Experimental Characterization and CFD Simulations of Gas Maldistribution Relevant to Fluidized Beds of Group A Particles

Troiano, M., Università degli Studi di Napoli Federico II
Amblard, B., IFP Energies nouvelles
Solimene, R., Consiglio Nazionale delle Ricerche
Salatino, P., Università degli Studi di Napoli Federico II
An experimental and computational study of gas maldistribution in fluidized bed reactors is hereby presented. Gas maldistribution can be often experienced in industrial fluidized bed reactors, affecting the performance in terms of operation and potential damages of internals within the reactors.

The present study aimed at investigating the hydrodynamic behaviour of the fluidized bed in presence of a gas maldistribution and the capability of the CFD code BarracudaTM to predict the effects of the maldistribution of gas in a fluidized bed under well controlled conditions. For this purpose, experiments were carried out in a 20 cm diameter cold flow fluidized bed column connected to a single stage cyclone to capture entrained particles and to re-insert them into the fluidized bed by means of a O.D. 40 mm cyclone dipleg and an automatic L-valve. The column is filled with Group A particles (FCC catalyst, ρP =1260 kg/m3, Dp50=75 mm), with a bed height around 1.1 m. Two configurations of gas distribution have been investigated under turbulent fluidization regime (superficial gas velocity of 0.7 m/s). In the first operative configuration, called “even fluidization”, all the gas is evenly fed at the bottom of the bed through a porous distributor. In the second operative configuration, the gas is partly fed to the bottom of the bed and partly injected through a nozzle located at the centre of the bed. This configuration is called “uneven fluidization” and it is aimed at purposely creating a gas maldistribution. Hydrodynamic characterization of the fluidized bed was carried out under even and uneven fluidization conditions by mapping solids concentration along radial and axial directions.

Local solid volume fraction profiles were measured by using optical probes at different bed heights and along the two radial directions of the column, normal to each other. Two types of optical probes are used: Type I, much larger than the particle diameter, and Type II, with a cross-sectional dimension closer to the particle diameter. Type I probes are one-channel fibre probes consisting of an 8 mm ID metallic rod. They are capable of detecting all the particles in the measuring volume and are commonly used to measure solids concentrations. Type II probes consist of two independent channels, separated by a calibrated distance. They can detect small groups of particles or single particles and are generally employed to determine particle velocities. In both configurations, channels consist of thousands of optical fibres arranged parallel to each other that can emit and reflect light. The arrangement is such that one layer consists of light emitting fibres and the next one consists of light receiving fibres. Both types have been used in the present experimental campaign, to measure the solid concentration. The optical probes were connected to a particle analyser consisting of photoelectric converter and amplifying circuits, signal pre-processing circuits, high-speed A/D interface card and acquisition software. The probe was inserted into the fluidized bed, perpendicular to the flow. When catalyst particles in the emulsion phase pass in front of the probe, a high percentage of the emitted light is reflected and transferred to the photocell, which converts it into a high voltage signal. In contrast, when a gas void passes, relatively little light is reflected to the probe and the photocell responds by giving a low voltage signal. The light reflection signals detected on each fibre was recorded with an acquisition frequency of 1 kHz. The sampling time was 60 s, chosen after trials with different sampling times (30 s, 60 s, 120 s), to minimize the sampling time without compromising reproducibility of average concentration profiles during the turbulent fluidization conditions. The concentrations of solid particles were measured in the axial direction (optical probes were inserted in the bed at 0.05, 0.1 and 0.15 m above the jet) and in the two main radial directions (11 measurement points for each direction). Solid concentration profiles have been compared with the average solid concentration obtained by means of pressure probes.

The experimental results have been then compared with CFD simulations in order to assess if the CFD code BarracudaTM is able to catch the effect of the gas injection configuration on the bed solid volume fraction. The gas/particles hydrodynamic modeling in BarracudaTM is based on the Multi-phase Particle in Cell method (MP-PIC). The gas phase is treated as a continuum in a Eulerian framework solving the averaged Navier-Stokes equations. Particles are treated with a hybrid Eulerian-Lagragian approach where equations are derived from the Liouville equation which governs the transport in an Eulerian framework of a particle distribution function f(x,u,m,t) where x is the spatial location, u and m are respectively the velocity and the mass of the particle and t is the time. The particle-phase equations are solved by discretizing the particle distribution function f(x,u,m,t) into computational parcels, each of which represents a certain number of real particles of identical size, velocity and position. For the resolution, computational parcels properties are interpolated onto the Eulerian grid to solve the solid-phase equations, once the equations are solved on the grid, the Eulerian grid properties such as gas velocities, gas pressure gradients and solids stress gradients are interpolated back to the parcel in order to update its position and velocity using a Lagrangian approach. Collisions between particles are modeled through the use of a particle stress function. The advantage of the MP-PIC method is that the coupling term between the gas and solid phases takes into account a distribution of particles with different sizes and velocities while a classic Eulerian approach for the solid phase only takes into account an averaged particle diameter and an averaged solid velocity for the interphase coupling term. In this paper, the particle–fluid drag force was expressed using the Gidaspow drag function which combines the Wen-Yu model and the Ergun model. Then, the turbulence of the gas phase is modeled through a Large Eddy Simulation (LES) approach where the large-scale motion due to the large eddies are computed directly from the gas flow equation while the small-scale eddies motion is captured through the LES Smagorinsky subgrid scale model. Concerning the interactions with the wall, for the gas phase a no slip boundary condition is applied while for the solid phase, the velocity vectors of the particles after bouncing on a wall are computed through normal and tangential momentum retention (or restitution) coefficients. In the current simulation, the default retention coefficients of the software were used: 0.3 for the normal retention coefficient which means that the particles keep 30% of its normal velocity after bouncing and 0.99 for the tangential retention coefficient which means that the particles keep 99% of its tangential velocity after bouncing. The effect of drag correlation, inlet gas boundary conditions and jet injections have been investigated. The radial profile of solid concentration at three distances from the jet injection have been calculated with and without the presence of the central gas jet.

The experimental results allowed to compare the solid concentration profiles using the two different optical probes (one-channel and two-channel probes). The experimental profiles highlight a very good qualitative and quantitative agreement for the tests in the even fluidization configuration when using the two kinds of probes. Regarding the tests in the uneven fluidization configuration, both the one-channel and two-channel probes are able to catch the presence of the gas stream. However, results obtained with the two probes highlight some differences, probably related to the different measurement volume of the two kinds of probes. When comparing experimental and CFD results, it is noteworthy that the choice of drag correlation and boundary conditions strongly influences the agreement between the experimental and CFD results in terms of radial and axial profiles of solid concentration.