Mathematical modelling and computational reduction of molten glass fluid flow in a furnace melting basin

Francesco Ballarin,Enrique Delgado Ávila,Andrea Mola,Gianluigi Rozza

2024-07-23

Abstract:In this work, we present the modelling and numerical simulation of a molten glass fluid flow in a furnace melting basin. We first derive a model for a molten glass fluid flow and present numerical simulations based on the Finite Element Method (FEM). We further discuss and validate the results obtained from the simulations by comparing them with experimental results. Finally, we also present a non-intrusive Proper Orthogonal Decomposition (POD) based on Artificial Neural Networks (ANN) to efficiently handle scenarios which require multiple simulations of the fluid flow upon changing parameters of relevant industrial interest. This approach lets us obtain solutions of a complex 3D model, with good accuracy with respect to the FEM solution, yet with negligible associated computational times.

Fluid Dynamics

What problem does this paper attempt to address?

The paper aims to address the mathematical modeling and numerical simulation of molten glass fluid flow in furnace melting pools and proposes a method based on non-intrusive Proper Orthogonal Decomposition (POD) combined with Artificial Neural Networks (ANN) to effectively handle fluid flow scenarios under different working conditions caused by parameter changes.

Specifically, the study first establishes a mathematical model of molten glass fluid flow, conducts numerical simulations based on the Finite Element Method (FEM), and then verifies the accuracy of the model by comparing it with experimental results. Considering that the fluid flow of molten glass in industrial applications is influenced by multiple parameters, such as the energy produced by methane combustion and the voltage of electric heaters (boosters), the paper proposes a non-intrusive POD-ANN model for rapidly solving fluid flow problems under different parameter combinations in complex three-dimensional models without the need for time-consuming full-scale numerical simulations.

The POD-ANN model can significantly reduce computation time while ensuring a certain level of accuracy, making it suitable for scenarios that require real-time monitoring of furnace conditions and timely measures. The model constructs a truncated singular value decomposition space during the training phase and then uses artificial neural networks to predict mode weights during the evaluation phase, thereby quickly obtaining an approximate solution for fluid flow. This method avoids the Galerkin projection step in traditional intrusive truncated singular value decomposition models, thus effectively handling nonlinear terms in the model and achieving real-time or near-real-time industrial applications. Additionally, the paper discusses how to evaluate bubble motion through post-processing steps, which is one of the key indicators for assessing the quality of molten glass, further demonstrating the model's practical value in real industrial problems.