Physics

We focus on a convolutional neural network (CNN), which has recently been utilized for fluid flow analyses, from the perspective on the influence of various operations inside the CNN considering some canonical regressions with fluid flow data. We consider two types of the CNN-based fluid flow analyses; 1. CNN metamodeling and 2. CNN autoencoder. For the first type of CNN with the additional scalar inputs, which is one of the common forms of CNN for fluid flow analysis, we investigate the influence of input placements in the CNN training pipeline. As an example, the estimation of force coefficients of laminar flows over a flat plate and two side-by-side cylinders are considered. We find that care should be taken for the placement of additional scalar inputs depending on the problems and the flows users handle. We then investigate the influence of various parameters and operations on CNN performance, with the utilization of autoencoder (AE). A two-dimensional turbulence is considered for the demonstration of AE. The results of AE highly rely on the decaying nature. The influence of padding operation at a convolutional layer is also investigated. The zero padding shows reasonable ability compared to other methods which account for boundary conditions of numerical data. Moreover, the effect of the dimensional reduction/extension methods inside CNN is also examined. The CNN model is robust to the dimension reduction operations, while being sensitive to the dimensional-extension methods. The findings through the paper can help us toward the practical uses of CNN-based fluid flow analyses.

Environmental pollutants are a source for reliability issues across data center and telecommunications equipment. A primary driver of this is the transport and deposition of particle matter (PM 2.5 , PM 10 ) on printed circuit boards, electronic components and heat exchange surfaces. This process is enhanced by turbulent air flows generated from cooling fans. Particle pollutants can persist after contemporary filtering, highlighting the importance of elucidating particle transport mechanisms and utilising this information to design robust equipment. This study investigates particle transport behaviour arising from axial fans operating under varied aerodynamic conditions. Transient, multi-phase numerical simulations were performed to model the flow of millions of microscale particles in air and determine their fate. Across a comprehensive range of fan operation conditions, from aerodynamic stall to free delivery, non-dimensional deposition velocities spanned an order of magnitude. Deposition profiles vary from monotonic to non-monotonic behaviour, influenced by local flow impingement, blade tip vortices, and shear velocity. A simple flow control solution that mitigates the factors influencing deposition has been demonstrated for equipment already deployed. The findings and numerical methods can be applied for the optimization of fan-cooled equipment intended for indoor and outdoor environments where air quality is poor, or pollution levels are high.

We propose a new class of wicks, consisting of closely packed circular rods, whose evaporative capacities have been measured at different heat loads ranging between 0.05W/cm^2 and 8W/cm^2. The experiments were performed with two different liquids, water and highly volatile pentane, in a specially designed setup to understand transient and steady state cooling characteristics of the proposed wicks. Heat interception and vapour release occur on the same side in these experiments. These wicks released ~50% of the supplied heat load as the latent heat; this value remained nearly constant between the explored heat loads. These wicks have the unique characteristic of potentially very high and rapid capillary rise induced by near-zero radii (NZR) of contacts formed between the rods in contact; liquid region reaching the end in NZR has been called corner meniscus. While the bulk liquid (present between three rods) may recede, depending on the heat load, the corner meniscus remains pinned; this unique feature thus leads to sustained high evaporation rate requirements. This remarkable characteristic seems advantageous compared to a regular wick, whose cooling performance depends on the heat loads. We also performed 3-D unsteady state numerical simulations to understand the effect of rod diameter and materials' thermal conductivity on the overall wick's performance. Steady state temperature value was in fairly good agreement with the ones measured in experiments. Based on the geometry of the corner film, fluid mechanics of liquid transport, and the heat transfer aspects, we present a design of suitable wicks as per the requirement. These new configurations can represent a separate class of wicks and may replace the regular wicks in current and futuristic cooling devices.

Miniature heaters are immersed in flows of quantum fluid and the efficiency of heat transfer is monitored versus velocity, superfluid fraction and time. The fluid is 4 He helium with a superfluid fraction varied from 71% down to 0% and an imposed velocity up to 3 m/s, while the characteristic sizes of heaters range from 1.3 μ m up to few hundreds of microns. At low heat fluxes, no velocity dependence is observed. In contrast, some velocity dependence emerges at larger heat flux, as reported previously, and three non-trivial properties of heat transfer are identified. First, at the largest superfluid fraction (71%), a new heat transfer regime appears at non-null velocities and it is typically 10% less conductive than at zero velocity. Second, the velocity dependence of the mean heat transfer is compatible with the square-root dependence observed in classical fluids. Surprisingly, the prefactor to this dependence is maximum for an intermediate superfluid fraction or temperature (around 2 K). Third, the heat transfer time series exhibit highly conductive short-lived events. These \textit{cooling glitches} have a velocity-dependent characteristic time, which manifest itself as a broad and energetic peak in the spectrum of heat transfer time series, in the kHz range. After showing that the velocity dependence can be attributed to the breaking of superfluidity within a thin shell surrounding heaters, an analytical model of forced heat transfer in a quantum flow is developed to account for the properties reported above. We argue that large scale flow patterns must form around the heater, having a size proportional to the heat flux (here two decades larger than the heater diameter) and resulting in a turbulent wake. The observed spectral peaking of heat transfer is quantitatively consistent with the formation of a Von Kármán vortex street in the wake of a bluff body.

To understand how internal flow structures manifest themselves in the global heat transfer, we study the correlation between different flow modes and the instantaneous Nusselt number ( Nu ) in a two-dimensional square Rayleigh-Bénard convection cell. High-resolution and long-time direct numerical simulations are carried out for Rayleigh numbers between 10 7 and 10 9 and a Prandtl number of 5.3. The investigated Nusselt numbers include the volume-averaged N u vol , the wall-averaged N u wall , the kinetic energy dissipation based N u kinetic , and the thermal energy dissipation based N u thermal . The Fourier mode decomposition and proper orthogonal decomposition are adopted to extract the coherent flow structure. Our results show that the single-roll mode, the horizontally stacked double-roll mode, and the quadrupolar flow mode are more efficient for heat transfer on average. In contrast, the vertically stacked double-roll mode is inefficient for heat transfer on average. The volume-averaged N u vol and the kinetic energy dissipation based N u kinetic can better reproduce the correlation of internal flow structures with heat transfer efficiency than that of the wall-averaged N u wall and the thermal energy dissipation based N u thermal , even though these four Nusselt numbers give consistent time-averaged mean values. The ensemble-averaged time trace of Nu during flow reversal shows that only the volume-averaged N u vol can reproduce the overshoot phenomena that is observed in the previous experimental study. Our results reveal that the proper choice of Nu is critical to obtain a meaningful interpretation.

In this work, we derive a model to investigate the generic thermoacoustic behavior of an idealized can-annular combustor containing N cans. We start from the acoustic wave equation, which simplifies to a Helmholtz equation in the frequency domain. By projecting this equation onto the dominant natural eigenmode of a single can, we derive a symmetric system of N coupled ordinary differential equations (ODEs) in the frequency domain for the dynamics of the dominant modal amplitudes. Assuming perfect symmetry, we use a Bloch wave ansatz to reduce this system to an equivalent single ODE in the frequency domain. The acoustic pressure fields in the cans are coupled at the annular turbine inlet. To model the effect of mean flow in the cans on the acoustic coupling, we use Howe's model for the Rayleigh conductivity of a rectangular aperture under turbulent grazing flow. The resulting low-order model allows us to study the influence of physical parameters such as natural eigenfrequency, the mean flow speed, the aperture width, the base linear growth rate and the can spacing on the frequency spectrum of an idealized can-annular combustor. We show that, depending on the values of the system parameters, the acoustic coupling can suppress or amplify thermoacoustic instabilities, raising the potential for instabilities in nominally stable systems.

Presented is a data-driven Machine Learning (ML) framework for the identification and modeling of traveling wave spatiotemporal dynamics. The presented framework is based on the steadily-propagating traveling wave ansatz, u(x,t)=U(ξ=x?�ct+a) . For known evolution equations, this coordinate transformation reduces governing partial differential equations (PDEs) to a set of coupled ordinary differential equations (ODEs) in the traveling wave coordinate ξ . Although traveling waves are readily observed in many physical systems, the underlying governing equations may be unknown. For these instances, the traveling wave ODEs can be (i) identified in an interpretable manner through an implementation of sparse regression techniques or (ii) modeled empirically with neural ODEs. Presented are these methods applied to several physical systems that admit traveling waves. Examples include traveling wave fronts, pulses, and wavetrains restricted to one-wave wave propagation in a single spatial dimension.

The state-of-the-art in wind-farm flow-physics modeling is Large Eddy Simulation (LES) which makes accurate predictions of most relevant physics, but requires extensive computational resources. The next-fidelity model types are Reynolds-Averaged Navier-Stokes (RANS) which are two orders of magnitude cheaper, but resolve only mean quantities and model the effect of turbulence. They often fail to accurately predict key effects, such as the wake recovery rate. Custom RANS closures designed for wind-farm wakes exist, but so far do not generalize well: there is substantial room for improvement. In this article we present the first steps towards a systematic data-driven approach to deriving new RANS models in the wind-energy setting. Time-averaged LES data is used as ground-truth, and we first derive optimal corrective fields for the turbulence anisotropy tensor and turbulence kinetic energy (t.k.e.) production. These fields, when injected into the RANS equations (with a baseline k−ϵ model) reproduce the LES mean-quantities. Next we build a custom RANS closure from these corrective fields, using a deterministic symbolic regression method to infer algebraic correction as a function of the (resolved) mean-flow. The result is a new RANS closure, customized to the training data. The potential of the approach is demonstrated under neutral atmospheric conditions for multi-turbine constellations at wind-tunnel scale. The results show significantly improved predictions compared to the baseline closure, for both mean velocity and the t.k.e. fields.

The need for single-camera 3D particle tracking methods is growing, among others, due to the increasing focus in biomedical research often relying on single-plane microscopy imaging. Defocusing-based methods are ideal for a wide-spread use as they rely on basic microscopy imaging rather than requiring additional non-standard optics. However, a wide-spread use has been limited by the lack of accessible and easy-to-use software. DefocusTracker is an open-source toolbox based on the universal principles of General Defocusing Particle Tracking (GDPT) relying solely on a reference look-up table and image recognition to connect a particle's image and its respective out-of-plane depth coordinate. The toolbox is built in a modular fashion, allowing for easy addition of new image recognition methods, while maintaining the same workflow and external user interface. DefocusTracker is implemented in MATLAB, while a parallel implementation in Python is in the preparation.

We study the dynamics of proppants carried by fluid driven into an evolving penny-shaped fracture. The behaviour of the slurry flow is investigated in two phases: pressurised injection and elastic closure. During injection the slurry is modelled using a frictional rheology that takes into account the shear-induced migration and jamming of the proppants. Making pragmatic assumptions of negligible toughness and cross-fracture fluid slip, we find self-similar solutions supporting a range of proppant concentration profiles. In particular, we define an effective viscosity, which equates the fracture evolution of a slurry flow with a given proppant volume fraction, to a Newtonian flow with a particular viscosity. Using this framework, we are able to make predictions about the geometry of the growing fracture and the significance of tip screen-out. In the closure phase, proppants are modelled as incompressible and radially immobile within the narrowing fracture. The effects of proppant concentration on the geometry of the residual propped fracture are explored in full. The results have important applications to industrial fracking and geological dike formation by hot, intruding magma.