Interpretable Machine Learning of Hydrogen Production from Reactive Molecular Dynamics Simulations of Polyethylene Oxidative Gasification
- 1Group of Applied Research in Advanced Materials for Sustainable Development, Faculty of Applied Sciences, Ton Duc Thang University, Ho Chi Minh City 756636, Vietnam
- 2Porelab, Department of Chemistry and Biomedical Science, Norwegian University of Science and Technology, NTNU, NO-7491 Trondheim, Norway
Abstract
Reactive molecular dynamics (ReaxFF MD) simulations of polyethylene (PE) oxidative gasification generate large trajectory datasets suited to machine learning (ML). We recast H2 production monitoring as a leakage-free forecasting task: given the simulated thermochemical state at time t, predict the time-weighted cumulative H2 production cumH2(t+h) a fixed horizon h=25 ps ahead. Data from 48 trajectories spanning six O2 loadings (1–900 molecules per 50 PE chains) and eight temperatures (1500–2900 K) yield 72,117 forecast pairs (45 simulations; three truncated trajectories excluded). Eight tabular regression models are benchmarked against persistence and physics-informed baselines under simulation-grouped nested cross-validation, a held-out test partition, temperature- and loading-holdout stress tests, and a paired ablation removing all H2-derived inputs. With the H2 history available, the best test model (random forest) attains R2=0.9999 at RMSE 121.8 molecules·ps, a 93.0% RMSE reduction relative to persistence, robust to unseen O2 loadings and largely robust to unseen temperatures. Without it, no model beats persistence in any evaluation: the best ML model (an artificial neural network) is 26.0% worse, and accuracy decreases substantially under temperature extrapolation. The ablation attributes 93.3–98.3% of the tree and kernel models’ RMSE reduction to the H2-derived features, quantifying how much forecasting skill derives from the target history rather than instantaneous observables.
Related
- Carbide Slag-Assisted Steam Gasification of Sewage Sludge for Hydrogen-Rich Syngas Production and Heavy Metal Stabilization
- Reactive Molecular Dynamics Simulations of Mechanochemical Ammonia Production
- Study on Gas Production and Heat Transfer Characteristics of Underground Coal Gasification with Groundwater-Related Operating and Reservoir Factor Influence
- Interfacial thermal conductance between a polyethylene glycol polymer chain and water: A molecular dynamics study
- Prospective multicenter study on patient-related outcomes of orbital floor reconstruction using ultrahigh-molecular-weight polyethylene
BibTeX
@article{truong2026interpretable,
title = {{Interpretable Machine Learning of Hydrogen Production from Reactive Molecular Dynamics Simulations of Polyethylene Oxidative Gasification}},
author = {Thi Be Ta Truong and Thuat T. Trinh},
journal = {International Journal of Molecular Sciences},
year = {2026},
month = oct,
doi = {10.3390/ijms27219435},
url = {https://www.mdpi.com/1422-0067/27/21/9435},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Interpretable Machine Learning of Hydrogen Production from Reactive Molecular Dynamics Simulations of Polyethylene Oxidative Gasification
AU - Thi Be Ta Truong
AU - Thuat T. Trinh
PY - 2026
DA - 2026/10/22
JO - International Journal of Molecular Sciences
DO - 10.3390/ijms27219435
UR - https://www.mdpi.com/1422-0067/27/21/9435
AB - Reactive molecular dynamics (ReaxFF MD) simulations of polyethylene (PE) oxidative gasification generate large trajectory datasets suited to machine learning (ML). We recast H2 production monitoring as a leakage-free forecasting task: given the simulated thermochemical state at time t, predict the time-weighted cumulative H2 production cumH2(t+h) a fixed horizon h=25 ps ahead. Data from 48 trajectories spanning six O2 loadings (1–900 molecules per 50 PE chains) and eight temperatures (1500–2900 K) yield 72,117 forecast pairs (45 simulations; three truncated trajectories excluded). Eight tabular regression models are benchmarked against persistence and physics-informed baselines under simulation-grouped nested cross-validation, a held-out test partition, temperature- and loading-holdout stress tests, and a paired ablation removing all H2-derived inputs. With the H2 history available, the best test model (random forest) attains R2=0.9999 at RMSE 121.8 molecules·ps, a 93.0% RMSE reduction relative to persistence, robust to unseen O2 loadings and largely robust to unseen temperatures. Without it, no model beats persistence in any evaluation: the best ML model (an artificial neural network) is 26.0% worse, and accuracy decreases substantially under temperature extrapolation. The ablation attributes 93.3–98.3% of the tree and kernel models’ RMSE reduction to the H2-derived features, quantifying how much forecasting skill derives from the target history rather than instantaneous observables.
SN - 1422-0067
ER -