Flow regime and volume fraction identification using nuclear techniques, artificial neural networks and computational...

Area of Science:

• Nuclear Engineering
• Computational Fluid Dynamics
• Artificial Intelligence

Background:

• Accurate knowledge of multiphase flow regimes and volume fractions is essential for system performance prediction.
• Traditional methods for multiphase flow analysis can be complex and time-consuming.

Purpose of the Study:

• To develop and validate an artificial neural network (ANN) model for recognizing multiphase flow regimes.
• To accurately predict the volume fraction in multiphase flows using ANN pattern recognition.
• To integrate computational fluid dynamics (CFD) simulations with ANN models for enhanced accuracy.

Main Methods:

• Utilized gamma-ray pulse height distribution pattern recognition with an ANN.
• Developed flow regime models (annular, stratified) using MCNPX code for ANN training and testing.
• Validated simulated results through experiments in the stratified flow regime.
• Employed Ansys-CFX for CFD simulations of different volume fractions, converting models to voxels for MCNPX.

Main Results:

• The ANN model correctly recognized all tested flow regimes.
• Volume fractions were predicted with high accuracy, showing relative errors below 1.1%.
• Integration of CFD simulations improved the realism and accuracy of the multiphase flow analysis.

Conclusions:

• The developed ANN-based approach is effective for real-time multiphase flow regime identification and volume fraction prediction.
• Combining CFD with ANN provides a powerful tool for understanding and optimizing multiphase flow systems.
• This methodology offers a significant advancement in the field, bridging simulation and experimental validation.