Stoichiometry Driven cPINN Modeling of Phenol Synthesis: Application in Enzyme Mediated, Biosynthetic Pathways, Bioengineering and Pharmaceutical Sciences

Abstract

This study introduces a causal Physics Informed Neural Network architecture to model and predict the chemical kinetics of phenol synthesis based on the alkaline fusion and acidification reaction mechanism. The reaction network involves sodium benzenesulfonate C₆H₅SO₃Na, sodium hydroxide NaOH, sodium phenoxide C₆H₅ONa, sodium sulfite Na₂SO₃, hydrochloric acid HCl, phenol C₆H₅OH, water H₂O, and sodium chloride NaCl, and is formulated as a nonlinear system of ordinary differential equations describing the time dependent evolution of all chemical species. Physical laws and stoichiometric constraints are embedded directly into the loss function, enabling stable and accurate learning of concentration profiles and kinetic parameters for stiff and strongly coupled reaction dynamics. The learned parameters give us smooth convergence on different initial conditions and different time interval, the time interval that we takes are $t\in [0,250]$, $[0,500]$, and $[0,750]$. Overall, the cPINN method is data efficient and scalable, which provides a reliable tool for stoichiometric analysis, stability assessment of chemicals, parameter estimation and kinetic modeling of phenol synthesis and related nonlinear chemical reaction systems.