Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Graph Machine Learning for Design of High-Octane Fuels

Jan G. Rittig, Martin Ritzert, Artur M. Schweidtmann, Stefanie Winkler, Jana M. Weber, Philipp Morsch, K. Alexander Heufer, Martin Grohe, Alexander Mitsos, Manuel Dahmen

TL;DR

The paper tackles the design of high-octane fuels by proposing a modular graph-ML CAMD framework that combines generative graph-ML models, graph neural networks for ignition-property prediction, and optimization to maximize $p= ext{RON}+ ext{OS}$ in a continuous molecular space. It analyzes three generator types (JT-VAE, MHG-VAE, MolGAN) and two optimizers (Bayesian optimization and genetic algorithms), augmented by an applicability-domain mechanism based on one-class SVMs to curb unreliable extrapolations. The study demonstrates that the framework can identify well-known octane boosters (e.g., MTBE, ETBE) and new candidates (notably 2,2-DMP), while also experimentally evaluating a selected candidate to reveal limitations due to data scarcity and model extrapolation. The results underscore the potential of graph-ML CAMD for fuel design, the importance of experimental validation, and the need for expanded ignition-property datasets to improve predictive accuracy in data-constrained settings. The framework is modular and extensible to additional properties and CAMD applications beyond fuels.

Abstract

Fuels with high-knock resistance enable modern spark-ignition engines to achieve high efficiency and thus low CO2 emissions. Identification of molecules with desired autoignition properties indicated by a high research octane number and a high octane sensitivity is therefore of great practical relevance and can be supported by computer-aided molecular design (CAMD). Recent developments in the field of graph machine learning (graph-ML) provide novel, promising tools for CAMD. We propose a modular graph-ML CAMD framework that integrates generative graph-ML models with graph neural networks and optimization, enabling the design of molecules with desired ignition properties in a continuous molecular space. In particular, we explore the potential of Bayesian optimization and genetic algorithms in combination with generative graph-ML models. The graph-ML CAMD framework successfully identifies well-established high-octane components. It also suggests new candidates, one of which we experimentally investigate and use to illustrate the need for further auto-ignition training data.

Graph Machine Learning for Design of High-Octane Fuels

TL;DR

The paper tackles the design of high-octane fuels by proposing a modular graph-ML CAMD framework that combines generative graph-ML models, graph neural networks for ignition-property prediction, and optimization to maximize in a continuous molecular space. It analyzes three generator types (JT-VAE, MHG-VAE, MolGAN) and two optimizers (Bayesian optimization and genetic algorithms), augmented by an applicability-domain mechanism based on one-class SVMs to curb unreliable extrapolations. The study demonstrates that the framework can identify well-known octane boosters (e.g., MTBE, ETBE) and new candidates (notably 2,2-DMP), while also experimentally evaluating a selected candidate to reveal limitations due to data scarcity and model extrapolation. The results underscore the potential of graph-ML CAMD for fuel design, the importance of experimental validation, and the need for expanded ignition-property datasets to improve predictive accuracy in data-constrained settings. The framework is modular and extensible to additional properties and CAMD applications beyond fuels.

Abstract

Fuels with high-knock resistance enable modern spark-ignition engines to achieve high efficiency and thus low CO2 emissions. Identification of molecules with desired autoignition properties indicated by a high research octane number and a high octane sensitivity is therefore of great practical relevance and can be supported by computer-aided molecular design (CAMD). Recent developments in the field of graph machine learning (graph-ML) provide novel, promising tools for CAMD. We propose a modular graph-ML CAMD framework that integrates generative graph-ML models with graph neural networks and optimization, enabling the design of molecules with desired ignition properties in a continuous molecular space. In particular, we explore the potential of Bayesian optimization and genetic algorithms in combination with generative graph-ML models. The graph-ML CAMD framework successfully identifies well-established high-octane components. It also suggests new candidates, one of which we experimentally investigate and use to illustrate the need for further auto-ignition training data.
Paper Structure (18 sections, 6 equations, 9 figures, 2 tables)

This paper contains 18 sections, 6 equations, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Preliminaries of graph machine learning
  3. Molecular graph
  4. Generative models
  5. Graph-based property prediction
  6. Graph-ML CAMD framework for high-octane fuels
  7. Molecule generation
  8. Property prediction
  9. Optimization
  10. Applicability domain
  11. Implementation and hyperparameters
  12. Results and discussion
  13. CAMD results
  14. Discussion of top candidates
  15. Promising classes of molecules
  16. ...and 3 more sections

Figures (9)

  • Figure 1: Schematic overview of the modular graph-ML CAMD framework for identification of high-octane fuels.
  • Figure 2: Schematic structure of (a) VAEs and (b) GANs.
  • Figure 3: Schematic structure of a graph neural network for molecular property prediction.
  • Figure 4: Detailed overview of the modular graph-ML CAMD framework for identification of high-octane fuels including methods for the individual modules.
  • Figure 5: Adapted MolGAN for high-octane fuels, modified from DeCao2018. The reward network is coupled with our GNN Schweidtmann2020_GNNs for predicting RON and OS values.
  • ...and 4 more figures