Physics-Informed Neural Networks as Fast Surrogate Models for Electrochemical Flow Reactors

This paper introduces a physics-informed neural network (PINN) to model a transient two-dimensional electrochemical flow reactor, trained without labeled concentration data by embedding governing equations and conditions into a loss function. The PINN accurately predicts concentration fields across a broad operating domain, showing strong agreement with finite-difference solutions and achieving a 5.34 times faster inference speed. This demonstrates PINNs as accurate and efficient parametric surrogates for electrochemical transport problems, foundational for low-computational-cost digital-twin modeling.