A machine learning based approach to reaction rate estimation

Matthew S. Johnson,William H. Green
DOI: https://doi.org/10.1039/d3re00684k
IF: 5.2002
2024-02-23
Reaction Chemistry & Engineering
Abstract:Chemical kinetic models are vital to accurately predicting phenomena in a wide variety of fields from combustion to atmospheric chemistry to electrochemistry. However, building an accurate chemical kinetic model requires the efficient and accurate estimation of many reaction rate coefficients for many reaction classes with highly variable amounts of available training data. Current techniques for fast automatic rate estimation tend to be poorly optimized and tedious to maintain and extend. We have developed a machine learning algorithm for automatically training subgraph isomorphic decision trees (SIDT) to predict rate coefficients for arbitrary reaction types. This method is fully automatic, scalable to virtually any dataset size, human readable, can incorporate qualitative chemical knowledge from experts and provides detailed uncertainty information for estimates. The accuracy of the algorithm is tested against the state of the art rate rules scheme in the RMG-database for five selected reaction families. The SIDT method is shown to significantly improve estimation accuracy across all reaction families and considered statistics. The estimator uncertainty estimates are validated against actual errors.
chemistry, multidisciplinary,engineering, chemical
What problem does this paper attempt to address?
The paper attempts to address the problem of quickly and accurately estimating reaction rates in chemical kinetics models. Specifically: - **Background Problem**: Chemical kinetics models are crucial for predicting phenomena in various fields, from combustion to atmospheric chemistry to electrochemistry. However, constructing an accurate chemical kinetics model requires efficiently and accurately estimating many different types of reaction rate coefficients, and the existing data varies greatly. - **Limitations of Existing Methods**: Current techniques for quickly and automatically estimating reaction rates are often poorly optimized and difficult to maintain and extend. Additionally, while quantum chemical calculations can estimate reaction rates with high precision, they are slow and not yet fully automated. - **Proposed Method**: The authors developed a machine learning-based method, the Subgraph Isomorphic Decision Tree (SIDT), for automatically training to predict rate coefficients for any reaction type. This method has the following advantages: - Fully automated - Scalable to any dataset size - Highly readable - Capable of incorporating expert qualitative chemical knowledge - Provides detailed uncertainty information - **Validation and Improvement**: The method was tested for its accuracy on 5 selected reaction families and demonstrated significantly better performance than existing state estimation methods. Additionally, the relationship between the estimated uncertainty and the actual error was validated.