Exploiting the prediction of mass transfer performance in aerated coaxial mixers containing biopolymer solutions using empirical correlations and neural networks

Abstract

The volumetric mass transfer coefficient is commonly used to assess the mixing effectiveness of gas–liquid bioreactor systems. Analyzing mass transfer performance in non‐Newtonian fluids inside coaxial mixers can be challenging due to the complex interaction between process variables, which requires developing robust characterization and estimation approaches. This study aims to investigate the gas dispersion in shear‐thinning biopolymers with yield stress using coaxial mixers in order to evaluate the effects of aeration and agitation on the volumetric mass transfer coefficient. A mixing configuration comprising a pitched blade turbine and an anchor was employed to disperse air into xanthan gum solutions, and the mass transfer performance was obtained at different impeller speeds and biopolymer concentrations by measuring the dissolved oxygen concentration. A dimensionless empirical correlation was proposed, and the results showed that the mass transfer was positively influenced by aeration intensity and agitation mechanism, quantified by the gas flow number and Reynolds number, respectively. Additionally, a strategy using stacking‐ensemble artificial neural networks was developed to accurately estimate the volumetric mass transfer coefficient, with a correlation coefficient of 0.998. The proposed mass transfer characterization approach overcame the complexities of analyzing aerated coaxial mixer systems and provided a reliable design model for bioreactor systems containing non‐Newtonian fluids.

Application of Machine Learning Models in Coaxial Bioreactors: Classification and Torque Prediction

“…This finding indicated that the most effective method for predicting the performance of the coaxial mixer was through the use of ensemble machine learning techniques. This was consistent with the findings of Barros et al [23], which demonstrated the superiority of stacked neural networks in predicting the mass transfer coefficient over other ANN models.…”

“…Kumar et al [22] demonstrated the potential of this method in developing a predictive model for the segregation index in particulate mixing. Based on our comprehensive literature review, only two studies have been published on the application of neural networks in coaxial mixers [23,24]. Barros et al [23] employed an artificial neural network (ANN) to predict the mass transfer coefficient in a coaxial bioreactor.…”

“…Based on our comprehensive literature review, only two studies have been published on the application of neural networks in coaxial mixers [23,24]. Barros et al [23] employed an artificial neural network (ANN) to predict the mass transfer coefficient in a coaxial bioreactor. They adopted two strategies: initially, they designed an ANN with two hidden layers, and subsequently, they implemented an ensemble neural network.…”