Experimental and Numerical Analysis of Mixing Ventilation at Laminar, Transitional and Turbulent Slot Reynolds Numbers (Experimentele en numerieke analyse van mengventilatie bij laminaire, transitionele en turbulente slot-Reynoldsgetallen)

The proper ventilation of buildings and other enclosures such as airplanes, trains, ships and cars is of primary interest in engineering with respect to human (thermal) comfort, energy efficiency and sustainability.One of the most commonly applied ventilation methods is mixing ventilation, which is based on the injection of an air jet in the upper part of the room. The momentum of the jet should ensure mixing of the fresh supply air with the room air, and the diluted air should subsequently be extracted from the room. Although a lot of research has been conducted on mixing ventilation in the past decades, there are still several issues that are not resolved. The dissertation consists of two parts, both of which address current issues in mixing ventilation studies: (I) experimental and numerical work on isothermal transitional mixing ventilation in anidealized simplified reduced-scale model; (II) experimental and numerical work on mixing ventilation in a full-scale complex enclosure in an urban environment, driven by both wind and buoyancy. Both parts consist ofa combination of unique measurements, either full-scale or reduced-scale, and state-of-the-art Computational Fluid Dynamics (CFD) simulations. Part I A wide range of experimental andnumerical studies have been conducted in the past to analyze the flow patterns associated with ventilation in general and with forced mixing ventilation in particular. However, the vast majority of these ventilationstudies focused on fully turbulent flows (high Reynolds numbers). Low Reynolds (Re) numbers can indicate the presence of a transitional flow regime inside the room, which can be distinguished from turbulent flow by the presence of relatively large coherent structures (vortices). Severalpublications have indicated the fact that transitional flow can be present in different types of room airflow, either in the supply jet region or in other regions of low velocities (e.g. corners of the room, vicinity of buoyant plumes). However, to the best knowledge of the author, onlya limited number of studies has dealt with room airflow at transitionalslot Reynolds numbers so far, either experimentally or numerically. In addition, there is no consensus on the capabilities of CFD to predict transitional room airflow. To be able to come to a conclusion regarding the capability of steady Reynolds-averaged Navier-Stokes (RANS) CFD simulations to predict transitional room airflow, high-quality experimental data sets should be available. The lack of such a data set, and subsequently the lack of consensus on the capabilities of CFD to predict transitional room airflow are the primary reasons for the work performed on this topic for a mixing ventilation case, and which is presented in Chapters 2-5. In Chapter 2, a reduced-scale experimental setup to study ventilation flow at low Reynolds numbers (transitional flow) is presented. The reduced-scale model is used to perform Particle Image Velocimetry (PIV) measurements of mixing ventilation flow at transitional slot Reynolds numbers for a free plane jet. The inlet height for the studied configuration is h/L = 0.0667, with L the characteristic dimension of the cubic test section (L = 0.3 m). Flow visualizations show that the roomairflow is transitional for the range of studied slot Reynolds numbers (800 A second set of PIV measurementsof forced mixing ventilation flow is presented in Chapter 3. The experimental setup is to a large extent similar to the one presented in Chapter 2. However, the experiments presented in Chapter 3 are conducted for an inlet height h/L = 0.1, which corresponds to a plane wall jet issued from a smooth contraction. The PIV measurements focus on both the instantaneous and the time-averaged velocity and vorticity fields, as well as on the turbulence intensity. The vorticity profiles indicate a solid-bodyrotation in the large recirculation cell. The instantaneous vector fields show Kelvin-Helmholtz-type instabilities as a result of the large velocity gradient in the shear layer of the wall jet. The Strouhal number based on the vortex formation frequency is shown to increase with increasing Reynolds number. Application of the Okubo-Weiss function indicates the presence of vortical structures in the wall jet region and the presence of a vortex train in the outer region of the wall jet. Chapter 4 presents steady RANS CFD simulations of forced mixing ventilation at transitional slot Reynolds numbers. The experimental data set presented in Chapter 3 is used to assess the capability of four commonly used RANS turbulence models to predict transitional room airflow. Three popular linear two-equation models are tested (RNG k-ε, low-Re numberk-ε, SST k-ω), as well as one second-order closure model (Reynolds Stress Model (RSM)). Both the dimensionless velocities and the turbulent kinetic energies are compared on three vertical lines in the enclosure. The results show that three out of the four turbulence models provide results that are in close agreement with the measurement results. The results obtained with the RNG k-ε model show the largest deviations with the measurements, which can be attributed to an overprediction of turbulent kinetic energy in the wall jet region. In addition, it is shown that the different turbulence models provide different predictions for the air exchange efficiency, with differences between two models being as high as 44%. In addition to the correct prediction of the time-averaged flow pattern, it is of interest to see whethersteady RANS models can predict the dispersion of pollutants in a room with sufficient accuracy, which is the topic of Chapter 5. CFD simulations with steady RANS models often employ the standard gradient-diffusion hypothesis, in which the turbulent mass fluxes are related to the mean mass gradient using the turbulent (or eddy) mass diffusivity. The relativeinfluence of convective and turbulent mass fluxes in the transport process is analyzed and the role of these fluxes in the prediction accuracy of RANS and Large Eddy Simulations (LES) is clarified for this particular case. It is shown that the standard gradient-diffusion hypothesis is not always valid. However, the turbulent mass fluxes are about one order of magnitude smaller than the convective fluxes. As a result, the invalidity of the standard gradient-diffusion hypothesis does not lead to significant deviations in the predicted mean pollutant concentration field using steady RANS CFD simulations in the case under study. Part II A literature study has shown that well-documented experimental data sets of complex ventilation flow are hardly available. As a result, there is a strong lack of experimental data to validate numerical models for realistic/complex situations. Furthermore, CFD studies of natural mixing ventilation are usually performed for relatively simple building geometries. This part of the dissertation presents full-scale measurements of wind velocity and a range of environmental conditions in and around a complex semi-enclosed stadium situated in an urbanarea. The measurement results are used to validate a CFD model of the stadium and its surroundings, which is subsequently used to assess the natural mixing ventilation of the interior air volume of the stadium. Chapter 6 presents an analysis of full-scale measurements of thermal conditions and natural ventilation in a large semi-enclosed stadium in Amsterdam, the Netherlands. Due to similarity requirements (Reynolds, Grashof, and Richardson numbers) that cannot be fulfilled in the windtunnel, full-scale measurements are the only means to obtain a reliabledata set for a realistic summer situation. The full-scale measurements indicate a certain degree of repeatability on three consecutive evenings; both the wind conditions and the indoor and outdoor thermal conditionsonly show small differences between the three evenings. As a result, the measured CO2 concentration decay curves, and the calculated air exchange rate (ACH) values only show small deviations between the three evenings. Although there might be problems with repeatability and uncontrollable boundary conditions when performing full-scale measurements, in some particular cases, as the one presented here, full-scale measurements canprovide useful experimental data to validate CFD models of natural ventilation. Chapter 7 presents a coupled CFD modeling approach for urban wind flow and indoor natural ventilation of a large semi-enclosed stadium on a high-resolution grid. The computational grid is constructed using a specific procedure to efficiently and simultaneously generate the complex geometry and the high-resolution body-fitted grid for both the outdoor and indoor environment, based on translation and rotation of pre-meshed cross-sections. A grid-sensitivity study indicates that a 5.5 million cell grid provides nearly grid-independent results. The coupled CFD simulations are validated using full-scale (on-site) wind velocity measurements. The natural ventilation of the current configuration,as well as alternative ventilation configurations is analyzed. From theCFD simulations it is concluded that small geometrical modifications can increase the ACH values by up to 43%. A CFD analysis of the influence of wind direction and urban surroundings on the computed air exchange rate is presented in Chapter 8. The computational model of the stadium is the same as the current stadium configuration as studied in Chapter 7. To assess the influence of the wind direction and urban surroundings, simulations are performed for eight wind directions and for acomputational model with and without the surrounding buildings. The simulated differences in ACH between wind directions can be as high as 152%(with surrounding buildings). Furthermore, comparing the simulations with and without taking into account the urban surroundings for each wind direction shows that neglecting the surrounding buildings can lead to overestimations of the ACH with up to 96%. Finally, Chapter 9 presents non-isothermal unsteady RANS CFD simulations of CO2 concentration decay from the abovementioned semi-enclosed stadium. The boundary conditions for the CFD simulations are based on the measured conditions. The CO2 concentration decay curves obtained with the unsteady CFD simulations are compared with measured CO2 concentration decay curves and showa fair to good agreement. The validated model is used to detect regionswith lower ventilation efficiencies, i.e. stagnant regions and recirculation zones inside the stadium, resulting in higher CO2 concentrations. The largest spatial gradients are present in the beginning of the CO2 concentration decay process, and can be as high as 700 ppm (= 37%) betweenthe northern and southern part of the stadium. In addition, a specific piecewise linear technique is applied for the concentration decay methodto determine the ACH values based for smaller time intervals. This is important because the semi-logarithmic decay curve itself is not linear because the value of ACH changes over time as a result of decreasing buoyancy forces. Using this technique, it is shown that the ACH values strongly decrease as a function of time, from about 2 h-1 at the beginning ofthe concentration decay simulations to about 0.3 h-1 at the end (t > 4000 s). Chapter 10 provides a discussion on the research andrecommendations for future work are listed. Finally, Chapter 11 (summary and conclusion) concludes this dissertation.

Jaar van publicatie:2012

Toegankelijkheid:Open