Hybrid soft sensor modeling for bisphenol‐A synthesis reaction process

Abstract

AbstractTo online estimate the concentrations of key components at the outlet of bisphenol‐A (BPA) synthesis reactor, a hybrid modeling approach is proposed in this paper. First, the simplified mechanism model for the synthesis of BPA in an industrial fixed‐bed reactor is deduced on the basis of mass and energy balance. Second, considering that the initial concentrations of reactants at the inlet of BPA synthesis reactor cannot be measured in real‐time, the modified stochastic gradient boosting‐Gaussian process regression approach previously proposed by the authors is employed to online estimate the initial concentrations of reactants. Furthermore, modified stochastic gradient boosting‐Gaussian process regression method is also adopted to model the nonlinear influence of the reaction factors ( i.e., initial concentrations of reactants and inlet temperature) on the reaction rate constants in the simplified mechanism model. Finally, the hybrid modeling approach for BPA synthesis process is achieved by integrating the simplified mechanism model with the estimation models of initial concentrations and reaction rate constants. The feasibility and validity of the proposed hybrid model is verified through an industrial BPA synthesis process. On the basis of the hybrid model, the sensitivity analyses of the initial concentrations and inlet temperature are conducted, which can provide theoretical guidance for the operation optimization study in future work.