$ Convection-dominance just means, in the convection-diffusion equation (2): $$D\ll \|\vec{b}\|.$$ The diffusion term is very small relative to the convection term.

Intuitively, diffusion means "smooth", while convection could possibly contain non-smoothness due to its derivation from an integral conservation law. For example, if we make the diffusion term $D\sim 0$ almost gone, the solution may looks like the following in a 2D problem:

adv

where the solution's discontinuity "flows" with the field $\vec{b}$ on the right. If the diffusion constant is huge, i.e., not convection-dominated, then the steep slope will be smoothened as the quantity being transported along with the flow $\vec{b}$.


As I said in the comments, traditional numerical methods relies on the smoothness of the solution. Solution like above can't be solved by tradition numerical schemes like you listed. That's why people using Discontinuous Galerkin finite element to numerically solve this type of problems, while we don't impose the continuity conditions of the numerical solutions, in the meantime the conservation can be guaranteed numerically.

For the last question, please just read 2.1.4 here on page 33.

$\endgroup$
5
  • $\begingroup$ Thank you so much for your help. That's great answer for me now. :) $\endgroup$ Jun 20 '13 at 9:29
  • $\begingroup$ @DinhAnhThi :-) $\endgroup$
    – Shuhao Cao
    Jun 21 '13 at 18:12
  • $\begingroup$ Hi Mr. @ShuhaoCao, I am using another method to solve this problem, it's not discontinuous galerkin finite element. It's streamline diffusion method. Do you know this method? The basic idea of this method consists in the addition of suitably weighted residuals to variational formulation. In this addition term, there is a parameter $\delta_K$. My problem is how to determine this parameter appropriately. $\endgroup$ Jun 29 '13 at 13:20
  • $\begingroup$ Mr. @ShuhaoCao, can you give me more details (equation, boundary condition, initial condition, parameters,...) about the example you gave in this answer in order that I can test them on FreeFEM++? Additionally, how can we recognize this problem visually? Thank you so much! $\endgroup$ Jun 29 '13 at 16:11
  • $\begingroup$ @DinhAnhThi Hi, the question for FreeFEM++ is better suited on scicomp.stackexchange.com . Also Claes Johnson's book amazon.com/… has chapter 9 dedicated to hyperbolic conservation law and introduces both streamline diffusion and DGfem. $\endgroup$
    – Shuhao Cao
    Jun 29 '13 at 17:24

Your Answer

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

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