TY - JOUR
T1 - In-depth analysis of reaction kinetics parameters of phenolic resin using molecular dynamics and unsupervised machine learning approach
AU - Bhesania, Abhishek S.
AU - Kamboj, Parvesh
AU - Peddakotla, Sai Abhishek
AU - Kumar, Rakesh
N1 - Funding Information:
The authors acknowledge the financial support provided by the Indian Space Research Organization through the Grant No. STC/AE/ 2018033, and the Ministry of Human Resource Development (now Ministry of Education) through the Grant No. SPARC/2018- 2019/P1103/SL. We also acknowledge the National Supercomputing Mission (NSM) for providing computing resources of ’PARAM Sanganak’ at IIT Kanpur, which is implemented by C-DAC and supported by the Ministry of Electronics and Information Technology (MeitY) and Department of Science and Technology (DST), Government of India. We would also like to express gratitude to Prof. Alejandro Strachan, Purdue University for making the Hands-on Machine Learning and Data Science Training Workshop material available on Nanohub [51]. We extend special thanks to Michael Sakano for guiding with the Machine Learning part. Last but not least, we would like to thank Dr. Kishore K. Kammara for critically reviewing the current work.
Funding Information:
The authors acknowledge the financial support provided by the Indian Space Research Organization through the Grant No. STC/AE/ 2018033 , and the Ministry of Human Resource Development (now Ministry of Education) through the Grant No. SPARC/2018- 2019/P1103/SL. We also acknowledge the National Supercomputing Mission (NSM) for providing computing resources of ’PARAM Sanganak’ at IIT Kanpur, which is implemented by C-DAC and supported by the Ministry of Electronics and Information Technology (MeitY) and Department of Science and Technology (DST), Government of India.
Publisher Copyright:
© 2022 Elsevier B.V.
PY - 2022/4/15
Y1 - 2022/4/15
N2 - Reaction kinetics parameters of a material depend on energy dynamics based on breaking and formation of all the types of bonds in the system. While employing molecular dynamics simulation, it can become tedious and a complicated job to use all the bond information for extracting reaction kinetics parameters. With this understanding, in the current study, the use of an unsupervised machine learning technique is demonstrated for extracting the reaction kinetics parameters from the molecular dynamics simulation of an ablative material. Molecular dynamics simulations are performed on crosslinked and non-crosslinked polymers in temperature regime where they would undergo pyrolysis decomposition. Non-negative Matrix Factorization (NMF) technique is used to reduce the bonding environment, obtained during the simulations, to the concentration profiles of a few principal components. A comparative analysis performed with polymers having different degrees of crosslinking reveals that the activation energy reduces with increase in the degree of crosslinking. The effect of heating rate on the reaction kinetics of phenolic polymers during the pyrolysis simulation is investigated in detail. The assumption of chemical equilibrium between gases and porous solid domain is frequently made in continuum level thermal response solvers. It is unknown if this assumption significantly affects the calculated reaction kinetics parameters. To understand the same, a molecular dynamics simulation, which eliminates the generated gas molecules in a systematic manner throughout the pyrolysis process, is carried out. Furthermore, to demonstrate the usefulness of reaction kinetics parameters extracted after manifestation of chemical equilibrium at microscale level, a one-dimensional heat conduction analysis is performed. The results obtained by not considering the gas particles in reaction modeling agree well with experiments. At the end, a multiscale thermal response analysis is performed over an axi-symmetric geometry for which, a relation is derived between pyrolysis gas species and solid material density evolution from MD simulations. Based on the relation, the axi-symmetric domain is segregated into different regions for their contribution in changing pyrolysis gas composition by either adding or consuming gas species.
AB - Reaction kinetics parameters of a material depend on energy dynamics based on breaking and formation of all the types of bonds in the system. While employing molecular dynamics simulation, it can become tedious and a complicated job to use all the bond information for extracting reaction kinetics parameters. With this understanding, in the current study, the use of an unsupervised machine learning technique is demonstrated for extracting the reaction kinetics parameters from the molecular dynamics simulation of an ablative material. Molecular dynamics simulations are performed on crosslinked and non-crosslinked polymers in temperature regime where they would undergo pyrolysis decomposition. Non-negative Matrix Factorization (NMF) technique is used to reduce the bonding environment, obtained during the simulations, to the concentration profiles of a few principal components. A comparative analysis performed with polymers having different degrees of crosslinking reveals that the activation energy reduces with increase in the degree of crosslinking. The effect of heating rate on the reaction kinetics of phenolic polymers during the pyrolysis simulation is investigated in detail. The assumption of chemical equilibrium between gases and porous solid domain is frequently made in continuum level thermal response solvers. It is unknown if this assumption significantly affects the calculated reaction kinetics parameters. To understand the same, a molecular dynamics simulation, which eliminates the generated gas molecules in a systematic manner throughout the pyrolysis process, is carried out. Furthermore, to demonstrate the usefulness of reaction kinetics parameters extracted after manifestation of chemical equilibrium at microscale level, a one-dimensional heat conduction analysis is performed. The results obtained by not considering the gas particles in reaction modeling agree well with experiments. At the end, a multiscale thermal response analysis is performed over an axi-symmetric geometry for which, a relation is derived between pyrolysis gas species and solid material density evolution from MD simulations. Based on the relation, the axi-symmetric domain is segregated into different regions for their contribution in changing pyrolysis gas composition by either adding or consuming gas species.
UR - https://www.sciencedirect.com/science/article/abs/pii/S0927025622000271?via%3Dihub
U2 - 10.1016/j.commatsci.2022.111215
DO - 10.1016/j.commatsci.2022.111215
M3 - Article
SN - 0927-0256
VL - 206
JO - Computational Materials Science
JF - Computational Materials Science
M1 - 111215
ER -