Towards Foundation Models for the Industrial Forecasting of Chemical Kinetics

TL;DR

The paper tackles forecasting stiff chemical kinetics within CFD using Scientific Machine Learning to reduce online solve costs. It proposes PatchTSMixer, an MLP-Mixer-based time-series model, and evaluates it on the ROBER benchmark with LSODA-generated data, using a context length $H=512$ and horizon $h=100$. Results show a mean relative error of $0.0166\%$ over 100-step forecasts, indicating strong alignment with the stiff-ODE ground truth and supporting industrial relevance. The study also notes robustness challenges in dynamic extrapolation and suggests hyperparameter tuning and architectural refinements to enable PatchTSMixer as a base for time-series foundation models in chemical kinetics.

Abstract

Scientific Machine Learning is transforming traditional engineering industries by enhancing the efficiency of existing technologies and accelerating innovation, particularly in modeling chemical reactions. Despite recent advancements, the issue of solving stiff chemically reacting problems within computational fluid dynamics remains a significant issue. In this study we propose a novel approach utilizing a multi-layer-perceptron mixer architecture (MLP-Mixer) to model the time-series of stiff chemical kinetics. We evaluate this method using the ROBER system, a benchmark model in chemical kinetics, to compare its performance with traditional numerical techniques. This study provides insight into the industrial utility of the recently developed MLP-Mixer architecture to model chemical kinetics and provides motivation for such neural architecture to be used as a base for time-series foundation models.

Figures (3)

• Figure 1: Schematic of the MLP-Mixer method PatchTSMixer.
• Figure 2: Batchwise extrapolation using PatchTSMixer. The variables were normalized in order to enhance the visualization.
• Figure 3: The batchwise extrapolation performed for two slightly different test cases.