<p>Due to strong interactions among various phases and among the phases and fluid motions, multiphase flows (MPFs) are so complex that lots of efforts have to be paid to predict its sequential patterns of phases and motions. The present paper applies the physical constraints inherent in MPFs and enforces them to a physics-informed neural network (PINN) model either explicitly or implicitly, depending on the type of constraints. To predict the unobserved order parameters (OPs) (which locate the phases) in the future steps, the conditional neural processes (CNPs) with long short-term memory (LSTM, combined as CNPLSTM) are applied to quickly infer the dynamics of the phases after encoding only a few observations. After that, the multiphase consistent and conservative boundedness mapping (MCBOM) algorithm is implemented the correction the predicted OPs from CNP-LSTM so that the mass conservation, the summation of the volume fractions of the phases being unity, the consistency of reduction, and the boundedness of the OPs are strictly satisfied. Next, the density of the fluid mixture is computed from the corrected OPs. The observed velocity and density of the fluid mixture then encode in a physics-informed conditional neural processes and long short-term memory (PICNP-LSTM) where the constraint of momentum conservation is included in the loss function. Finally, the unobserved velocity in future steps is predicted from PICNP-LSTM. The proposed physics-informed neural processes (PINPs) model (CNP-LSTM-MCBOM-PICNP-LSTM) for MPFs avoids unphysical behaviors of the OPs, accelerates the convergence, and requires fewer data. The proposed model successfully predicts several canonical MPF problems, i.e., the horizontal shear layer (HSL) and dam break (DB) problems, and its performances are validated.</p>

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第21723号 / 工博第4540号 / 新制||工||1708(附属図書館) / 京都大学大学院工学研究科社会基盤工学専攻 / (主査)教授 中川 一, 准教授 竹林 洋史, 准教授 川池 健司 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

No description available.

This thesis reports on investigations carried out to study of the effect of horizontal wall jets on rough, fixed and mobile beds in open channel flow. Experimental tests were carried out, using fixed and mobile sediment beds. Computer simulation models for the flow within the jet and resulting sediment transport were developed and their results analysed in this study. In the experimental phase, tests were carried out with both fixed and mobile sediment beds. The shape of the water surface, numerous point velocity measurements and measurements of the evolving scour hole shape were made. Detailed descriptions of the turbulent flow field over a fixed rough bed and for scour holes at equilibrium were obtained for a range of initial jet conditions. Fully turbulent, multiphase flow was modelled using the Fluent Computational Fluid Dynamics software. This was used to analyze the flow caused by a jet in a rectangle open-channel with a rough bed, and also the flow pattern in a channel with a local scour hole. The volume of fluid (VOF) multiphase method and K- model was used to model the fluid flow in both cases. The model predictions of velocity and shear stress were compared against experimental observations. The experimental data was used to develop new empirical relationships to describe the pattern of boundary shear stress caused by a wall jet over fixed beds and in equilibrium scour holes. These relationships were linked with existing bed-load transport rate models in order to predict the temporal evolution of scour holes. An analytical model describing the relationship between the wall jet flow and the development of a local scour hole shape was reported and its predictions compared with experimental data.

No description available.

<p>Understanding the drying of droplets is of interest for processes such as spray drying, where particulate materials are produced by evaporating moisture. Even though spray-drying is a widely used method, there are still challenges, such as undesired agglomeration or controlling the morphology and size of the final dried product. This dissertation develops a physics based model that is used to examine the droplet dynamics and drying kinetics at large and small scales.  In addition, the model simulates the internal motion of insoluble particles and  is used to better understand particle formation during spray drying type processes.</p> <p><br></p> <p>The first part of this work examines the effect of droplet-droplet collisions on evaporation and the size distribution at a large scale. Droplet collision dynamics are implemented into an Eulerian-Lagrangian framework, where droplets are tracked in the Lagrangian frame, and the background gas is modeled as a continuum. The modeling framework includes fully coupled interphase heat and momentum transfer between the droplet and gas phases. Binary collision of droplets could result in coalescence, reduction in surface area, or separation of droplets, resulting in the generation of satellite droplets and an increase in total surface area. By capturing the change in size distribution due to the collision of particles, our results show a linear relationship between the Weber number and the evaporation rate at low droplet number densities. Further, it is shown that droplet number density is a critical factor influencing the evaporation rate. At high droplet number densities, the relationship between the evaporation rate and the Weber number becomes non-linear, and at extremely high droplet number densities, the evaporation rate decreases even at high Weber numbers.</p> <p><br></p> <p>In the next part of this dissertation, the drying of a single droplet containing insoluble solid particles is investigated. Using a volume-of-fluid framework coupled with the Lagrangian phase, we study the particle transport within a droplet, and how it is affected by airflow, phase properties (e.g., viscosity and density of each phase), surface tension, and evaporation. Unlike the traditional one-dimensional modeling approach, our multi-dimensional model can capture the generation of internal flow patterns due to shear flow and the accumulation of solid particles on the surface of the drying droplet. Our results show that the surface tension effect is more pronounced at larger droplet diameters and low airflow velocities. Our approach also provides a quantitative method for modeling crust growth and formation. </p> <p>Our results show that increasing solids mass fraction, and decreasing particle diameter, slow down the internal transport of solid particles, leading to a more quick accumulation near the surface of the droplet. Further, despite the droplet undergoing a constant-rate drying stage, the accumulation of solids near the surface is non-linear. In addition, the inclusion of solids within the droplet drastically reshapes the formation of internal vortices compared to the uncoupled case, which determines solids distribution.</p>

No description available.

The collisions frequency of dispersed phases (particles, droplets, bubbles) in a turbulent carrier phase is a fundamental quantity that is needed for modeling multiphase flows with applications to chemical processes, minerals flotation, food science, and many other industries. In this dissertation, numerical simulations are performed to determine collisions frequency of bi-dispersed particles (solid particles and bubbles) in homogeneous isotropic turbulence. Both direct numerical simulations (DNS) and Large Eddy simulations (LES) are conducted to determine velocity fluctuations of the carrier phase. The DNS results are used to validate existing theoretical models as well as the LES results. The dissertation also presents a CFD-based flotation model for predicting the pulp recovery rate in froth flotation machines. In the direct numerical simulations work, particles and bubbles suspended in homogeneous isotropic turbulence are tracked and their collisions frequency is determined as a function of particle Stokes number. The effects of the dispersed phases on the carrier phase are neglected. Particles and bubbles of sizes on the order of Kolmogorov length scale are treated as point masses. Equations of motion of dispersed phases are integrated simultaneously with the equations of the carrier phase using the same time stepping scheme. In addition to Stokes drag, the pressure gradient in the carrier phase and added-mass forces are also included. The collision model used here allows overlap of particles and bubbles. Collisions kernel, radial relative velocity, and radial distribution function found by DNS are compared to theoretical models over a range of particle Stokes number. In general, good agreement between DNS and recent theoretical models is obtained for radial relative velocity for both particle-particle and particle-bubble collisions. The DNS results show that around Stokes number of unity particles of the same group undergo expected preferential concentration while particles and bubbles are segregated. The segregation behavior of particles and bubbles leads to a radial distribution function that is less than one. Existing theoretical models do not account for effects of this segregation behavior of particles and bubbles on the radial distribution function. In the large-eddy simulations efforts, the dissertation addresses the importance of the subgrid fluctuations on the collisions frequency and investigates techniques for predicting those fluctuations. The cases studied are of particles-particles and particles-bubbles collisions at Reynolds number Re_λ = 96. A study is conducted first by neglecting the effects of subgrid velocity fluctuations on particles and bubbles motions. It is found that around Stokes number of unity solid particles of the same group undergo the well known preferential concentration as observed in the DNS. Effects of pressure gradient on the particles are negligible due to their small sizes. Bubbles as a low inertia particles are very sensitive to subgrid velocity and acceleration fields where the effects of pressure gradient in the carrier phase are dominant. However, particle-bubble radial distribution functions from LES are not as low as that from DNS. To account for the effects of subgrid field on the dispersion of particles and bubbles, a new multifractal methodology has been developed to construct a subgrid vorticity field from the resolved vorticity field in frame work of LES. A Poisson's solver is used to obtain the subgrid velocity field from the subgrid vorticity field. Accounting for the subgrid velocity fluctuations (but neglecting pressure gradient) produced minor changes in the radial distribution function for particle-particle and particle-bubble collisions. We conclude from this study that for accurate particle tracking in LES the subgrid velocity fluctuations must be dynamically realizable field (temporally and spatially correlated with the large scale motion). Adding random SGS velocity fluctuations is not enough to capture the correct radial distribution functions of dispersed phases especially for bubbles-particles collisions where the pressure gradient term ( or acceleration Du_f′/Dt) is responsible for particle-bubble segregation around particle Stokes number near one. A CFD-based model for minerals flotation machines has been developed in this dissertation. The objective of flotation models is to predict the recovery rate of minerals from a flotation cell. The developed model advances the state-of-the-art of pulp recovery rate prediction by incorporating validated theoretical collisions frequency models and detailed hydrodynamics from two-phase flow simulations. Spatial distributions of dissipation rate and air volume fraction are determined by the two-phase hydrodynamic simulations. Knowing these parameters throughout the machine is essential in understanding the effectiveness of different components of flotation machine (rotor, stator or disperser, jets) on the flotation efficiency. The developed model not only predicts the average pulp recovery rate but also it indicates regions of high/low recovery rates. The CFD-based flotation model presented here can be used to determine the dependence of recovery rate constant at any locality within the pulp based on particle diameter, particle specfic gravity, contact angle, and surface tension. / Ph. D.

A finite element model is developed for multiphase flow through soil involving three immiscible fluids: namely air, water, and an organic fluid. A variational method is employed for the finite element formulation corresponding to the coupled differential equations governing the flow of the three fluid phase porous medium system with constant air phase pressure. Constitutive relationships for fluid conductivities and saturations as functions of fluid pressures which may be calibrated from two-phase laboratory measurements, are employed in the finite element program. The solution procedure uses iteration by a modified Picard method to handle the nonlinear properties and the backward method for a stable time integration. Laboratory experiments involving soil columns initially saturated with water and displaced by p-cymene (benzene-derivative hydrocarbon) under constant pressure were simulated by the finite element model to validate the numerical model and formulation for constitutive properties. Transient water outflow predicted using independently measured capillary head-saturation data agreed well with observed outflow data. Two-dimensional simulations are presented for eleven hypothetical field cases involving introduction of an organic fluid near the soil surface due to leakage from an underground storage tank. The subsequent transport of the organic fluid in the variably saturated vadose and ground water zones is analysed. / Ph. D.

The presence of gas in engine liquid cooling systems can have severe consequences for engine efficiency and life. The presence of stagnant, trapped gases will result in cooling system hotspots, causing gallery wall degradation through thermal stresses, fatigue and eventual cracking. The presence of entrained, transient gases in the coolant flow will act to reduce its bulk thermal properties and the performance of the system s coolant pump; critically the liquid flow rate, which will severely affect heat transfer throughout the engine and its ancillaries. The hold-up of gas in the pump s impeller may cause the dynamic seal to run dry, without lubrication or cooling. This poses both an immediate failure threat should the seal overheat and rubber components melt and a long term failure threat from intermittent quench cooling, which causes deposit formation on sealing faces acting to abrade and reduce seal quality. Bubbles in the coolant flow will also act as nucleation sites for cavitation growth. This will reduce the Net Positive Suction Head available (NPSHA) in the coolant flow, exacerbating cavitation and its damaging effects in locations such as the cylinder cooling liners and the pump s impeller. This thesis has analysed the occurrence of trapped gas (air) during the coolant filling process, its behaviour and break-up at engine start, the two-phase character of the coolant flow these processes generate and the effects it has on coolant pump performance. Optical and parametric data has been acquired in each of these studies, providing an understanding of the physical processes occurring, key variables and a means of validating numerical (CFD) code of integral processes. From the fundamental understanding each study has provided design rules, guidelines and validated tools have been developed, helping cooling system designers minimise the occurrence of trapped air during coolant filling, promote its breakup at engine start and to minimise its negative effects in the centrifugal coolant pump. It was concluded that whilst ideally the prevention of cooling system gases should be achieved at source, they are often unavoidable. This is due to the cost implications of finding a cylinder head gasket capable of completely sealing in-cylinder combustion pressures, the regular use of nucleate boiling regimes for engine cooling and the need to design cooling channel geometries to cool engine components and not necessarily to avoid fill entrapped air. Using the provided rules and models, it may be ensured stagnant air is minimised at source and avoided whilst an engine is running. However, to abate the effects of entrained gases in the coolant pump through redesign is undesirable due to the negative effects such changes have on a pump s efficiency and cavitation characteristics. It was concluded that the best solution to entrained gases, unavoidable at source, is to remove them from the coolant flow entirely using phase separation device(s).