| NSF Org: |
CBET Division of Chemical, Bioengineering, Environmental, and Transport Systems |
| Recipient: |
|
| Initial Amendment Date: | March 9, 2018 |
| Latest Amendment Date: | March 9, 2018 |
| Award Number: | 1751720 |
| Award Instrument: | Standard Grant |
| Program Manager: |
Harsha Chelliah
hchellia@nsf.gov (703)292-7281 CBET Division of Chemical, Bioengineering, Environmental, and Transport Systems ENG Directorate for Engineering |
| Start Date: | July 1, 2018 |
| End Date: | June 30, 2024 (Estimated) |
| Total Intended Award Amount: | $503,888.00 |
| Total Awarded Amount to Date: | $503,888.00 |
| Funds Obligated to Date: |
|
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
360 HUNTINGTON AVE BOSTON MA US 02115-5005 (617)373-5600 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
360 Huntington Ave Boston MA US 02115-5005 |
| Primary Place of
Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): | CFS-Combustion & Fire Systems |
| Primary Program Source: |
|
| Program Reference Code(s): |
|
| Program Element Code(s): |
|
| Award Agency Code: | 4900 |
| Fund Agency Code: | 4900 |
| Assistance Listing Number(s): | 47.041 |
ABSTRACT
Halogenated hydrocarbons (HHCs) are widely used as both refrigerants and fire suppressants. Driven by environmental and economic considerations, there is rapid innovation in the industry, but the next generation of HHC compounds raise fire safety concerns. Predicting the combustion behavior of these novel HHCs earlier in the design process will save much time, effort, and expense. The chemical kinetic models for describing HHC combustion are highly complex, comprising thousands of elementary reactions involving hundreds of chemical species. To effectively predict these combustion behaviors, we must automate the construction of kinetic models. This project will use a computational approach known as machine learning to help model these complex reacting systems. This breakthrough will enable us to develop an automated reaction mechanism generation tool to create detailed kinetic models for combustion of HHCs. The methodology proposed in this work are not only novel and necessary, but will be widely applicable in other aspects of automated mechanism generation. The integrated educational objective of this CAREER project is to develop a series of computational modules teaching students to solve problems throughout their chemical engineering curriculum.
The research approach is to extend and apply automated Reaction Mechanism Generator (RMG) software to create detailed kinetic models for combustion of any mix of hydrocarbons containing any combination of halogen atoms. Optimized decision-tree and novel convolutional neural network algorithms from the field of machine learning will be extended to enable the necessary restructuring of parameter estimation codes. Quantum chemistry calculations will be automated to supplement literature searches to generate the necessary training data. The model-generating tool will be validated against available experimental data from key example compounds, and used to explain the remarkable combustion behavior of these compounds. The educational program is aligned with the research, developing a series of computational modules that will be integrated into existing classes. These modules will teach students to use Python and SciPy to solve chemical engineering problems. The integration of teaching modules for scientific computing throughout the undergraduate chemical engineering curriculum will help prepare a generation of graduate engineers for a workplace in which data analysis, processing, and computation are increasingly important.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
PUBLICATIONS PRODUCED AS A RESULT OF THIS RESEARCH
Note:
When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external
site maintained by the publisher. Some full text articles may not yet be available without a
charge during the embargo (administrative interval).
Some links on this page may take you to non-federal websites. Their policies may differ from
this site.
PROJECT OUTCOMES REPORT
Disclaimer
This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.
This NSF CAREER project aimed to harness machine learning and computational chemistry to revolutionize the modeling of combustion processes for halogenated hydrocarbons (HHCs), critical components in refrigerants and fire suppressants. The work has advanced automated reaction mechanism generation, enhanced scientific understanding of HHC chemistry, and contributed to chemical engineering education.
Intellectual Merit.
Automated quantum calculations: We developed a quantum chemistry workflow to calculate thermochemical properties (enthalpy, entropy, heat capacity) of halogenated species, and used it to calculate approximately 67,000 unique molecules (at wB97X-D3 level) and about 17,000 molecules with a more accurate workflow (G4 energies with 1D hindered rotors based on DFT). This work presents the most accurate thermochemistry for many halocarbons to date.
Machine Learning for Parameter Estimation: Using the computed thermochemistry, we developed new Benson functional group values, and long-range interactions, to allow prediction of novel halocarbons, and added these to RMG. We also used the data to develop, train, and test new Graph Neural Networks (GNNs) for predicting chemical properties, exploring transfer learning, fine-tuning, and attention models, with many GNN architectures.
Extending RMG: We extended the open-source Reaction Mechanism Generator (RMG) software by adding atom types for fluorine, chlorine, and bromine, and created new reaction families and extended existing families to include their reactions. We calculated transition states for over 1000 elementary reactions, and properties for 17,000 molecules, and used these data to improve our parameter estimation routines. In addition to the thermochemistry and kinetics parameters, we developed estimators for vibrational frequencies (for densities of states needed for pressure-dependent kinetics) and transport properties (needed for flame speed predictions).
Validation through Modeling and Simulations: Using RMG's new reaction families, Benson groups, and thermochemical and kinetic data, we generated kinetic models for combustion of many halogenated hydrocarbons, including methyl chloride, CF3Br, CH2=CBrCF3 (2-BTP), and blends. Using these models to predict laminar flame speeds and comparing with measurements from the literature and other kinetic models, confirms the effectiveness of the automated RMG approach. RMG is able to discover the flame inhibition cycle of bromine radicals, as well as the "fuel-effect" of 2-BTP when it is added to lean flames.
Extension to PFAS chemistry: although not part of the original project scope, the new features make RMG helpful for predicting reactions of per- and poly-fluoroalkyl substances (PFAS) which are a special case of halogenated hydrocarbon. We have further extended RMG with data specific to these compounds and used it to help model their incineration.
Broader Impacts.
Enhancing Fire Safety and Sustainability: The project provides tools for safer, faster evaluation of next-generation refrigerants and fire suppressants, helping industry meet environmental regulations while improving fire safety. It will also help in the urgent need to address PFAS pollutants.
Open Science Contributions: We made datasets, ML models, and software tools publicly available on GitHub and the RMG website, fostering global collaboration and accessibility in combustion research.
Educational Innovations: We integrated Python-based computational chemistry into the undergraduate and graduate chemical engineering curriculum, enhancing computational literacy. We developed and shared virtual labs and teaching materials, facilitating modernized education for chemical engineering students across multiple institutions.
Community Engagement: We co-organized workshops for early-career researchers and led sessions on integrating computational tools into education, establishing a community for sharing combustion-related teaching resources.
This project has successfully blended ML, quantum chemistry, and computational tools to address complex problems in HHC combustion while contributing to education and open science. The results promise significant impacts on fire safety, sustainable chemical design, and chemical engineering education.
Last Modified: 11/17/2024
Modified by: Richard H West
Please report errors in award information by writing to: awardsearch@nsf.gov.
