1
$\begingroup$

I want to discretize the LWR(Lighthill-Whitham-Richards)-v partial differential equation (as shown in 1) $$ \frac{\partial u}{\partial t} + \frac{\partial R(u)}{\partial x} = 0 \quad \text{i.e.,} \quad \frac{\partial u}{\partial t} + \frac{dR}{du}\,\frac{\partial u}{\partial x} = 0 \quad \dots\dots \quad \text{(1)} $$ In the equation $u(x,t) \equiv u_x^t$ is the speed at time $t$ at section $x$ and $R(u)$ is given by Greenshield’s speed-density relationship $$ R(u) = u^2 - u\,u_f \quad \implies \quad \frac{dR}{du} = 2u - u_f =: a(u) $$ where $u$ is the stream speed and $u_f$ is the free flow speed or the maximum speed in the section.

I want to use an implicit finite difference scheme (backward time, back ward space as shown in (2) to discretize equation (1): $$ \frac{u_x^{t+1}-u_x^t}{\Delta t} + a(u_x^{t+1})\,\frac{u_x^{t+1}-u_{x-1}^{t+1}}{\Delta x} = 0 \quad \dots\dots \quad \text{(2)} $$

In literature I could find solutions for explicit schemes alone like upwind scheme, downwind scheme, Godunov scheme etc. Can anyone suggest a good discretization scheme or some references in literature for the implicit case shown above in equation (2)?

Thanks for the help!

$\endgroup$
  • 1
    $\begingroup$ To build on what DavidKetcheson has said, the standard advice would be to use something developed for hyperbolic problems, which tend to use explicit schemes. You could try something like a second-order Lax-Wendroff approach with an explicit SSP time integrator (forward Euler, for instance) and a limiter of some sort for the shocks. If you want to go higher-order in space and time, you could use ENO/WENO methods in space and higher-order SSP integrators in time. $\endgroup$ – Geoff Oxberry Mar 14 '14 at 8:31
1
$\begingroup$

Nobody solves this implicitly because it would be very inefficient. If you step with large CFL number, you will smear the shocks and get very poor accuracy.

$\endgroup$

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.