X.-Q. Chen

Computation of turbulent evaporating sprays with well-specified measurements: a sensitivity study on droplet properties

Abstract An extensive numerical study was carried out for a confined evaporating spray in a turbulent, heated gas flow using a published well-defined experimental dataset. The Eulerian-Lagrangian stochastic models were employed for spray calculations wherein the gas turbulence was modeled using the second-moment transport model for the Reynolds stresses and heat-flux vectors, and the droplet dispersion was modeled using the Lagrangian stochastic models with or without temporal correlations. Two fashions of the infinite-conduction-evaporation model were studied, both of which have taken into account the variable gas-film properties by the 1 3 - rule . Numerical results for the droplet phase, i.e., the mean diameters, mass fluxes, mean and fluctuating velocities were presented and discussed by comparison with the experimental data. The sensitivity of various droplet properties to the number of droplet trajectories at the inlet, the drift correction approaches for the improvement of mass-flux predictions, and the evaporation models was investigated in terms of the well-defined experimental dataset. Results show that the droplet mean velocities are generally not sensitive to all the factors considered, that droplet r.m.s. velocities downstream are sensitive to the number of trajectories, that the droplet mass-flux accumulation near the centreline can be substantially improved by using a new drift correction approach, and that mass-flux predictions are sensitive to the evaporation models.

PREDICTION OF EVAPORATING SPRAY IN ANISOTROPICALLY TURBULENT GAS FLOW

Numerical investigation was conducted for a confined evaporating isopropyl alcohol spray issuing into a coflowing, heated turbulent air stream. The Eulerian-Lagrangian stochastic model was used for the spray calculations. The gas phase turbulence was modeled using either the isotropic eddy viscosity model or the second-moment transport model for both Reynolds stresses and heat fluxes. Two droplet dispersion models were studied for the Lagrangian trajectory calculations; the conventional particle-eddy encounter model and the time-correlated dispersion model. In the time-correlated model, gas phase turbulent velocity fluctuations were correlated temporally and directionally between two successive time steps in modeling the droplet dispersion. The droplet evaporation was accounted for by the infinite-conduction evaporation model, where the gas-film variable properties were considered using the one-third rule. Detailed numerical results of the liquid droplet phase, i.e., the droplet mean diameters, mass fluxe...

Efficient particle tracking algorithm for two-phase flows in geometries using curvilinear coordinates

Particle dispersion in turbulent carrier flows is described using a Eulerian-Lagrangian hybrid model The body-fitted coordinates are used for the finite-volume solution to the Eulerian equations, whereas the mesh-free Lagrangian formulation is used for the solution to the Lagrangian equations. A novel numerical approach is developed to locate whether a particle stays in an arbitrary Eulerian control volume and to handle particle-wall interactions. The developed approach is characterized by its robustness and high computational efficiency. Details of the approach are given for a two-dimensional case. However, it is straightforwardly extendible to three-dimensional two-phase flows. As a demonstration, numerical results are presented for a three-dimensional, particle-laden turbulent gas flow in an industrial gas meter. It is found that the developed approach offers much higher computational efficiency than the existing circular-search approach. As a result, the particle-locating procedure with an optimal sea...

Efficient computation of particle dispersion in turbulent flows with a stochastic-probabilistic model

Abstract This paper describes a three-dimensional, stochastic probabilistic, efficiency-enhanced dispersion (SPEED) model for the prediction of particle dispersion in turbulent flows. A stochastic procedure is used to compute particle-trajectory mean and variance, and a probabilistic procedure is used to compute a physical-particle spatial distribution among Eulerian control volumes. An additional Lagrangian equation is derived to govern the evolution of particle-trajectory variance. The SPEED model is aimed at tracking a relatively small number of particle trajectories, while efficiently reducing computational shot noise in the conventional stochastic dispersion model. Two tests cases with available measurements are used to validate the efficiency of the SPEED model. Numerical results of the SPEED model using only 5 × 102 particle trajectories are compared with those of the conventional stochastic model using as high as 1.5 × 104 particle trajectories. It is found that the SPEED model offers better agreement with the experimental measurements, and that the computational efficiency can be substantially enhanced by a factor of 20.

Numerical calculations of unsteady heavy gas dispersion

This paper concerns both near-field and far-field numerical predictions of liquefied gas releases in atmospheric environment. The near-field prediction was related to sudden depressurization of liquefied propane into atmospheric environment. Three phases of propane vapor, propane droplets, and entrained air were considered. Simplification was made that air and vapor have the same velocity and temperature but different volume fractions so that an air-vapor mixture phase could be assumed, and was treated using an Eulerian formulation. The droplet phase was handled using a Lagrangian formulation by which droplet trajectories were computed. A thin-skin evaporation model was used to account for droplet evaporation for the near-field prediction. Present numerical results for the near-field modelling were compared with those obtained with a twin-fluid Eulerian-Eulerian model. The far-field prediction was associated with heavy gas dispersion of Burro 8 LNG field test in a flat terrain. The conventional κ-e eddy viscosity-diffusivity model was modified to account for the anisotropy of turbulence characterized by heavy gas dispersion close to a ground. Numerical results were presented for the Burro 8 LNG field test. Results for the far field simulation were also compared with those obtained with the commercial code DEGADIS.

Computation of Particle-Laden Turbulent Gas Flows Using Two Dispersion Models

Ahybrid Eulerian‐ Lagrangianmodelwasused tostudy thesolidparticles,witha meandiameterof49 :3πmand a standard deviation of 4 :85πm, dispersing from a turbulent wall-adjacent gaseous jet. Two Lagrangian particle dispersion models, the conventional stochastic discrete delta function model and a recently developed stochasticprobabilistic efe ciency-enhanced dispersion (SPEED) model, were used to account for the particle dispersion induced by gas turbulence. The present work also investigated theeffects of two different methods for determining the particle‐ eddy interaction time on the predicted particle property. Numerical predictions obtained with the different models using the conventional and modie ed particle‐ eddy interaction timescales were compared with each other and with experimental measurements. It was found that betteragreementwith experimental data could be achieved by using the modie ed particle‐ eddy interaction timescale. Moreover, it was demonstrated that the SPEED model could not only signie cantly improvethe computational efe ciency of the trajectory solver by a factor of 10 for the present e ow considered but also yield smoother proe les and better agreement with the measurements than could the conventional Lagrangian stochastic discrete particle model.

Stochastic-Probabilistic Efficiency Enhanced Dispersion Modeling of Turbulent Polydispersed Sprays

A stochastic-probabilistic, efficiency enhanced dispersion (SPEED) model is developed for the prediction of turbulent two-phase flows. The SPEED model computes both the mean and variance of droplet positions at each Lagrangian integral time step. The mean position is determined with an improved conventional stochastic model, whereas the variance is determined by a newly derived Lagrangian equation with a Lagrangian autocorrelation function. A memoryless Markovian chain is used to determine the autocorrelation function. The distribution of a physical droplet in space is determined with a prescribed probability density function. The efficiency of the SPEED model is that a minimal number of droplet trajectories are required for Lagrangian trajectory computations during which a large amount of smooth noise-free solution can be attained. The developed SPEED model is first validated against a benchmark test where the measured mean-squared dispersion width is available. Then the results include the prediction of a polydispersed turbulent spray with detailed experunental measurements. Numerical results of the SPEED model, using only a total number of 6 x 10 2 droplet trajectories, are compared with those of a conventional stochastic discrete delta-function model using a total number of 2.1 x 10 4 trajectories, and with a previous stochastic dispersion-width transport model. It is found that the SPEED model is numerically more efficient than the dispersion-width transport model and needs much fewer number of droplet trajectories than the standard model.

Computation of particle dispersion in turbulent liquid flows using an efficient Lagrangian trajectory model

SUMMARY The dispersion of solid particles in a turbulent liquid flow impinging on a centrebody through an axisymmetric sudden expansion was investigated numerically using a Eulerian‐Lagrangian model. Detailed experimental measurements at the inlet were used to specify the inlet conditions for two-phase flow computations. The anisotropy of liquid turbulence was accounted for using a second-moment Reynold stress transport model. A recently developed stochastic‐probabilistic model was used to enhance the computational efficiency of Lagrangian trajectory computations. Numerical results of the stochastic‐probabilistic model using 650 particle trajectories were compared with those of the conventional stochastic discrete-delta-function model using 18 000 particle trajectories. In addition, results of the two models were compared with experimental measurements.

Computational modeling of dilute gas-particle flows in an ultrasonic gas flowmeter

Abstract A computational procedure is presented to predict particle-laden turbulent gas flows in an ultrasonic flowmeter using curvilinear coordinates. The Eulerian-Lagrangian hybrid model is used for the two-phase flow predictions. The turbulent gas phase is formulated using the Eulerian governing equations whereas the particle phase is formulated using the Lagrangian governing equations. The effect of the gas turbulence on discrete particle dispersion is accounted for using a particle-eddy interaction model. An efficient numerical algorithm is described to locate particles with an optimized search path in three-dimensional Eulerian control volumes. Numerical results are compared with experimental measurements for both the gas and particle phases. In addition, the sensitivity of particle deposition onto the meter-duct wall is investigated by using different particle diameters and densities.

Experimental and numerical study of a water spray in the wake of an axisymmetric bluff body

An experimental and numerical study was made of a water spray issuing into a recirculating flow behind a bluff body disk mounted in a nozzle. A two-component laser-Doppler/phase-Doppler anomometry system was used to characterize the mean and turbulent dispersed phase. The numerical prediction of the hollow-cone spray was based on an Eulerian-Lagrangian stochastic hybrid model. The continuous gas flow field was predicted using a differential Reynolds stress transport model whereas the particulate droplet flow field was predicted using an improved Lagrangian stochastic dispersion model. Comparison of numerical predictions and experimental measurements was carried out for droplet mean and fluctuating velocities, number mean diameters, and mass fluxes. Results indicate that the droplet dispersion characteristics are strongly influenced by the presence of flow recirculation due to different particle Stokes numbers, yielding recirculation or penetration of particles through the separated flow region. Predictions of droplet axial and radial rms velocities obtained with the Lagrangian stochastic model are less satisfactory than those of other simple gas spray flows.