Forward physics-informed neural networks for catalytic CO2 methanation via isothermal fixed-bed reactor

Abstract

Abstract We developed physics-informed neural networks (PINNs) to solve an isothermal fixed-bed (IFB) model for catalytic CO 2 methanation. The PINN is composed of a feed-forward artificial neural network (FF-ANN) with two inputs and physics-informed constraints for governing equations, boundary conditions, initial conditions, and nonlinear reaction kinetics. The forward PINN showed excellent extrapolation performance for the IFB model. The calculation speed of the PINN surrogate model is faster significantly than a stiff ODE numerical solver. These results suggest that forward PINNs can be used as a surrogate model for chemical reaction kinetics.