Anchoring CFD Models to Real-World Bubble Columns Using Wall Pressure Fluctuations

Bubble column reactors (BCRs) are widely used in chemical and other allied industries. Computational fluid dynamics (CFD) models are often used to simulate and optimize BCRs. However, these models still require ad hoc adjustments of model parameters to anchor simulations to real-world reactors. A recently proposed idea suggests that wall pressure fluctuations could be exploited to obtain relevant closure models. In this work, we evaluated this idea for simulating BCRs. A published case of a bubble column with 0.1 m diameter and 2 m height, for which the gas holdup and wall pressure fluctuation data were available, was considered. Three-dimensional transient flow in the BCR was simulated by using the Euler–Euler models. In the first step, following the conventional approach, the values of the key parameter (C_D/d_B) for estimating interphase drag were obtained by matching simulated results with experimental gas holdup data. An artificial neural network (ANN) was then trained to relate experimental wall pressure fluctuations data with those of (C_D/d_B). The hybrid model demonstrated high accuracy, with an R² of 0.99, and the predicted values fall within ±5% of the experimental data. The trained ANN was subsequently used to estimate the (C_D/d_B) for the cases unseen by the model. The wall pressure fluctuations-based approach successfully captured the influence of gas velocity on gas holdup for both the water and the 1.5% ethanol case. The approach was also found to be reasonably successful for simulating the influence of ethanol concentration on gas holdup. These findings demonstrate the feasibility of combining wall pressure fluctuations with machine learning to estimate relevant model parameters. This work establishes a robust foundation for advancing the presented approach toward anchoring CFD models to real-world reactor systems.