Answers are grounded in this paper and cite their sources.
Tailoring the dimensional properties of carbon nanotubes (CNTs) during chemical vapor deposition is hindered by complex nonlinear interactions among catalyst formulations and operating conditions. This study aims to develop a highly accurate and interpretable predictive framework to map these complex multivariable relationships and precisely forecast carbon nanotube diameter. A comprehensive dataset comprising varied catalyst compositions and synthesis parameters was utilized to train a gradient boosting decision tree algorithm. To identify the optimal model architecture, the algorithm was independently coupled with four distinct hyperparameter tuning techniques including Gaussian Process Optimization (GPO), Evolution Strategies (ES), Bayesian Probability Improvement (BPI), and Batch Bayesian Optimization (BBO) within a rigorous cross-validation structure. Statistical evaluation demonstrated that the evolution strategy optimized model achieved the highest predictive accuracy and generalization capability, yielding a superior coefficient of determination and minimal mean squared error on unseen testing data. Conversely, alternative optimization methods exhibited varying susceptibilities to overfitting. Subsequent implementation of game-theoretic feature attribution revealed that molybdenum concentration acts as the primary determinant of diameter, followed closely by the mass fractions of active transition metals and specific spatial velocity parameters. The integration of evolutionary optimization with ensemble machine learning provides a highly reliable framework for catalyst design. This predictive capability, combined with mechanistic interpretability, offers a powerful computational approach to accelerate the targeted synthesis of nanomaterials by replacing heuristic trial and error with data-driven precision.
No takes yet. Share an insight, caveat, or question.
Almansour et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: