Skip to content

Reaction centered reduction of RMG generated kinetic models

NICK VANDEWIELE edited this page Jul 5, 2016 · 1 revision

Reaction-centered reduction of RMG generated kinetic models

Authors: Nick M. Vandewiele, Chiamaka Obianyor, and William H. Green.

Usage

A command line interface to the reduction driver script is contained in rmgpy/reduction/main.py. It accepts four files:

  • input.py: RMG-Py input file containing the settings to evaluate state variables.
  • reduction_input.py: Reduction input file containing the target variables and associated error tolerances to allow in the reduced model
  • chem_annotated.inp: the reaction mechanism to reduce.
  • species_dictionary.txt: the species dictionary associated with the reaction mechanism to reduce.

Please visit $RMG/examples/reduction for examples of input file.

The script is called as follows:

python $RMG/rmgpy/reduction/main.py input.py reduction_input.py chemkin/chem_annotated.inp chemkin/species_dictionary.txt

If everything goes as planned, a chem_reduced.inp is generated containing the reduced mechanism. In addition, a number of files chem_reduced_{i}.inp are created and correspond to the intermediate reduced mechanisms. They can replace the final reduced model, in case the reduction algorithm does not terminate normally.

Background

The reduction algorithm computes the ratio of species reaction rate (rij) to the total rate of formation/consumption (Ri) of all species i, and compares this ratio to a tolerance (epsilon), with values of epsilon between 0 and 1. If the ratio of a reaction is greater than epsilon it is deemed important for the species in question. When a reaction is not important for a single species, at any given time between t=0 and the user-defined end time, then it is deemed unimportant for the given system. As a result, the reaction is removed from the mechanism.

The value of epsilon is determined by an optimization algorithm that attempts to reduce the model as much as possible given the constraints of the user-defined target variables. A logarithmic bisection optimization algorithm is used to provide guesses for the value of epsilon based on the two previous guesses that undershoot and overshoot the user-defined relative deviation of the target variables

A value of 5% for the relative deviation of the target variable implies that the mole fraction of the target variable at the end time of the batch reactor simulation as computed by the reduced mechanism may deviate up to 5% w.r.t. to the mole fraction of the target variable at the end time of the batch reactor simulation as computed by the full mechanism.

Clone this wiki locally