An experimental and numerical investigation of heat transfer enhancement in a three dimensional channel using wall mounted porous baffles was conducted. The module average heat transfer coefficients were measured in a uniformly heated rectangular channel with staggered positioned porous baffles. A numerical procedure was implemented, in conjunction with a commercially available Navier-Stokes solver, to model the turbulent flow in porous media. The Brinkman-Forchheimer-Extended Darcy model was used for modeling fluid flow through the porous baffles. Conventional, oneequation, and two-equation models were used for heat transfer modeling. The accuracy and characteristics of each model were investigated and discussed. The results were compared with experimental data. Baffles were mounted alternatively on the top and bottom walls. Heat transfer coefficients and pressure loss for periodically fully developed flow and heat transfer were obtained for different pore densities (10, 20, and 40 pores per inch (PPI)) with two different baffle heights ( / h h B D = 1/3 and 2/3), and two baffle thicknesses ( / t h B D = 1/3 and 1/12). The Reynolds number (Re) was varied from 20,000 to 50,000. To compare the effect of foam metal baffles, the data for conventional solid-type baffles was obtained for ( / t h B D =1/3). The maximum uncertainties associated with the module Nusselt number and friction factor were 5.8% and 4.3%, respectively. The experimental procedure was validated by comparing the data for the straight channel without baffles ( / h h B D = 0) with those in the literature. The use of porous baffles resulted in heat transfer enhancement as high as 300% compared to heat transfer in straight channels without baffles. However, the heat transfer enhancement per unit increase in pumping power was less than one for the range of parameters studied in this work. Correlation equations were developed for the heat transfer enhancement ratio and the heat transfer enhancement per unit increase in pumping power in terms of Reynolds number. The conventional theoretical model, the dispersion conductivity model, and the modified two-phase model using the local thermal non-equilibrium theory were considered. The results from each model were compared against the experimental data, and compared to each other to investigate the efficiency of each model. Also, the characteristics of each model were discussed.

Study of passive dispersion of a neutral scalar in turbulentflows is highly important due to its numerous applications in the areas of turbulent flow visualization, turbulent heat transfer and transport of pollutants and other substances in the environment. Over the past few decades, many analytical, numerical, and experimental studies have been conducted on this topic to obtain a better understanding of the physical process. In the present work, Large Eddy Simulations (LES) of scalar dispersion in turbulent flow over smooth and rough channels is conducted to contribute to the further understanding of the relation between the turbulent velocity field and the concentration field. The LES results from the present work showed good agreement with a recently com-pleted experimental study(Rahman and Webster [2005]). An in-depth comparison of in-stantaneous concentration and velocity fields revealed thecorrelation between scalar dis-persion and coherent structures of the turbulent flow. Also,a three dimensional visual-ization of concentration iso-surfaces at different instants provided a good picture of the concentration structures transported as a result of hairpin vortices of turbulent flow, which is quite difficult to accomplish using experimental studies.

No description available.

This thesis presents a computational analysis of multi-element guard-heated sensors designed to overcome the most severe limitation of conventional thermal sensors for wall shear stress (WSS) measurement in turbulent flows –that of indirect heat conduction through the substrate. The objectives of this thesis are the study of guard-heated sensors {i} to quantify the reduction, over conventional single-element sensors, of substrate heat conduction losses and resultant errors over a range of applied shear and {ii} to examine a range of values of guard heater geometric parameters, in two common fluids, air and water and identify the best designs. Wall-turbulence, the turbulent flow in the vicinity of solid boundaries, has proved difficult to model accurately, due to the lack of accurate WSS measurements. Examples of areas of impact are drag force reduction on transport vehicles in land, sea, air, which today largely translate to reduced fossil fuel use and dependence; aerodynamic noise and control for flight and for wind energy conversion; atmospheric and oceanic transport studies for weather, climate and for pollutant transport; riverbank erosion. Constant-temperature anemometry with MEMS devices, flush-mounted hot-film thermal sensors, is non-intrusive, affords the best temporal resolution and is well-established. However, these hot-film probes suffer from unwanted heat transport to the fluid through the substrate, with errors and nonlinearity large enough to overwhelm quantitative utility of the data. Microfabrication techniques have enabled multi-element guard-heated prototypes to be fabricated. Our results show that errors in sensing-element signals, contributing to spectral distortion, are sensitive to sensor location within the guard heater. These errors can be reduced to below 1% of the signal with proper location of the sensor. Guard heating also reduces the large variation in spatial averaging due to substrate conduction. This makes them suitable for turbulent flows with a large range of fluctuations. / Graduate

- / -

The distance required for flow entering a laboratory channel to become fully-developed and uniform can be substantial. Given the need to establish fully-developed uniform flow, if the length of a laboratory channel is not substantial then it likely that the flume cannot be used to conduct open-channel flow research. In laboratory studies where the channel bed is hydraulically rough, the noted problem can be lessened by minimizing the length over which the flow becomes fully-developed and uniform (Lunif). For this study it is hypothesized that if bed material with a roughness height (ks, ∆) is placed at the channel entrance and ks, ∆ is greater has the roughness height of bed material placed throughout the channel (ks, bed) then Lunif can be reduced. The length over which the larger bed material is referred to as the tripping zone length (∆). A second hypothesis for this study is that if ∆ is longer, then Lunif will be shorter. The primary objective of this study is to test the above mentioned hypothesis and to develop a relationship for predicting Lunif as a function of Δ. For this study, physical tests were performed in a rectangular Plexiglas flume with a variable slope. The flume was 6.1 m long, 45.7 cm wide, and 45.7 cm deep. The channel has smooth walls and the bed was lined with gravel (median particle size, d50 = 8.5 mm or 22 mm). Similarly tripping zone was lined with gravel of larger size (median particle size, d50 = 13 mm or 58 mm).Twelve tests were conducted for the study. For each test, longitudinal point velocity measurements (u) were made along the channel center, at five elevations (z), and at twelve longitudinal stations (x). An Acoustic Doppler Velocimeter was used to measure u. Lunif was determined by considering four indications of flow uniformity. Results indicate that having a tripping zone decreases Lunif and the magnitude of the decrease in Lunif was dependent on ∆. A function is presented for predicting Lunif /H = f (Rep, Fr, and Δ/H) where Rep is the Reynold's particle number, Fr is the Froude number and H is the flow depth.

Orientador: Erick de Moraes Franklin / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecânica / Made available in DSpace on 2018-08-27T18:01:24Z (GMT). No. of bitstreams: 1 CardonaFlorez_JorgeEduar_M.pdf: 8246209 bytes, checksum: 9860a8456aed83e182b9580704940549 (MD5) Previous issue date: 2015 / Resumo: Na natureza e na indústria é comum encontrar fluidos em movimento transportando grãos quando escoam sobre superfícies granulares. Por exemplo, na natureza o transporte de areia por rios e mares quando o fluido é um liquido, ou na formação e migração de dunas nos desertos quando o fluido é um gás, são casos bem conhecidos. Na indústria apresenta-se por exemplo, na extração e transporte de petróleo nos oleodutos, assim como em variados processos químico-industriais envol- vidos com sedimentação, ou simplesmente por utilização de matérias-primas impuras. O fenômeno apresenta-se porque o fluido em escoamento transfere parte da quantidade de movimento à superfí- cie granular, e alguns grãos são postos em movimento mantendo contato com o fundo se a força de arrasto devida ao movimento do fluido supera levemente à força de resistência devida ao peso dos grãos. Os grãos em movimento formam uma camada móvel (leito móvel ou bed-load em inglês), tornando-se instável em alguns casos, gerando rugas e/ou dunas que interagem com o escoamento. Na indústria este fenômeno está associado principalmente a problemas de perda de carga, e a tran- sientes de pressão e vazão, quando fluidos misturados com material granular são transportados por dutos. O presente trabalho faz um estudo experimental da formação e migração deste tipo de ins- tabilidades quando escoamentos turbulentos de água a diferentes velocidades são impostos sobre leitos granulares inicialmente planos de diferentes granulometrias. Os experimentos foram feitos em um canal retangular de material transparente. A evolução do leito foi filmada por uma câmera de alta definição. O comprimento de onda e a celeridade de rugas formadas foram determinados mediante o desenvolvimento de um código numérico em Matlab, utilizado para o tratamento au- tomático das imagens obtidas nos experimentos. Os resultados são comparados com os de outros trabalhos publicados na literatura / Abstract: In nature and industry, it is common to find moving fluids carrying grains, when they flow over granular surfaces. For instance, in nature it can be found in the sand transported by rivers and ocean currents, or on the formation and migration of sand dunes in deserts by air; and in industry, it is common to find them in the extraction and transport of oil through pipelines, as well as in several chemical processes involving granular material. This phenomenon arises because part of the momentum is transferred from the fluid flow to the granular surface. Therefore, if the drag force, resulting from the movement of the fluid, exceeds slightly the resistance force occasioned by the weight of the grains, some of these particles are set in motion keeping in touch with the bottom. Consequently, the moving grains form a mobile layer (bed-load) that becomes unstable in some cases producing ripples and/or dunes, which interact with the flow. In industry, this is mainly associated with problems of pressure loss in closed conduits. This work presents an experimental study on the formation and migration of ripples in a turbulent flow. In the experiments, fully-developed turbulent water flows were imposed over a granular bed of known granulometry in a transparent closed-conduit. For different grain sizes and flow rates, a high-definition camera filmed the growth and migration of bedforms. The wavelength and the celerity of bedforms were determined from the acquired images through a numerical code developed in Matlab. The obtained results are compared with other published works in the area / Mestrado / Termica e Fluidos / Mestre em Engenharia Mecânica

A dynamic hybrid Reynolds-averaged Navier-Stokes (RANS)-Large Eddy Simulation (LES) modeling framework has been investigated and further developed to improve the Computational Fluid Dynamics (CFD) prediction of turbulent flow features along with laminar-to-turbulent transitional phenomena. In recent years, the use of hybrid RANS/LES (HRL) models has become more common in CFD simulations, since HRL models offer more accuracy than RANS in regions of flow separation at a reduced cost relative to LES in attached boundary layers. The first part of this research includes evaluation and validation of a dynamic HRL (DHRL) model that aims to address issues regarding the RANS-to-LES zonal transition and explicit grid dependence, both of which are inherent to most current HRL models. Simulations of two test cases—flow over a backward facing step and flow over a wing with leading-edge ice accretion—were performed to assess the potential of the DHRL model for predicting turbulent features involved in mainly unsteady separated flow. The DHRL simulation results are compared with experimental data, along with the computational results for other HRL and RANS models. In summary, these comparisons demonstrate that the DHRL framework does address many of the weaknesses inherent in most current HRL models. Although HRL models are widely used in turbulent flow simulations, they have limitations for transitional flow predictions. Most HRL models include a fully turbulent RANS component for attached boundary layer regions. The small number of HRL models that do include transition-sensitive RANS models have issues related to the RANS model itself and to the zonal transition between RANS and LES. In order to address those issues, a new transition-sensitive HRL modeling methodology has been developed that includes the DHRL methodology and a physics-based transition-sensitive RANS model. The feasibility of the transition-sensitive dynamic HRL (TDHRL) model has been investigated by performing numerical simulations of the flows over a circular cylinder and a PAK-B airfoil. Comparisons with experimental data along with computational results from other HRL and RANS models illustrate the potential of TDHRL model for accurately capturing the physics of complex transitional flow phenomena.

The discrete-element method for predicting skin friction for turbulent flow over rough surfaces considers the drag on the surface to be the sum of the skin friction on the flat part of the surface and the drag on the individual roughness elements that protrude into the boundary layer. The discrete-element method considers heat transfer from a rough surface to be the sum of convection through the fluid on the flat part of the surface and the convection from each of the roughness elements. The discrete-element method has been widely used and validated for roughness composed of sparse, ordered, and deterministic elements. Modifications made to the discrete-element roughness method to extend the validation to real surface roughness are detailed. These modifications include accounting for the deviation of the roughness element cross sections from circular configurations, determining the location of the computational "surface" that differs from the physical surface, and accounting for temperature changes along the height of the roughness elements. Two randomly-rough surfaces found on high-hour gas-turbine blades were characterized using a Taylor-Hobson Form Talysurf Series 2 profilometer. A method for using the three-dimensional profilometer output to determine the geometry input required in the discrete-element method for randomly-rough surfaces is presented. Two randomly-rough surfaces, two elliptical-analog surfaces, and two cone surfaces were generated for wind-tunnel testing using a three-dimensional printer. The analog surfaces were created by replacing each random roughness element from the original randomly-rough surface with an elliptical roughness element with the equivalent planorm area and eccentricity. The cone surfaces were generated by placing conical roughness elements on a flat plate to create surfaces with equivalent values of centerline-averaged height or root-mean-square (RMS) height as the randomly-rough surfaces. The results of the wind tunnel skin friction coefficient and Stanton number measurements and the discrete-element method predictions for each of the six surfaces are presented and discussed. For the randomly-rough surfaces studied, the discrete-element method predictions are within 7% of the experimentally measured skin friction coefficients. The discrete-element predictions are within 16% of the experimentally measured Stanton numbers for the randomly-rough surfaces.

No description available.