 |
 |
Troops: I have reviewed your lab 11 submissions, most of which contained most of the required information and were well presented. The M = 51 uniform mesh produces a monotone Pe = 100 solution, hence M = 501 would be required for Pe = 1000. Several reports presented nodal data close ups confirming that getting a nonuniform mesh genuinely!! monotone solution for Pe = 100 and 1000 is very sensitive to pr. Most of you reported a pr TOO small for Pe = 100. The precise LEAST values are 0.90 and 0.65 for Pe = 100, 1000, as determined by tests at pr increments of 0.01 done by a previous class participant. M refinement for the traveling wave always improves the solution, ie., peak better symmetric and trailing dispersion error wake magnitude smaller. As you change M you must change t-final and the print interval to get the solutions to plot having traveled the same distance, which many of you did not do. Think about it! Increasing Courant number basically increases time step size, hence solution is computed faster but solution has aggravated dispersion error. The solution is unstable for theta < 0.5, and is artificially diffused for any theta > 0.5 is the essential conclusion. Lab 12 asks you evaluate the multi-dimensional forms of these two verification problems. For the steady Pe problem, you must implement the TWS algorithm to generate monotone solutions for large Pe. For Pe = 1000, go to a M = 160x160 mesh at least and then you should be able to validate the theory predicted Order h^4 convergence for TWS. Plot the theory straight line adjacent to your data to assess this, as the coarse mesh solutions are artificial diffusion-polluted. End of semester is near. Put on your calendar that we will MEET as a class in CFD Lab, Perkins 318, on the last class day, Tuesday Nov 30 at 2:10 pm. If it is impossible for you to attend, notify me asap to make an alternate arrangement. The final exam will be opened and discussed and you will fill out class critique forms. AJB