GreenPolyProducts/cantera
5
0
Fork 0
mirror of https://github.com/Cantera/cantera.git synced 2026-08-19 01:14:44 -05:00
Code Issues Packages Projects Releases Wiki Activity

[Examples] Pin finite-difference Jacobian for high-pressure regrid cases

Browse Source 
The high-pressure H2/O2 counterflow diffusion cases in'
flame_solver_timestep_controls.py sit at the edge of solvability, where
the analytic and finite-difference Jacobians (which agree to ~1e-7) steer
the solver onto different trajectories. With the analytic Jacobian now the
default, these cases no longer converged within the example's original
regrid/timestep budgets.

Pin finite differences for the counterflow diffusion helper and document
that FD is empirically more robust for these high pressure diffusion flames,
restoring the original example behavior.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Ray Speth
2026-07-03 09:54:37 -05:00
committed by Ingmar Schoegl
co-authored by Claude Opus 4.8
parent 5f666e4b25
commit a6a5f5ede6
1 changed files with 10 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
samples/python/onedim/flame_solver_timestep_controls.py
+10
View File
@@ -159,6 +159,15 @@ plt.show()
# too coarse for the steady-state solver to converge. The ``time_step_regrid`` option
# lets the solver refine the grid and retry when a timestepping sequence has reached
# ``max_time_step_count``.
#
# These cases pin the finite-difference Jacobian
# (``jacobian_mode = "finite-difference"``) rather
# than using the default analytic Jacobian. While the analytic Jacobian is more accurate
# and generally substantially faster, for high-pressure diffusion flames like this
# one, the finite-difference Jacobian can be more robust: the (small) differences
# between the two steer the pseudo-transient timestepping onto different paths, and here
# the finite-difference path grows the timestep more freely and reaches steady state
# within the modest regrid and timestep budgets used below.
diffusion_width = 30e-3
diffusion_initial_points = 20
diffusion_fuel_mdot = 0.3
@@ -176,6 +185,7 @@ def make_regrid_flame(
flame = ct.CounterflowDiffusionFlame(
gas, grid=np.linspace(0.0, diffusion_width, diffusion_initial_points)
)
flame.flame.jacobian_mode = "finite-difference"
flame.max_time_step_count = max_time_step_count
flame.set_refine_criteria(ratio=2.0, slope=0.06, curve=0.08, prune=0.02)
flame.time_step_regrid = regrid_max