Allow the grid refinement procedure to limit the number of points added per refinement stage

[GoogleCodeExporter]

What is the expected output?

During a 1-D solution, I have found that the grid refinement parameters slope, 
curve and ratio can sometimes be set too aggressively such that many new grid 
points are added at one time. These new grids can be difficult to solve.

In these cases, it often helps to have the ability to limit the number of grid 
points that can be added during the refinement stage so that the solution stays 
within the steady-state domain of convergence. Luckily, Cantera's grid 
refinement procedure has been written in a way that allows this feature to be 
added easily. I have attached one possible way to implement this.

While adding this, I also suggest that the function "setMaxPoints" be added to 
the flat interface so that the maximum number of grid points (m_npmax) could be 
controlled in cython.

Original issue reported on code.google.com by watso...@gmail.com on 10 Apr 2013 at 7:55

Attachments:

• refine.cpp
• refine.h

[GoogleCodeExporter]

Do you have an example of a case where this approach improves the behavior of 
the solver, either allowing convergence of a case that failed before or even 
just improving the overall convergence rate? It's not immediately clear that 
spreading the point additions over multiple refinement passes is a good idea, 
since it guarantees that you will have to do more steady-state solves.

Original comment by yarmond on 12 Apr 2013 at 3:11

• Added labels: Type-Enhancement
• Removed labels: Type-Defect

[GoogleCodeExporter]

Original comment by yarmond on 8 May 2013 at 2:40

• Changed state: Feedback

[GoogleCodeExporter]

Original comment by yarmond on 25 Sep 2013 at 11:11

• Changed state: WontFix