A unified Physics-Informed Neural Operator (PINO) framework was developed to model coupled transport and reaction systems by combining spectral operator learning with physics information. The approach enabled both forward prediction of concentration fields and inverse estimation of kinetic and transport parameters. The forward PINO accurately reproduced nonlinear transient diffusion-reaction concentration fields with an average relative deviation below 0.002. The backward PINO identified intrinsic Thiele modulus and effective factors from steady-state concentration profiles with a mean relative deviation of 0.055 from limited experimental-like data. In a convective reactive flow, it inferred the reaction rate constant and reaction orders from limited data while substantially reducing computational cost by approximately 95% at a comparable prediction accuracy relative to traditional CFD-based optimization. These results demonstrated that PINO provided a scalable and physically consistent tool for reactor modeling and catalyst characterization, with potential extensions to multireaction and multiphase systems.
Zhang et al. (Mon,) studied this question.