Sparse Surrogate Modelling for Combustion Chemistry via Time-Lag Autoencoders and Gradient-Based Clustering

Key Points

• Surrogate modelling enhances prediction accuracy in combustion chemistry, optimizing the process significantly.
• Key evidence shows that integrating sparse time-lag autoencoders can reduce dimensionality by over 50% in data sets.
• Analysis utilizes gradient-based clustering techniques to group data efficiently, enabling robust surrogate models.
• Findings may enable more precise simulations of combustion systems, though practical applications need further validation.