Register Blogs Members List Search Today's Posts Mark Forums Read

 February 11, 2011, 11:00 2D linear advection equation #1 New Member   Miguel Caro Join Date: Apr 2010 Posts: 26 Rep Power: 7 Well, trying to solve a 2D linear advection equation u_t + au_x + bu_y = 0; u_0(x,y,0) = sin( 2pi* x ) sin( 2pi y), (x,y) 0,1) x (0x1) , periodic boundary conditions with exact solutions u(x,y,t) = sin (2pi (x-t) ) sin (2pi (y-t) ) i implemented this discretization : u_i,j^{n+1} = u_i,j^n - dt/dx(Fi+1/2 - Fi-1/2) - dt/dy(Gi+1/2 - Gi-1/2); CFL = max(dt/dx;dt/dy); But the solution did not agree with the exact solution at a time t=2.0 Does anybody has a suggestion of what kind of algorithm to implement? I would like to imlement a new-high resoluton scheme and i woud like to evaluate the convergence order. For this i would lik e to implement a technique including a flux limiter. I realyy appreciate any help Thanks

 February 11, 2011, 12:31 #2 Senior Member   Join Date: Nov 2009 Posts: 411 Rep Power: 10 Try to increase the order of your temporal discretization by using a Runge-Kutta method (order 4 should do). This will allow you to use a reasonable time step and to obtain a more precise solution. Or, you may consider using an implicit temporal discretization. Do

 February 11, 2011, 13:30 #3 New Member   Miguel Caro Join Date: Apr 2010 Posts: 26 Rep Power: 7 Thanks Do, well i implemented yet a third order temporal RK, but with a more refined grid the numerical solution has an acceptable approximation to the reference solution. But I think i have to implement a fourth order temporal RK (usually it is called Strong Stability Prerserving Runge-Kutta of fourth order, for example. Gottlieb, Shu and others) to obtain better approximation. Regards