Multidimensional machine learning algorithms to learn liquid velocity inside a cylindrical bubble column reactor.

Area of Science:

• Chemical Engineering
• Fluid Dynamics
• Computational Science

Background:

• Multiphase chemical bubble column reactors exhibit complex fluid behaviors crucial for process efficiency.
• Predicting bubble flow dynamics is essential for optimizing reactor design and performance.

Purpose of the Study:

• To develop a hybrid computational fluid dynamics (CFD) and adaptive network-based fuzzy inference system (ANFIS) model for predicting bubble flow in a chemical bubble column reactor.
• To evaluate the effectiveness of ANFIS with varying input parameters (liquid node coordinates) for predicting liquid velocity.

Main Methods:

• Employed the Euler-Euler model within CFD to simulate flow patterns and turbulence.
• Utilized ANFIS as an artificial intelligence method, training it with CFD-generated data sets.
• Assessed ANFIS performance using one, two, and three input parameters (x, y, z coordinates) to predict liquid velocity in the x-direction (Ux).

Main Results:

• The ANFIS model demonstrated accurate prediction capabilities for liquid velocity.
• The best prediction accuracy was achieved when using three input parameters (x, y, z coordinates).
• The study highlights the successful integration of computational methods and computer science for modeling complex physical and chemical phenomena.

Conclusions:

• The combined CFD-ANFIS approach provides a powerful tool for understanding and predicting bubble flow in chemical reactors.
• This methodology showcases the potential of artificial intelligence and big data analytics in advancing chemical engineering processes.
• The findings confirm the capability of ANFIS to learn and predict complex fluid dynamics based on spatial data.