Fluids in computer generated imagery can add an impressive amount of realism to a scene, but are particularly time-consuming to simulate. In an attempt to run fluid simulations in real-time, recent efforts have attempted to simulate fluids by using machine learning techniques to approximate the movement of fluids. We explore utilizing machine learning to simulate fluids while also integrating the Fluid-Implicit-Particle (FLIP) simulation method into machine learning fluid simulation approaches.

Extracorporeal Membrane Oxygenation (ECMO) is a life-saving therapy usedfor support in critical heart and/or lung failure. Patient’s blood is pumped viaan artificial lung for oxygenation outside of the body. The circuit is composedof a blood pump, cannulae for drainage and reinfusion, a membrane lung,tubing and connectors. Its use is associated with thromboembolic complicationsand hemolytic damage. Detailed numerical studies of two blood pumps anda lighthouse tip drainage cannula were undertaken to characterize the flowstructures in different scenarios and their link to platelet activation. The pumpsimulations were modelled according to manufacturer’s proclaimed use but alsoin off-design conditions with flow rates used in adult and neonatal patients.Lagrangian Particle Tracking (LPT) was used to simulate the injection ofparticles similar in size to platelets to compute platelet activation state (PAS).The results indicated that low flow rates impacted PAS similarly to high flowrates due to increased residence time leading to prolonged exposure to shearstress despite the fact that shear per se was lower at low flow rate. Regardingthe cannula, the results showed that a flow pattern similar to a jet in crossflowdeveloped at the side holes. A parameter study was conducted to quantifydrainage characteristics in terms of flow rate distribution across the holes wheninput variables of flow rate, modelled fluid, and hematocrit were altered. Thefindings showed, across all the cases, that the most proximal hole row drainedthe largest fraction of fluid. The effects due to the non-Newtonian nature ofblood were confined to regions far from the cannula holes and the flow structuresshowed very limited dependence on the hematocrit. A scaling law was found tobridge the global drainage performance of fluid between water and blood. / <p>QC 210906</p>

Inviscid fluid inside thin pipe system is first analytically solved for eigenfrequencies and eigenmodes using Modal Interaction Model method with fluid-structure interaction condition at boundary. Shear-diaphragm boundary condition is used for comparing and validating Analytic results with Simulation using COMSOL Multiphysics. Effect of viscosity is also compared using Newtonian fluid model. Experiment is performed using simple pipe geometry and fluid to measure transfer accelerance which is post-processed to extract cirumferential modes up to order 4; this is used to compare and validate Experiemental results with Simulation. Good correlation is obtained between Analytic, Experiment and Simulation results with n=0 breathing modes requiring modification of governing equations to incorporate compressibility effects due to changing pipe cross-section area. / En analytisk lösning för egenfrekvenser och egenmoder för en icke-viskös fluid inuti ett tunt rörsystem är först framtagen med användning av en modbaserad modell för interaktion mellan fluid och struktur som randvillkor. Idealiserad randvillkor används för att jämföra och validera analytiska resultat med simulationer i COMSOL Multiphysics. Effekten av viskositet jämförs också med hjälp av en Newtonsk fluidmodell. Experiment genomförs med simpel rörgeometri samt fluid för att mäta acceleransen som är analyserad för till att få ut mo-der i omkretsled upp till fjärde ordningen; detta i sin tur används för att jämföra och validera de experimentella resultaten med simulering-ar. Det erhålls bra korrelation mellan de analytiska-, simulerade- samt experimentella resultaten. Undantaget för n=0 grundmoder då krävs modifikation av differentialekvationerna till att inkorporera kompressibilitetseffekter

Climate change is an evermore urging existential treat to the human enterprise. Mean temperature and greenhouse gas emissions have in-creased exponentially since the industrial revolution. But solutions are also mushrooming with exponential pace. Renewable energy technologies, such as wind and solar power, are deployed like never before and their costs have decreased significantly. In order to allow for further transformation of the energy system these technologies must be refined and optimised. In wind energy one important field with high potential of refinement is aerodynamics. The aerodynamics of wind turbines constitutes one challenging research frontier in aerodynamics today.   In this study, a novel approach for calculating wind turbine flow is developed. The approach is based on the partially parabolic Navier-Stokes equations, which can be solved computationally with higher efficiency as compared to the fully elliptic version. The modelling of wind turbine thrust is done using actuator-disk theory and the torque is modelled by application of the Joukowsky rotor. A validation of the developed model and force implementation is conducted using four different validation cases.   In order to provide value for industrial wind energy projects, the model must be extended to account for turbulence (and terrain in case of onshore projects). Possible candidates for turbulence modelling are parabolic k-ε and explicit Reynolds stress turbulence models. The terrain could possibly be incorporated consistently with the used projection method by altering the finite difference grid layout.

Although it is possible to directly solve an entire flow-acoustics problem in one computation, this approach remains prohibitively large in terms of the computational resource required for most practical applications. Aeroacoustic problems are therefore usually split into two parts; one consisting of the source computation and one of the source propagation. Although both these parts entail great challenges on the computational method, in terms of accuracy and efficiency, it is still better than the direct solution alternative. The source usually consists of highly turbulent flows, which for most cases will need to be, at least partly, resolved. Then, acoustic waves generated by these sources often have to be propagated for long distances compared to the wavelength and might be subjected to scattering by solid objects or convective effects by the flow. Numerical methods used solve these problems therefore have to possess low dispersion and dissipation error qualities for the solution to be accurate and resource efficient. The wave expansion method (WEM) is an efficient discretization technique, which is used for wave propagation problems. The method uses fundamental solutions to the wave operator in the discretization procedure and will thus produce accurate results at two to three points per wavelength. This thesis presents a method that uses the WEM in an aeroacoustic context. Addressing the propagation of acoustic waves and transfer of sources from flow to acoustic simulations. The proposed computational procedure is applied to a co-rotating vortex pair and a cylinder in cross-flow. Overall, the computed results agree well with analytical solutions. Although the WEM is efficient in terms of the spatial discretization, the procedure requires that a Moore-Penrose pseudo-inverse is evaluated at each unique node-neighbour stencil in the grid. This evaluation significantly slows the procedure. In this thesis, a method with a regular grid is explored to speed-up this process. / <p>QC 20161121</p>