Phase optimized skeletal mechanisms for engine simulations
(2010) In Combustion Theory and Modelling 14(3). p.295-313- Abstract
- Adaptive chemistry is based on the principle that instead of having one comprehensive model describing the entire range of chemical source term space (typically parameters related to temperature, pressure and species concentrations), a set of computationally simpler models are used, each describing a local region (in multidimensional space) or phases (in zero-dimensional space). In this work, an adaptive chemistry method based on phase optimized skeletal mechanisms (POSM) is applied to a 96 species n-heptane-isooctane mechanism within a two-zone zero-dimensional stochastic reactor model (SRM) for an spark-ignition (SI) Engine. Two models differing only in the extent of reduction in the phase mechanism, gave speed-up factors of 2.7 and 10.... (More)
- Adaptive chemistry is based on the principle that instead of having one comprehensive model describing the entire range of chemical source term space (typically parameters related to temperature, pressure and species concentrations), a set of computationally simpler models are used, each describing a local region (in multidimensional space) or phases (in zero-dimensional space). In this work, an adaptive chemistry method based on phase optimized skeletal mechanisms (POSM) is applied to a 96 species n-heptane-isooctane mechanism within a two-zone zero-dimensional stochastic reactor model (SRM) for an spark-ignition (SI) Engine. Two models differing only in the extent of reduction in the phase mechanism, gave speed-up factors of 2.7 and 10. The novelty and emphasis of this study is the use of machine learning techniques to decide where the phases are and to produce a usable phase recognition. The combustion process is automatically divided up into an 'optimal' set of phases through machine learning clustering based on fuzzy logic predicates involving a necessity parameter (a measure giving an indication whether a species should be included in the mechanism or not). The mechanism of each phase is reduced from the full mechanism based on this necessity parameter with respect to the conditions of that phase. The algorithm to decide which phase the process is in is automatically determined by another machine learning method that produces decision trees. The decision tree is made up of asking whether the mass fraction values were above or below given values. Two POSM studies were done, a conservative POSM where the species in each phase are eliminated based on a necessity parameter threshold (speed-up 2.7) and a further reduced POSM where each phase was further reduced by hand (speed-up 10). The automated techniques of determining the phases and for creating the decision tree are very general and are not limited to the parameter choices of this paper. There is also no fundamental limit as to the size of the original detailed mechanism. The interfacing to include POSM in an application does not differ significantly from using the original detailed mechanism. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/1658000
- author
- Blurock, Edward LU ; Tunér, Martin LU and Mauss, Fabian
- organization
- publishing date
- 2010
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- engine, mechanism, chemical kinetics, Adaptive chemistry, reduction, simulation
- in
- Combustion Theory and Modelling
- volume
- 14
- issue
- 3
- pages
- 295 - 313
- publisher
- Taylor & Francis
- external identifiers
-
- wos:000279632800001
- scopus:77954303907
- ISSN
- 1364-7830
- DOI
- 10.1080/13647830.2010.483018
- language
- English
- LU publication?
- yes
- id
- 71103871-823a-4a0b-a4d3-440432320c1e (old id 1658000)
- date added to LUP
- 2016-04-01 10:13:02
- date last changed
- 2022-03-04 17:17:17
@article{71103871-823a-4a0b-a4d3-440432320c1e,
abstract = {{Adaptive chemistry is based on the principle that instead of having one comprehensive model describing the entire range of chemical source term space (typically parameters related to temperature, pressure and species concentrations), a set of computationally simpler models are used, each describing a local region (in multidimensional space) or phases (in zero-dimensional space). In this work, an adaptive chemistry method based on phase optimized skeletal mechanisms (POSM) is applied to a 96 species n-heptane-isooctane mechanism within a two-zone zero-dimensional stochastic reactor model (SRM) for an spark-ignition (SI) Engine. Two models differing only in the extent of reduction in the phase mechanism, gave speed-up factors of 2.7 and 10. The novelty and emphasis of this study is the use of machine learning techniques to decide where the phases are and to produce a usable phase recognition. The combustion process is automatically divided up into an 'optimal' set of phases through machine learning clustering based on fuzzy logic predicates involving a necessity parameter (a measure giving an indication whether a species should be included in the mechanism or not). The mechanism of each phase is reduced from the full mechanism based on this necessity parameter with respect to the conditions of that phase. The algorithm to decide which phase the process is in is automatically determined by another machine learning method that produces decision trees. The decision tree is made up of asking whether the mass fraction values were above or below given values. Two POSM studies were done, a conservative POSM where the species in each phase are eliminated based on a necessity parameter threshold (speed-up 2.7) and a further reduced POSM where each phase was further reduced by hand (speed-up 10). The automated techniques of determining the phases and for creating the decision tree are very general and are not limited to the parameter choices of this paper. There is also no fundamental limit as to the size of the original detailed mechanism. The interfacing to include POSM in an application does not differ significantly from using the original detailed mechanism.}},
author = {{Blurock, Edward and Tunér, Martin and Mauss, Fabian}},
issn = {{1364-7830}},
keywords = {{engine; mechanism; chemical kinetics; Adaptive chemistry; reduction; simulation}},
language = {{eng}},
number = {{3}},
pages = {{295--313}},
publisher = {{Taylor & Francis}},
series = {{Combustion Theory and Modelling}},
title = {{Phase optimized skeletal mechanisms for engine simulations}},
url = {{http://dx.doi.org/10.1080/13647830.2010.483018}},
doi = {{10.1080/13647830.2010.483018}},
volume = {{14}},
year = {{2010}},
}