Mass transfer study of water deoxygenation in a rotor–stator reactor based on principal component regression method

Abstract

This study investigated mass transfer characteristics in a rotor–stator reactor (RSR) in terms of the overall volumetric mass-transfer coefficient (Kxa) using nitrogen stripping (N2–H2O–O2) and nitrogen stripping coupled with vacuum degassing (vacuum–N2–H2O–O2) processes. Principal component regression (PCR) method was used to establish a mass transfer model, and the number of principal components (PCs) was examined by three different techniques including cumulative percent variance (CPV), average eigenvalue (AE), and cross validation (CV). The prediction performance of the PCR model was compared to that of the multiple linear regression (MLR) model. Results reveal that the number of PCs determined by CV based on the predicted residual error sum of squares can be used to determine the optimal number of PCs to express the relationship between various modeling variables and Kxa. The values of Kxa predicted by the PCR and MLR models in the N2–H2O–O2 system were in agreement with the experimental values with deviations within 15% while those in the vacuum–N2–H2O–O2 system generally agreed with the experimental values with deviations within 15% and 30%, respectively. These results indicate that the PCR method is best suited for mass transfer modeling in an RSR.