Adaptive neuro-fuzzy prediction of flow pattern and gas hold-up in bubble column reactors

The prediction of fluid dynamics in multiphase bubble column reactors is a subject of major concern to appropriately design and optimize them. This paper employs the combination of computational fluid dynamics (CFD) (i.e., Euler-Euler approach) and adaptive neuro-fuzzy inference system (ANFIS) to propose new a viewpoint for multiphase modeling, including the accuracy of soft computing technique in prediction of a 3D bubble column reactor. Existing experimental, numerical and correlations results in the literature have been used to validate the implementation of the Euler-Euler approach. The results of Euler-Euler approach for a 3D bubble column reactor has been used for input training data which are liquid velocity, turbulent kinetic energy and gas hold-up. The ANFIS results have been also compared with Eulerian results, using root-mean-square error (RMSE) and coefficient of determination and Pearson coefficient. The results show that, flow pattern and gas hold-up are mainly affected by bubble column height, meaning towards sparger region, gas hold-up has a higher value near the ring sparger. According to the results, a greater improvement in estimation has been achieved through the ANFIS. Overall, the results show that ANFIS is a robust method to predict bubble column hydrodynamics parameters (e.g., liquid flow pattern and gas hold-up) as input. In addition, the exactness of the proposed ANFIS model may be boosted by considering more meteorological parameters as input values.

Automated summary

This study combines CFD and ANFIS to propose a new viewpoint for multiphase modeling and validates the accuracy of the Euler-Euler approach, showing that the flow pattern and gas hold-up of a bubble column reactor are mainly influenced by the height of the column. The results suggest that ANFIS is a robust method for predicting hydrodynamics parameters and the model's accuracy can be further enhanced by considering more meteorological parameters as input values.