Troops: I have
reviewed and commented individually on your lab 12 reports. More than half
the class got things pretty correct. No one documented the asymptotic
convergence rate of near order 3 for the TWS solution at finer M meshes for Pe =
1000. FYI, the theory underlying
the TWS algorithm predicts linear basis convergence rate is 4 on a sufficiently
refined mesh – good stuff!
In general, about half of the reports comparing the GWS and TWS solutions
plots for all M for Pe = 10 show no difference, which is correct with both being
non-monotone for M < 40^2. Note
further that for all M > 40^2, GWS and TWS solutions appear identical (some
reported no convergence – good!) which means the algorithms produced an
interpolation of the exact solution!!
Hence the computed
convergence rate dropped to near zero – precisely correct!
.
There are three key points with this computer lab exercise. The first remains, “you get out what you
put in” and to your credit none had input coding errors for TWS. The second is that the TWS beta term for
small Pe does not influence the GWS solution at all, as the theory
predicts. The third is that use of
the solution adapted non-uniform meshing process with the TWS algorithm could
produce the precisely correct monotone solution at Pe = 1000 on an M = 20^2
mesh!
There are a few of you who failed to mount
this report, which earned you a 0/6 grade.
The archive is open. Recall we meet Tuesday for class
wrap-up. AJB