ML-based Method for Solving the Microkinetic Model of Fischer-Tropsch Synthesis with Varying Catalyst/Reactor Parameters

2025/03/28 by Demchuk, Taras, Nikolaienko, Tymofii, Panda, Aniruddha +3
#Chemical Physics (physics.chem-ph) #Disordered Systems and Neural Networks (cond-mat.dis-nn) #FOS: Physical sciences

paper · doi:10.48550/arxiv.2503.22304

Abstract

This study introduces a physics-informed machine learning framework to accelerate the computation of the microkinetic model of Fischer-Tropsch synthesis. A neural network, trained within the NVIDIA Modulus framework, approximates the fraction of vacant catalytic sites with high accuracy. The combination of implicit differentiation and the Newton-Raphson method enhances derivative calculations, ensuring physical consistency. Computational efficiency improves significantly, with speedups up to 10⁴ times on a GPU. This versatile methodology generalizes across catalysts and reactors, offering a robust tool for chemical engineering applications, including model approximation and catalyst parameter fitting from experimental data.

Related

• Accurate and Fast Fischer-Tropsch Reaction Microkinetics using PINNs
• Simplest Mechanism Builder Algorithm (SiMBA): An Automated Microkinetic Model Discovery Tool
• Data Driven Reaction Mechanism Estimation via Transient Kinetics and Machine Learning
• Machine Learning in Chemical Kinetics: Predictions, Mechanistic Analysis, and Reaction Optimization.
• Physics-informed neural networks need a physicist to be accurate: the case of mass and heat transport in Fischer-Tropsch catalyst particles