Adiabatic Gibbs Reactor Convergence Issues
Simulate chemical processes using advanced thermodynamic models
Brought to you by:
danwbr
Menu
▾
▴
1 Attachments
#234 Adiabatic Gibbs Reactor Convergence Issues
There seems to be issues with the Gibbs reactor running in adiabatic mode (convergence error). However, when it is run with a specified outlet temperature and the inlet temperature is adjusted so that the energy streem is 0 or near 0, convergence is fine. Perhaps some issue with the adiabatic calculation?
Convergence errors also occur often when a pressure drop is specified above a certain value (i.e. 10 psi works ok in attached file, but 15 psi fails).
Related
Discussion
Hi Hector,
This is not a bug. The Gibbs Reactor is very sensitive to initial estimates due to usage of newton's method for convergence. You need good initial estimates for the final mole flows, but you can also play with some convergence parameters like damping factor limits (default value for lower is 0.001 but a good one is 1E-20). It is always good to copy initial estimate values from outlet streams every time you change a condition in the reactor on in the material stream so newton's method can start the convergence process very close to the new solution point.
Try 1E-20 for the lower limit of the damping factor and copy initial estimates from the inlet stream, you'll see that it will converge for adiabatic mode and different pressure drop values.
I too have had the problem and found the solution used by Hector, i.e. set the temperatre and enjoy quick convergence. I then itterate on the temperature until the energy from the reactor is near zero. With that complete, I copy the "good initial estimates for the final mole flows" from the results and then have to sigificantly tighten the damping factor to get convergence as Daniel suggested. That final itteration is hit or miss as often the maximum number of itterations is exceded before convergence.
You can also play with the numerical derivative perturbation, default is 0.0001 but it can range from 0.1 to 1E-10 in some cases.
Great suggestions. They worked. Thank you.
Hector
On Fri, Apr 19, 2019 at 9:10 AM Daniel Medeiros danwbr@users.sourceforge.net wrote:
Related
Tickets (Inactive):
#234