OATAO - Open Archive Toulouse Archive Ouverte Open Access Week

Electrostatic forces in fluidized bed reactors : numerical and experimental analysis

Montilla Estrella, Carlos. Electrostatic forces in fluidized bed reactors : numerical and experimental analysis. PhD, Dynamique des fluides, Institut National Polytechnique de Toulouse, 2021

[img]
Preview
(Document in English)

PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
4MB

Abstract

Fluidized bed reactors are one of the unit processes most commonly used in the industry. Plastic production, energy conversion, petroleum refining, and medicine manufacturing are just a few examples of the fields benefiting from this type of technology. Although important advances have been made towards the understanding and prediction of the dynamics of fluidized beds, many important questions remain unanswered. One of the most important open challenges is the study of the effects of electrostatic forces inside the reactor. This electrostatic interaction is known to be the cause of some important problems such as the accumulation of material at the reactor's wall, the risk of explosion, the perturbation of nearby electronic devices and even the complete loss of the fluidization state. Despite the important research efforts in the last few decades, many problems are still unsolved. Amongst these, we find the use of non-invasive measurements techniques to characterize the hydrodynamics effects of electrostatic forces inside the reactor; and the macroscopic mathematical modeling of the charging dynamics in the bed. These are the issues that this research program tries to address. As part of the project ANR-IPAF, this Ph.D. thesis aims at improving the understanding of the effects of electrostatic forces in a fluidized bed reactor. On the modeling front, we use the kinetic theory of rapid granular flow to derive the most complete Eulerian model of the particle electric charge dynamics in monodispersed gas-solid flow systems. In this work, we show how to lift some of the most restrictive hypotheses of previous models. We show that the transport equation for the mean particle electric charge can be obtained without assuming the shape of the particle electric charge probability density function. In addition to this, we also derive and close the transport equation for the second order terms: the particle charge-velocity covariance and the particle charge variance. Our results show that a correct modeling of the second order moments is needed in dilute or highly electrically charged regions. Given that this complete model also adds many more partial differential equations to be solved, we study possible simplifications. Two algebraic models, one neglecting the effects of the charge variance and one taking it into account are proposed. The former proved to be suitable in configurations with low electric potential energy. However, the latter must be use with caution as it can become nonphysical in high charged situations. Finally, a semi-algebraic model is also proposed to solve the important limitations of the coupled algebraic model. On the experimental front, we study the use of an ECVT system to characterize the dynamics inside the bed. We focus our attention to the image reconstruction algorithm. We test the traditional reconstruction algorithms found in the literature. However, our results show that they are, either too inaccurate, or too computationally expensive. For these reasons, we explore the use of a novel reconstruction technique using machine learning algorithms. In this thesis, we propose two different strategies to train a feed forward artificial neural network to handle the image reconstruction step in a ECVT device. The first strategy is based on CFD-generated data which is coupled with the sensitivity matrix model to deduce the capacitance measurements. The second approach relies exclusively on real experimental data and it seeks to reconstruct an image that could explain the capacitance measurements. Our results show that artificial neural networks can be as accurate as the best image reconstruction algorithms found in the literature. However, they can reduce the computational cost by several order of magnitudes.

Item Type:PhD Thesis
Uncontrolled Keywords:
Institution:Université de Toulouse > Institut National Polytechnique de Toulouse - Toulouse INP (FRANCE)
Laboratory name:
Research Director:
Ansart, Renaud and Simonin, Olivier
Statistics:download
Deposited On:09 Dec 2021 10:55

Repository Staff Only: item control page