Search

The hydrodynamics of secondary flow phenomena in a disc-shaped gas vortex unit (GVU) is investigated using experimentally validated numerical simulations. The simulation using ANSYS FLUENT® v.14a reveals the development of a backflow region along the core of the central gas exhaust, and of a counterflow multivortex region in the bulk of the disc part of the unit. Under the tested conditions, the GVU flow is found to be highly spiraling in nature. Secondary flow phenomena develop as swirl becomes stronger. The backflow region develops first via the swirl-decay mechanism in the exhaust line. Near-wall jet formation in the boundary layers near the GVU end-walls eventually results in flow reversal in the bulk of the unit. When the jets grow stronger the counterflow becomes multivortex. The simulation results are validated with experimental data obtained from Stereoscopic Particle Image Velocimetry and surface oil visualization measurements.

A gas–liquid vortex reactor (GLVR) is considered one of the promising Process-Intensified Equipment for applications that require high mixing and mass transfer efficiency. To gain understanding of the fundamentals of GLVR operation, the gas–liquid hydrodynamics in the GLVR has been investigated by high-speed imaging, digital images analysis and cross correlation. Gas bubbles and liquid ligaments are observed in the GLVR over a wide operating window. The dynamic liquid layer thickness in the GLVR chamber varies in the range of 26–36 mm. The instantaneous and time-average velocity fields of the liquid phase are obtained with a guaranteed high signal-to-noise ratio (cross correlation peak ratio higher than 10), allowing the analysis of liquid turbulent properties including the turbulent kinetic energy and its dissipation rate. Results show that the high turbulent kinetic energy dissipation rate ranging from 23 to 114 m2/s3 contributes to a high mixing and mass transfer efficiency in the GLVR.

To study fouling in steam cracker convection section tubes, accurate tube wall temperature profiles are needed. In this work, tube wall temperature profiles are calculated using a hybrid model, combining a one-dimensional (1D) process gas side model and a computational fluid dynamics (CFD) flue gas side model. The CFD flue gas side model assures the flue gas side accuracy, accounting for local temperatures, while the 1D process gas side model limits the computational cost. Flow separation in the flue gas side at the upper circumference of each tube suggests the need for a compartmentalized 1D approach. A considerable effect is observed. The hybrid CFD-1D model provides accurate tube wall temperature profiles in a reasonable simulation time, a first step towards simulation-based design of more efficient steam cracker convection sections.

A three-dimensional numerical analysis of the flow in a Gas-Solid Vortex Unit (GSVU) is carried out. The numerical model is first validated by comparing the bed pressure drop and solids velocity with experimental data. Next, the influence of gas flow rate, solids density, and particle diameter on the pressure drop, solids velocity, bed void fraction and slip velocity between the two phases is studied. A stable, solids bed is achieved for the entire range of gas flow rates tested (0.1–0.6 Nm3/s). No particle entrainment is observed when varying the solid density (1800–450 kg/m3) or the particle diameter (2–0.5 mm). A shift to bubbling fluidization regime is observed for small particle diameters (0.5 mm). The observed changes in the GSVU flow patterns are discussed by analyzing the changes in the cumulative centrifugal to drag force ratio over the bed.

To clarify the three-dimensional structure of near-wall jets observed in disc-like gas vortex units, experimental and numerical studies are performed. The experimental results are obtained using Particle Image Velocimetry (PIV), Laser Doppler Anemometry (LDA), pressure probes and surface oil flow visualization techniques. The first three techniques have been used to investigate the bulk flow hydrodynamics of the vortex unit. Surface oil flow visualization is adopted to visualize streamlines near the end-walls of the vortex unit. The surface streamlines help determine the azimuthal and radial velocity components of the radial near-wall jets. Simulations of the vortex unit using FLUENT® v.14a are simultaneously performed, computationally resolving the near-wall jet regions in the axial direction. The simulation results together with the surface oil flow visualization establish the three-dimensional structure of the near-wall jets in gas vortex units for the first time in literature. It is also conjectured that the near-wall jets develop due to combined effects of bulk flow acceleration and swirl. The centrifugal force diminishes in the vicinity of the end-walls. The radially inward pressure gradient in these regions, no longer balanced by the centrifugal force, pushes gas radially inward thus developing the near-wall jets. This article is protected by copyright. All rights reserved.

The radiation and convection section of a steam cracker are thermally coupled. Optimization and design requires a coupled simulation of both sections. In this work a 1D model for the convection section, CONVEC-1D, is developed. Several models for the different heat transfer phenomena are implemented and evaluated. For flow boiling, an empirical and a mechanistic model are developed and compared for both single- and multicomponent hydrocarbon feeds. The latter is performing best over a wide range of operating conditions, taking into account the different two-phase flow regimes. The coupled iterative procedure is demonstrated for an n-pentane steam cracker convection section.

Coincident (simultaneous) three‐component particle velocity measurements performed using two laser Doppler anemometry probes at the outlet section of a 9 m high cylindrical riser are for the first time presented for dilute flow conditions. Near the blinded extension of the T‐outlet a solids vortex is formed. Particle downflow along the riser wall opposite the outlet tube is observed, which is restricted to higher riser heights at higher gas flow rates. Increased velocity fluctuations are observed in the solids vortex and downflow region as well as at heights corresponding to the outlet tube. Contrary to the rest of the riser, in the downflow region time and ensemble velocity averages are not equal. Given the local bending of the streamlines, axial momentum transforms to radial and azimuthal momentum giving rise to the corresponding shear stresses. Turbulence intensity values indicate the edges of the downflow region. © 2016 The Authors AIChE Journal published by Wiley Periodicals, Inc. on behalf of American Institute of Chemical Engineers AIChE J, 62: 3575–3584, 2016

Regime maps that predict the heavy hydrocarbon droplet impingement behavior on a heated wall (Weber number of the impinging droplet v/s wall temperature) are constructed based on CFD simulations using the Volume of Fluid model with the geo-reconstruct scheme. Based on the simulation results, maps are constructed for single-component droplets with a diameter of 50 and of 100 µm. The applied CFD model is validated by comparing these with regime maps available in literature, constructed based on experimental data for model liquids and liquid mixtures. The impingement regimes of Splash, Stick, Rebound and Breakup are well-predicted. Two distinct types of Splash (Splash with ligament formation and Splash with ring detachment) are reported for the first time. Using the validated CFD model, regime maps are constructed for multi-component heavy hydrocarbon droplets with a diameter of 50 and of 100 µm. Significant differences between the single-component and the multi-component droplet impact behavior are observed. Improved and new correlations for regime transitions, droplet stretching on the wall, droplet rebounding velocity and number of splashed droplets are derived based on energy balances. They are found to correlate well with CFD predictions.

Coincident three-component particle velocity measurements are performed in an almost 9 m high cylindrical riser (internal diameter 0.10 m) of a pilot-scale cold-flow Circulating Fluidized Bed set-up, using two Laser Doppler Anemometry (LDA) probes. Experiments are performed with superficial gas velocities of 3.5 m/s and 5.3 m/s near the solids inlet line and with an average solids volume fraction of 0.0002. The particle flow is observed to be highly disturbed due to the asymmetrical position of both the gas and the Y-shaped solids inlet line. A jet-like particle flow and a by-passing streaming gas jet around the particles are formed. Just below the solids inlet line a downward oriented particle velocity is measured. Particles are immediately lifted by the ascending gas and spread over the entire cross-sectional tube area. Near the solids inlet line the axial and radial particle velocity fluctuations are of the same order of magnitude as the corresponding mean particle velocity components, showing that the flow is anisotropic. The radial particle velocity fluctuations decrease fast as the radial inflow of particles is transformed into axial particles flow. The turbulence intensity in the inlet section of the dilute riser flow is found to be extremely high, indicating that the flow is far from being fully developed. The total particle shear stress profiles near the solids inlet line indicate a high momentum transfer from radial to axial particle transport. The particle fluctuation energy values calculated based on experimental data are nearly double than the values corresponding in fully developed flow studied previously.

In a Gas–Solid Vortex Reactor (GSVR), also referred to as a Rotating Fluidized Bed in Static Geometry, a fluidized bed is generated in a centrifugal field by introducing the gas via tangential inlet slots to the reactor chamber. Better heat and mass transfer are observed, making this a promising reactor type for Process Intensification. Developing GSVRs on industrial scale requires, amongst other, a good insight and understanding of the hydrodynamics of the granular flow. In the present work experiments are performed over a wide range of operating conditions in a cold flow pilot-scale set-up. The set-up has a diameter of 0.54 m, a length of 0.1 m and 36 tangential inlet slots of 2 mm. Different materials with solids density between 950–1800 kg/m3 and particle diameters of 1–2 mm, at varying gas injection velocities from 55 to 110 m/s are tested between minimum and maximum solids capacities. All these operating conditions are used to follow the change of granular flow by performing PIV. The rotating fluidized bed can change from a smoothly rotating, densely fluidized bed to a highly bubbling rotating fluidized bed depending on the operating conditions. Bubbling diminishes with increasing solids density and particle diameter. Experimental measurements of azimuthal particle velocity fields in a GSVR are for the first time reported. Azimuthal solids velocity is found to decrease with higher solids density and larger particle diameter. The critical minimum fluidization velocity, that is the minimum velocity at which the complete bed is fluidized, is calculated and the centrifugal bed behavior is mapped in terms of a dimensionless radial gas velocity and a dimensionless particle diameter, as conventionally done for gravitational beds.

A drag model for the Gas-Solid Vortex Unit is presented. A large set of experimental data, obtained using static gauge pressure measurements, Particle Image Velocimetry and Digital Image Analysis in different GSVU geometries is used to estimate the model parameters with regression analysis. A large range of radial particle Reynolds numbers, particle densities and particle diameters is investigated. The newly proposed drag coefficient correlation, including 95% confidence intervals, reads: C D , GSVU = 15.00 ± 4.65 e 2 R e p , R − 0.28 ± 0.05 S 0.76 ± 0.03 The drag coefficient correlation is found to depend on the swirl ratio S, a variable determined by the GSVU geometry. The model suggests that, for the operating conditions in the GSVU, the particles mostly behave independently of one another. The performance of the model to predict the pressure drop over fluidized beds in a centrifugal field is high as compared to the standard drag models for fluidized beds in the gravitational field. Additionally, the drag model eventually allows to calculate the azimuthal solids velocity in a GSVU under varying operating conditions using an Archimedes/Reynolds number correlation derived from the radial momentum balances for both phases.

Coincident three-component solids velocity measurements are performed in a 8.7-m-high cylindrical riser (internal diameter 0.10 m) of a pilot-scale cold-flow Circulating Fluidized Bed set-up, using two Laser Doppler Anemometry (LDA) probes. Experiments are performed with superficial gas velocities of 3.5 and 5.3 m/s at riser heights from 0.9 to 5.5 m and with an average solids volume fraction of 0.0002. At the lower riser heights, the solids flow is observed to be highly disturbed due to the asymmetrical position of both the air and solids inlet. A fully developed parabolic axial solids velocity field is observed about 2 m above the solids inlet line. The measured radial and azimuthal particle velocity components are significantly smaller than the axial one. The particle velocity fluctuations in the axial flow direction are larger than in the radial and azimuthal directions, showing that the solids flow is anisotropic. However, both the radial and azimuthal velocity fluctuations are higher than the corresponding mean velocities, as observed when studying the turbulence intensity in the three directions. All fluctuating velocities are decreasing significantly with increasing riser height, i.e. as the flow becomes more developed. No average down-flow of solids is recorded, not even close to the wall, due to the highly diluted flow. The total particle shear stress profiles designate a mean transport of fluctuating momentum along the riser radius due to velocity fluctuations. The particle fluctuation energy calculated based on experimental data indicates that the solids flow is dominated by wall-particle rather than intra-particle collisions.

The hydrodynamic behavior of dispersed gas–solid flow is simulated for an industrial-scale Fluid Catalytic Cracking (FCC) riser using the multi-fluid approach, with complementary information from the Kinetic Theory of Granular Flow (KTGF) for the transport coefficients of the solid phase. A continuous Particle Diameter Distribution (PDD) is considered for the solid phase catalyst. The three-dimensional gas–solid flow is considered to be non-isothermal and turbulent. The hydrodynamics of the riser reactor is coupled to a 12-lump FCC kinetic model to predict the influence of a polydisperse distribution on gas–solid reactive flow. The kinetic model involves lumped species consisting of paraffins, naphthenes, aromatic rings, and aromatic substituent groups in medium and heavy fuel oil fractions and includes the effect of aromatic ring adsorption and catalyst deactivation due to coke formation. In this study, a Computational Fluid Dynamics (CFD) model using the Eulerian–Eulerian multi-fluid approach and a Population Balance Model (PBM) are coupled. The hydrodynamic equations are solved by means of a finite volume method, while the population balance equations are solved using the Direct Quadrature Method of Moments (DQMOM). The moments of the solids phase velocity are modeled using classical kinetic theory, while the moments of the PDD are described using quadrature weights and abscissas. To account for the catalyst PDD, three different approaches namely simple, surface-averaged and volume-averaged coupling have been attempted. The latter two approaches, assuming surface and volumetric reaction kinetics respectively, predict with a good precision the yields of the different product families obtained in an industrial FCC unit showing a slight improvement from a conventional heterogeneous reactor model with a monodispersed distribution.

A Eulerian–Lagrangian method is implemented to simulate turbulent two-phase gas–solid riser flow, using a mean-field/probability density function (PDF) method. The mean-field method is applied to the gas phase while the PDF method is applied to the solid phase. Using the PDF method for the solid phase is advantageous as there is no need for closure models for the convection term in the momentum equation contrary to the situation where the mean-field method is used. The present method is implemented to investigate the influence of a side inlet on the flow pattern in a dilute gas–solid riser configuration. When applying the Eulerian–Lagrangian method, a perfect correspondence with the experimental observations is obtained. On the contrary, when using the Eulerian–Eulerian method significant differences with the experimental data are observed.

A hybrid Lagrangian-Eulerian methodology is developed for the numerical simulation of turbulent reactive gas-solid flow. The SO2-NOx Adsorption Process (SNAP) in a riser reactor with dilute gas-solid flow is taken as a test case. A threedimensional time-dependent simulation is performed. By using the transported composition PDF method [1], modeling of the mean chemical source term and mass transfer terms in the gas-solid flow model equations is no longer needed. A notional particle-based Monte-Carlo algorithm is used to solve the transported composition PDF equations. A Finite-Volume technique is used to calculate the hydrodynamic fields from the Reynolds Averaged Navier Stokes (RANS) equations combined with the k-e turbulence model for the gas phase and the Kinetic Theory of Granular Flow (KTGF) for the solid phase [2]. The newly developed hybrid solution technique is tested with the SNAP chemistry that has a total of 13 scalars (i.e., 5 gas phase components and 8 solid phase species) for which the composition fields of the reactive species are calculated. A good agreement between simulated and experimental gas-outlet composition of a demonstration unit is obtained.

The presence of liquid hydrocarbon droplets in the convection section of a steam cracker may cause serious fouling problems due to coke formation, especially in high temperature zones. In order to investigate these fouling problems, a model has been developed and implemented in a CFD code to accurately simulate the behavior of a hydrocarbon droplet impinging on a hot surface. The impact energy of the droplet and the hot surface temperature are found to determine the impact behavior. On the basis of the newly developed model, the positions where droplets preferentially collide with the convection section tube walls and liquid material is deposited are determined. Furthermore, a kinetic model describing coke formation out of liquid hydrocarbon droplets in the temperature range of 450−700 K has been developed to calculate the rate of coke formation in each zone of the convection section tube. The calculated coke layer thickness on the most vulnerable tube locations and the industrially available values corre...

Computational fluid dynamics (CFD) simulations of the hydrodynamic behavior of rotating fluidized beds in static geometry (RFB-SG) are carried out for gas–solid flows. The rotating motion of the reactor bed is induced by the tangential injection of the gas along the circumference of the fluidization chamber. Steep gradients in the gas velocity fields both in radial and tangential direction generate turbulence. The radial and tangential drag forces fluidize the particle bed in both radial and tangential direction. An Eulerian two-fluid model is used. Gas phase turbulence is accounted for by a k–e model adapted for rotational flows. The RFB-SG simulations provide guidelines for a design and operation with a high efficiency in gas–solid momentum transfer, excellent gas–solid separation and limited solids losses. Hydrodynamic variables like the centrifugal force, the injection pressure, the radial and tangential slip velocities, solids hold-up are calculated for both polymer particles (300 μm, 950 kg/m3, Geldart Group B) and glass beads (70 μm, 2500 kg/m3, Geldart Group A) to allow for a comparison among different fluidization chamber designs. Unstable bed behavior, like slugging and channeling, is also numerically predicted.

A hybrid Lagrangian-Eulerian methodology is developed for the numerical simulation of turbulent reactive gas-solid flow. The SO2-NOx Adsorption Process (SNAP) in a riser reactor with dilute gas-solid flow is taken as a test case. A three-dimensional time-dependent simulation is performed. By using the transported composition PDF method [1], modeling of the mean chemical source term and mass transfer terms in the gas-solid flow model equations is no longer needed. A notional particle-based Monte-Carlo algorithm is used to solve the transported composition PDF equations. A Finite-Volume technique is used to calculate the hydrodynamic fields from the Reynolds Averaged Navier Stokes (RANS) equations combined with the k-epsilon turbulence model for the gas phase and the Kinetic Theory of Granular Flow (KTGF) for the solid phase [2]. The newly developed hybrid solution technique is tested with the SNAP chemistry that has a total of 13 scalars (i.e., 5 gas phase components and 8 solid phase species) for which the composition fields of the reactive species are calculated. A good agreement between simulated and experimental gas-outlet composition of a demonstration unit is obtained.

A model has been developed allowing to simulate the flow boiling process of a hydrocarbon feedstock. For the calculation of the different horizontal two-phase flow regimes that evolve during flow boiling, use is made of the Volume Of Fluid (VOF) model that uses a Piecewise Linear Interface Calculation (PLIC) method to reconstruct the interface between both phases in each computational cell. In-house developed codes calculate the mass and energy transfer phenomena occurring during this flow boiling process. As such, an existing CFD code is completed with a newly developed complete evaporation model. The developed model is used to simulate the flow boiling process of a hydrocarbon feedstock in the tubes of a convection section heat exchanger of a steam cracker. The simulation results show a succession of horizontal two-phase flow regimes in agreement with the literature.

A 3D RANS simulation of lean premixed turbulent combustion in an industrial scale steam cracking furnace is performed. Turbulence is represented by a renormalization group (RNG) k − e model, while combustion is modeled with a skeletal mechanism. The eddy dissipation concept (EDC) model is used to account for turbulence/combustion interaction. The turbulent combustion regime is identified as a “thin flame regime with pockets” or a “corrugated flamelet regime”. A weighted sum of gray gases model (WSGGM) is used to account for the gas radiative properties. The radiative transfer equation is solved using the P-1 and discrete ordinates method in the conservative finite volume formulation. NOx calculations are performed as a postprocessing step, with the flow field, temperature, and species concentrations fixed. The formation of NO through thermal, prompt, and N2O intermediate mechanisms is considered. The in situ adaptive tabulation technique is used to decrease the computational time required to treat the combustion chemistry. The effect of radiation models on the predicted furnace wall, tube skin, and flue gas temperature profiles and heat fluxes toward the reactor tubes, as well as on the predicted species and NO concentration profiles and structure of the furnace flames under normal firing conditions, is discussed. © 2007 American Institute of Chemical Engineers AIChE J, 2007

Inlet phenomena in a 0.1 m diameter cold-flow riser with a 35 ∘ inclined side inlet are studied experimentally using 3D-Laser Doppler Anemometry for solids fluxes of 0.5– 4.5 kg / m 2 / s and gas velocities of 5.3–7.4 m/s. In the vicinity of the solids inlet, radial gas-solids mixing is hindered and bypassing of the solids jet occurs, resulting in steep velocity gradients and off-centre maxima in the velocity field. The feeding conditions and the type of the solids affect the bottom operation and gas–solids mixing to a large extent: compared to FCC particles, silica particles extend the acceleration zone in the riser. Low gas flow rates and/or high solids feeding rates result in an increased penetration depth of the solids jet and in explicit bypass zones in the plane facing the inlet. High root mean square fluctuating particle velocities are observed at the solids jet boundaries. A non-aerated Y-inlet configuration causes vortex formation, inducing a small reflux into the upper dilute part of the standpipe. The influence of dilution of the inlet solids jet is also investigated using an “aerated” inlet configuration. Aerated inlets lead to better entrainment, improved radial mixing, less pronounced broader bypass zones and a firm reduction of the penetration depth. In the 0.1 m diameter riser, radial mixing quickly dissipates the non-uniformities introduced by the solids inlet. Reflection phenomena can, however, occur in the case of a non-aerated solids inlet.

Simultaneous solution algorithms for Eulerian-Eulerian gas-solid flow models are presented and their stability analyzed. The integration algorithms are based on dual-time stepping with fourth-order Runge-Kutta in pseudo-time. The domain is solved point or plane wise. The discretization of the inviscid terms is based on a low-Mach limit of the multi-phase preconditioned advection upstream splitting method (MP-AUSMP). The numerical stability of the simultaneous solution algorithms is analyzed in 2D with the Fourier method. Stability results are compared with the convergence behaviour of 3D riser simulations. The impact of the grid aspect ratio, preconditioning, artificial dissipation, and the treatment of the source terms is investigated. A particular advantage of the simultaneous solution algorithms is that they allow a fully implicit treatment of the source terms which are of crucial importance for the Eulerian-Eulerian gas-solid flow models and their solution. The numerical stability of the optimal simultaneous solution algorithm is analyzed for different solids volume fractions and gas-solid slip velocities. Furthermore, the effect of the grid resolution on the convergence behaviour and the simulation results is investigated. Finally, simulations of the bottom zone of a pilot-scale riser with a side solids inlet are experimentally validated.

A three-dimensional simulation of a dilute phase riser reactor (solid mass flux: 2 kg m −2 s −1 ) is performed using a novel density based solution algorithm. The model equations consisting of continuity, momentum, energy and species balances for both phases, are formulated following the Eulerian–Eulerian approach. The kinetic theory of granular flow is applied. The gas phase turbulence is accounted for via a k – e model. An extra transport equation describes the correlation between the gas and solid phase fluctuating motion. The solution algorithm allows a simultaneous integration of all the model equations in contrast to the sequential multi-loop solution in the conventional pressure based algorithms, used so far in riser simulations. The simulations show an unsteady behaviour of the flow, but a core-annulus flow pattern emerges on a time-averaged basis. The abrupt nature of the T type outlets causes a significant recirculation of gas and solid from the top of the riser. The flow near the outlets is highly non-symmetric and has a three-dimensional character. A significant decrease of the gas phase turbulence and particle granular temperature across the riser length is attributed to the presence of small particles, which is qualitatively consistent with the experimental data from literature.

Gas–solid flow in circulating fluidized beds is calculated using the Eulerian–Eulerian approach with the kinetic theory of granular flow. The usefulness of this approach and the necessity of performing 3D calculations are illustrated by calculating the exit effects of single abrupt outlet configurations of different outlet surface area.

Bubbles in confined geometries serve an important role for industrial operations involving bubble-liquid interactions. However, high Reynolds number bubble dynamics in confined flows are still not well understood due to experimental challenges. In the present paper, combined experimental and numerical methods are used to provide a comprehensive insight into these dynamics. The bubble behaviour in a vertical Hele-Shaw cell is investigated experimentally with a fully wetting liquid for a variety of gap thicknesses. A numerical model is developed using the volume of fluid method coupled with a continuum surface force model and a wall friction model. The developed model successfully simulates the dynamics of a bubble under the present experimental conditions and shows good agreement between experimental and simulation results. It is found that with an increased spacing between the cell walls, the bubble shape changes from oblate ellipsoid and spherical-cap to more complicated shapes, while the bubble path changes from only rectilinear to a combination of oscillating and rectilinear; the bubble drag coefficient decreases and this results in a higher bubble velocity caused by a lower pressure exerted on the bubble; the wake boundary and wake length evolve gradually accompanied by vortex formation and shedding.

Segregation and mixing effects of particle diameter distributions are numerically investigated in the riser section of a circulating fluidized bed. A granular kinetic theory based approach, which implements the Eulerian quadrature-based moment method to describe the particle phase, is coupled to an Eulerian multi-fluid solver through a population balance model. The gas-solid multiphase flow is two-dimensional, transient and isothermal. The particle distributions fed from each side inlet of the riser have different variances, but the same mean diameter. The core-annular regime used as a numerical benchmark for riser flows is well predicted. A comparison in the homogeneity of particle mixing is made from the lower-order moments of the particle distribution obtained at various positions and at different axial lengths along the riser. It is seen that the relative standard deviation of the particle distribution varies spatially, indicating dynamic mixing inside the riser.

For the calculation of gas–solid flow in circulating fluidized beds, a Eulerian–Eulerian approach is taken. An integration scheme based on dual time stepping and a finite volume technique is developed and implemented in 3D. The inviscid part of the equations is treated following an extension of the preconditioned advection upstream splitting method (AUSMP) to two-phase flows. Calculations on an industrial size straight riser are performed. The influence of the inelasticity of particle–particle collisions on the stability of the flow is investigated. Further, the effects of a double abrupt side outlet configuration are shown.