Gas-flow distribution in bubbling fluidized beds: CFD-based analysis and impact of operating conditions

Highlights

• Computational framework developed for gas-flow and residence-time distribution

• Flow-field coupled with bubble detection tool (MS3DATA) for detailed analysis

• Dense-phase close to minimum fluidization, except near bubble boundaries

• Bypass related to solids permeability and bubble rise; increases in bubble chains

• Throughflow constitutes 30–40% of gas flow with gas residence 50% shorter.

Abstract

Gas-flow distribution plays a critical role in the performance of fluidized beds because it directly affects gas residence-time and solids mixing. However, measuring it accurately in the harsh conditions of larger reactors is not possible. Therefore, this study is focused on the development of a rigorous computational framework for quantifying gas-flow distribution during fluidization. To this end, fine-grid simulations are conducted for the bubbling fluidization of two distinct Geldart B particles - 1.15 mm LLDPE and 0.50 mm glass particles, at superficial gas velocities U/U_mf = 2 and 3 in a 50 cm diameter bed. The two-fluid model (TFM) is employed to describe the solids motion efficiently and in-house developed tool MS3DATA (Multiphase-flow Statistics using 3D Detection and Tracking Algorithm) to compute detailed bubble statistics. The overall gas flow is divided into three phases: (a) dense flow in areas relatively rich is solids concentration (b) “visible” bubble flow associated with rising bubbles and (c) throughflow accounting for the gas flow which mostly bypasses through bubbles. It is found that conditions within the dense-phase depend largely on the particle properties while bubbling dynamics are significantly affected by superficial gas velocity. Calculations show that the throughflow increases in areas frequented by bubbles because the voidage distribution around bubbles increases the local dense-phase permeability. Throughflow may account for up to 40% of the overall gas flow, especially in the fluidization of large particles. This is not desirable because its residence-time is almost 2 × shorter (as compared to the dense flow) and contributes minimally to solids mixing. Finally, it is shown that in comparison to lab-scales, larger beds exhibit more homogeneous gas mixing. Insights from this study and the methodology developed will be useful in investigating gas flow distribution in complex fuel conversion systems.

Introduction

Fluidized beds are widely used in the chemical and petroleum industries, because of their high heat and mass transfer rates resulting from large gas-solids contacting [1]. However, design and performance optimization of commercial-scale fluidized beds continues to be challenging because of limitations of diagnostic techniques in the harsh conditions most fluidized beds operate in. Recently, there has been considerable progress in multiscale, multiphase modeling based on Computational Fluid Dynamics (CFD), which can accurately simulate local and global fluidization hydrodynamics and hence, can be used to investigate different physical and chemical phenomena occurring in these systems. This study is focused on developing a framework for describing the bubbling dynamics and gas-flow distribution, using fine-grid CFD simulations.

Distribution of gas flow in fluidized beds adversely affects gas residence time and solids mixing [1], [37], [38], [39], and therefore, impacts bed reactor's operation and performance. Recognizing this, substantial experimental and analytical effort have been undertaken to describe gas flow accurately and reliably [2], [3], [4], [5], [6]. The two-phase theory first proposed by Toomey and Johnstone [7] was based on the premise that the dense phase (or emulsion) is minimally fluidized (at U_mf) while the excess gas (=U − U_mf) flows through rising bubbles. Subsequent experimental measurements indicated that this theory grossly overestimates the visible bubble flow and demonstrated that a significant fraction of the gas bypasses through bubbles, especially in the fluidization of Geldart B and D particles [1], [2]. To account for the bypassed gas, commonly referred to as through-flow, modifications to this theory were proposed (most notably the n-type theory [2]), all of which are based on the division of the gas flow into three components: (a) gas contributing to the dense phase (with axial velocity U_d) (b) “visible” bubble flow Q_b and (c) throughflow (with axial velocity U_tf) which, at steady state, can be related by volume conservation at any axial location:

$$UA=(1-\delta )U_{d}A+Q_b+\delta {\mathrm{U}}_{tf}A$$

where U is the superficial gas velocity, A is the cross-sectional area of the bed and δ is the bubble fraction (fraction of cross-sectional area occupied by bubbles). From their experiments using thin rectangular fluidized beds, Grace and Harrison [2] suggested that U_d can be approximated as U_mf; this was later corroborated by the analytical modeling of Valenzuela and Glicksman [4] for large (Geldart B) particles. Additionally, they suggested that U_d must not depend on bubble characteristics and is, therefore, independent of the superficial gas velocity.

Meanwhile, visible bubble flow Q_b, which is the gas-flow associated with rising bubbles, has been conventionally quantified using cine photography and high speed video in thin-rectangular beds (e.g. [2], [8]) or probes (fiber optic or capacitance) in 3D beds [9], [10]. Despite considerable uncertainty associated with bubble measurements, it is generally established that Q_b is sensitive to both bed geometry and operating conditions. Inferences from thin-rectangular beds may not be applicable to real cylindrical bed systems because the hydrodynamics are significantly affected by the close proximity of walls in the spanwise direction [11], [12]. Further, in lab-scale beds, large bubbles and slugs are frequently observed which is not the case in relatively larger beds, where bubbles laterally coalesce towards the center over considerable axial distance resulting in significantly different solids circulation patterns [10], [13]. On the other hand, particle properties directly affect bubble structure: fluidization of small Geldart A particles is characterized by fast bubbles associated with clouds separating gas flow within bubbles from the dense-phase. In case of larger particles (Geldart B and D), the interstitial gas (through the dense-phase) is faster than the speed of typical bubble rise and is, therefore, able to bypass through bubbles [6].

Throughflow is frequently observed in large-particle systems and constitutes the component of gas flow that escapes into the freeboard through swarms of bubbles which offer low-resistance pathways. This results in minimal interaction with the dense-phase and adversely affects mixing of solid particles. Therefore, quantifying throughflow is critical for analyzing bed reactor performance and optimization, but direct measurement is not possible because of limitations in diagnostic techniques. Most studies (e.g. [3], [4], [6], [14], [15], [16]) have estimated throughflow indirectly, based on bubble measurements and assumptions regarding the dense-phase, while only recently [17], [18], attempts have been made to characterize throughflow using detailed analysis of experimental and numerical data. While the methodologies prescribed by the latter studies are useful for investigating throughflow fundamentally, their conclusions are based on observations in thin rectangular beds (where walls in the spanwise direction significantly impact fluidization hydrodynamics) and therefore, cannot be extrapolated to larger 3D fluidized beds [13], [19]. Overall, throughflow is expected to increase in areas of high coalescence activity, which are usually correlated with high density of small bubbles.

Given the chaotic nature of the interactions and the complex physical phenomena associated with bubbling fluidization, it is not surprising that a generalized description of the gas flow across a range of bed sizes and operating conditions has not been possible. This challenge is further exacerbated by technical limitations with simultaneous measurements of the dense and bubble phases. In cylindrical beds, measurements based on point probes (fiber-optic and capacitance) require assumptions regarding bubble shape and trajectory while accurate quantification using more sophisticated techniques (e.g. γ- and X-ray imaging) is only possible in lab-size beds and at low bubble fraction ( <0.1) [4], [8]. Thus, highly resolved CFD simulations can provide valuable insights into the bubbling behavior and gas distribution, and the developed framework can be employed in modeling different systems.

This work is part of a series of studies investigating bubbling fluidization of Geldart B particles. In [19], [20], [21], the computational framework and suitable metrics were developed to validate critical sub-models at the lab-scale. For predicting bubbling dynamics at large-scales, MS3DATA (Multiphase-flow Statistics using 3D Detection and Tracking Algorithm) was developed in [22] which uses time-resolved volumetric void fraction data from simulations. In [13], the effect of reactor size (bed diameter D and initial bed height H₀) on the fluidization hydrodynamics was examined using fine-grid CFD simulations and predictions were analyzed qualitatively using time-resolved visualization, bubble centroid and solids velocity vector maps and quantitatively using detailed bubble statistics and solids circulation metrics. Overall, it was shown that (a) scalability of predictions is only possible when bubble size and spatial distributions are consistent across scales (initial bed height must be lower than the critical height for gulf-stream circulations) and (b) for fluidization of Geldart B particles, predictions in a 50 cm diameter bed are reasonably independent of bed size.

This study is focused on developing a rigorous framework for describing gas-flow distribution in bubbling fluidized beds. To this end, simulations are conducted for the fluidization of two distinct particles (1.15 mm LLDPE and 0.5 mm glass) at superficial gas velocities U/U_mf = 2 and 3 in 15–70 cm diameter beds. Observations and insights are also compared with previous modeling efforts and experimental evidence. The simulation setup and fluidization metrics (phase-specific statistics) are first discussed briefly in 2 Simulation setup, 3 Fluidization metrics, respectively. Next, the computational framework for gas-flow distribution is detailed for the fluidization of LLDPE particles at U/U_mf = 2 and 3 in Sections 4.1–4.4. Finally, the effects of particle properties and reactor size are analyzed in 4.5 Effect of particle properties, 4.6 Effect of bed diameter, respectively. All simulations are performed using MFiX (Multiphase Flow with Interface eXchanges) [23], [28], an open-source code developed at the National Energy Technology Laboratory, USA to describe the hydrodynamics in solid-gas systems.