Hexadecane mechanisms: Comparison of hand-generated and automatically generated with pathways, Mersin, I. E., Blurock, E., Soyhan, H. S., & Konnov, A., Fuel, 115, 132-144 (2014). DOI:10.1016/j.fuel.2013.06.055
This paper represents an non-trivial large application of the software system REACTION (Blurock, 1995) for the combustion of large molecules (representing fuels in energy and transport) from over 500 reaction classes and 10 reaction pathways, representing reactivity. The complex chemistry is represented as a generated mechanism with 2176 molecules and 7269 (reversible) reactions. In addition to the simulation of this complex chemistry, this paper represented the success of systematically reproducing a similar mechanism that was produced solely by hand, proving the techniques within the REACTION software system simulate human reasoning in this regard. The total system can be considered a GUI based database management and expert system.
Databases of complex data stuctures representing the varied types of data that was needed to simulate the process is the heart of this system. There is numerical data used to set up numerical constants for each of the entities of the modeling. There is graphical data representing the molecular structures and the reactions (viewed as transformations) of the molecular structures.
Graph Theory plays a large role in this modeling process in not only the 2D graph representation of the molecules, but also in the network representation of the reactions between these molecules. Molecules are modeled as Lewis structures, basically atom configurations, including electronic, lone pair and resonance information, as nodes and the bonding between these nodes, including multiple bonds (multiple bonds can also be a part of the node representation). Graph isomorphism plays a large role is deciding not only whether molecules are equivalent, but also determine the existence of substructures, which is key to deciding the molecules reactivity. The entire reaction network is modeled in several ways, depending on need. One type representation is molecules as nodes and the directed bonding representing the connection between reactants and products, i.e. a molecule, as reactant, is connected to all the product molecules. This is useful in interpreting the concentration flux during reactivity. This representation is also key to reduction of complex mechanisms by determining if two isomers, molecules with the same atomic make up, have the same reactivity. If so, they can be combined into a pseudo molecule.
Algorithms and data structures play a large role in the sense that the REACTION system, on which this paper is based, is an expert database management system. To produce the complex reactivity of thousands of reaction and molecules a multitude of different algorithms and data structures needed to be used to not only generate the network of reactions, but also to analyse the results. The REACTION software, built upon the machine learning and fundamental data structure system ANALYSIS, is a composite of several language platforms, C, C++, JAVA and web technologies (including web-services). The REACTION software is not only used for detailed mechanism generation(heavily based on graphical techniques), but also for its analysis (where statistics, machine learning and other software techniques are employed).
The fundamental philosophy of the ANALYSIS system is to be able to combine different methods in an effective way. To this end, an “Algorithm” and “Goal” class was used to manage and define complex combinations of algorithms.
Artificial Intelligence plays a role in that this paper gave systematic proof that a complex mechanism can be automatically generating using the same reasoning as a modeler trying to accomplish the same task by hand. The key is to replace the (usual) combinatoric method, which humans do not consistently do, with the use a reactive pathways of reaction classes, which is acutally how human modelers approach the problem. The extensive knowledge base, and its management and use can be considered an expert system.
Other key publications in this area:
1. A book outlining the general principles of detailed mechanisms in combustion:
Development of detailed chemical kinetic models for cleaner combustion, Editors: F. Battin-Leclerc, J. M. Simmie, E. Blurock, ISBN: 978-1-4471-5306-1 (Print) 978-1-4471-5307-8 (Online), Springer, 2013
Chapter 1: Introduction, Battin-Leclerc, Frédérique; Blurock, Edward S.; Simmie, John M.; Alzueta, Maria U.; Tomlin, Alison S.; Olzmann, Matthias
Chapter 2: Modeling Combustion with Detailed Kinetic Mechanisms, Blurock, Edward; Battin-Leclerc, Frédérique
Chapter 3: Automatic Generation of Detailed Mechanisms, Blurock, Edward; Battin-Leclerc, Frédérique; Faravelli, Tiziano; Green, William H.
2. Review article:
Towards cleaner internal combustion engines through groundbreaking detailed chemical kinetic models, Frédérique Battin-Leclerc, Edward Blurock, René Fournet a Pierre-Alexandre Glaude, Olivier Herbinet and Baptiste Sirjean, Chemical Society Reviews, 40 4762-4782 (2011).
3. Another application
Automatic Generation of a Detailed Mechanism for the Oxidation of n-Decane, Moréac, G., Blurock, Edward S., and Mauss, F.; Combustion Science Technology 178 (10-11): 2025-2038 (2006).
4. The original paper with a description of the fundamental method:
Reaction: System for Modeling Chemical Reactions, Blurock, Edward S.; J. Chem. Inf. Comput. Sci., 35, 607-616 (1995).
Detailed Mechanism Generation 1: Generalized Reactive Properties as Reaction Class Substructures, Blurock, Edward S.; Journal of Chemical Information and Computer Sciences, 44 (4) 1336-1347 (2004).
Detailed Mechanism Generation 2: Aldehydes, Ketones and Olefins, Blurock, Edward S.; Journal of Chemical Information and Computer Sciences, 44 (4) 1348-1357 (2004).