(582h) Deconstructing Detailed Reaction Mechanisms to Identify Species and Mine for Data | AIChE

(582h) Deconstructing Detailed Reaction Mechanisms to Identify Species and Mine for Data


West, R. H. - Presenter, Northeastern University
Seyedzadeh Khanshan, F. - Presenter, Northeastern University
Bhoorasingh, P. L. - Presenter, Northeastern University
Slakman, B. L. - Presenter, Northeastern University

Detailed kinetic models have become integral to combustion research over the last 40 years. The latest models can explain many complicated phenomena and offer accurate simulations for novel engines and fuels.  These models can be very large (eg. the LLNL model for 2-methylalkanes has over 7,000 species and 30,000 reactions) and there are now dozens of these large published models. Unfortunately, these ever-proliferating detailed kinetic models are usually incompatible and inconsistent, are seldom compared directly, and often contain undetected mistakes. 

The usual publication format remains a “Chemkin file” for use in compatible simulation tools. This format, devised in the 1970’s when input was limited by the width of 80-column punch-cards, forces model-builders to abbreviate species’ names, thereby losing their chemical identity, and to discard other metadata. The main challenge in comparing these models is in recognizing, for example, that the name “C3KET12” in one model represents 1-hydroperoxypropan-2-one, which another research group may have named “CH3COCH2O2H” in a different model.

The clues available to determine the molecule corresponding to a given nickname are: the name itself (often cryptic and sometimes misleading), thermochemical data (usually estimated), and a list of reactions (usually incomplete) in which the species participates, connecting it to other species in the model (that are also initially unknown). 

We have developed an algorithm and tools to facilitate the identification of chemical species in a kinetic model “Chemkin file”, and then to allow comparison of the models. The tools are built on top of the open source Python version of Reaction Mechanism Generator software (RMG-Py)1, originally designed to create detailed kinetic models of its own. By comparing reported reactions with its own predicted reactions between already-identified species, it is able to propose new species and eliminate unlikely matches. A web-based user interface allows a team of humans to quickly review the evidence and confirm or block the proposed matches.

We will present how the tool works, opportunities for improvement, and some findings from analyzing recent publications in the combustion kinetics literature. This material is based upon work supported by the National Science Foundation under Grant No. 1403171.

Green, W.H., Allen, J.W. et al. RMG – Reaction Mechanism Generator, Python Version. http://rmg.mit.edu, 2013.