Search

A numerical study of turbulent combustion systems was pursued to examine different computational modeling techniques, namely computational fluid dynamics (CFD) and chemical reactor network (CRN) methods. Both methods have been studied and analyzed as individual techniques as well as a coupled approach to pursue better understandings of the mechanisms and interactions between turbulent flow and mixing, ignition behavior and pollutant formation. A thorough analysis and comparison of both turbulence models and chemistry representation methods was executed and simulations were compared and validated with experimental works. An extensive study of turbulence modeling methods, and the optimization of modeling techniques including turbulence intensity and computational domain size have been conducted. The final CFD model has demonstrated good predictive performance for different turbulent bluff-body flames. The NOx formation and the effects of fuel mixtures indicated that the addition of hydrogen to the fuel and non-flammable diluents like CO2 and H2O contribute to the reduction of NOx. The second part of the study focused on developing chemical models and methods that include the detailed gaseous reaction mechanism of GRI-Mech 3.0 but cost less computational time. A new chemical reactor network has been created based on the CFD results of combustion characteristics and flow fields. The proposed CRN has been validated with the temperature and species emission for different bluff-body flames and has shown the capability of being applied to general bluff-body systems. Specifically, the rate of production of NOx and the sensitivity analysis based on the CRN results helped to summarize the reduced reaction mechanism, which not only provided a promising method to generate representative reactions from hundreds of species and reactions in gaseous mechanism but also presented valuable information of the combustion mechanisms and NOx formation. Finally, the proposed reduced reaction mechanism from the sensitivity analysis was applied to the CFD simulations, which created a fully coupled process between CFD and CRN, and the results from the reduced reaction mechanism have shown good predictions compared with the probability density function method.

The goal of this study is to validate the accuracy of the SC/Tetra CFD code by modeling four cases: a 2D zero pressure gradient flat plate, an axisymmetric separated boundary layer, flow over an NACA 0012 airfoil, and flow over an NACA 4412 airfoil trailing edge separation. The results will be compared to experimental data and CFL3D results. This study also investigates the sensitivity of SC/Tetra results to different set up details such as mesh resolution, turbulence model, and under-relaxation settings. Automation using the Visual Basic programming language was used to vary these set up details and efficiently process a large number of simulations for NACA 0012. The Flat plate case focuses on skin friction coefficient and velocity profiles. The axisymmetric separated boundary layer case concentrates on pressure coefficient and velocity profiles. For the NACA 0012 airfoil, evaluation focuses on Cp profiles. Cp and velocity profiles and streamlines are considered for the NACA 4412 airfoil. In general, coarser meshes are shown to be less accurate than finer meshes and the SST k-OMG is more accurate.

A major challenge in the design of turbomachinery components for aircraft gas turbine engines is high cycle fatigue failures due to flutter. Of particular concern is the subsonic/transonic stall flutter boundary which occurs at part speed near the stall line. At these operating conditions the incidence angle is large and the relative Mach number is high subsonic or transonic. Viscous effects dominate for high incidence angles. In order to predict the flutter phenomena, accurate calculation of the steady and unsteady aerodynamic loading on the turbomachinery airfoils is necessary. The development of unsteady aerodynamic models to predict the unsteady forces and moments acting on turbomachine airfoils is an area of fundamental research interest. Unsteady Reynolds Averaged Navier-Stokes (RANS) models have been developed to accurately account for viscous effects. For these Reynolds averaged equations turbulence models are needed for the Reynolds stress terms. A transition model is also necessary. The transition onset location is determined by a transition onset model or specified at the suction peak. Usually algebraic, one or two-equation or Reynolds stress turbulence models are used. Since the Reynolds numbers in turbomachinery are large enough to guarantee the flow is turbulent, suitable transition and turbulence models are crucial for accurate prediction of steady and unsteady separated flow. The viscous flow solution of compressor airfoils at off-design conditions is challenging due to flow separation and transition to turbulent flow within separation bubbles. Additional complexity arises when the airfoils are vibrating as is encountered in stall flutter. In this investigation calculations are made of a transonic compressor airfoil in steady flow and with the airfoils oscillating in a pitching motion about the mid-chord at 0° and 10° of chordal incidence angle, and correlated with experiments conducted in the NASA GRC Transonic Flutter Cascade. To model the influence of flow transition on the steady and unsteady aerodynamic flow characteristics, the Solomon, Walker, and Gostelow (SWG) transition model is utilized. The one-equation Spalart-Allmaras model is used to model turbulence. Different transition onset models including fixed onset are implemented and compared for the two incidence angle cases. At each incidence angle, the computational model is compared to the experimental data for the steady flow case and also for pitching oscillation at a reduced frequency of 0.4. The 10° incidence angle case has flow separation over front 40% of the airfoil chord. The operating conditions considered are an inlet Mach number of 0.5 and a Reynolds number of 0.9 Million.

The primary purpose of this thesis is to quantify the effects of RANS turbulence modeling on the resolution of free shear vortical flows. The simulation of aerodynamic wing-tip vortices is used as a test bed. The primary configuration is flow over an isolated finite wing with aspect ratio, , and Reynolds number, . Tip-vortex velocity profiles, vortex core and wake turbulence levels, and Reynolds stresses are compared with wind tunnel measurements. Three turbulence models for RANS closure are tested: the Lumley, Reece, and Rodi full Reynolds stress transport model and the Sparlart-Allmaras model with and without a proposed modification. The main finding is that simulations with the full Reynolds stress transport model show remarkable mean flow agreement in the vortex and wake due to the proper prediction of a laminar vortex core. Simulations with the Spalart-Allmaras model did not indicate a laminar core and predicted over-diffusion of the tip-vortex. Secondary investigations in this work include the study of wall boundary layer treatment and simulating the wake-age of an isolated rotorcraft in hover using a steady-state RANS solver. By comparing skin friction plots over the NACA 0012 airfoil, it is shown that wall functions are most effective in the trailing edge half of the airfoil, while high velocity gradient and curvature of the leading edge make them more vulnerable to discrepancies. The rotorcraft simulation uses the modified Spalart-Allmaras turbulence model and shows proper, qualitative, resolution of the interaction between the vortex sheet and the tip vortex.

This study investigated two most important parts for numerical simulation of UV disinfection reactors: fluence rate distribution and turbulence modeling. In the first part, the evaluation of alternative fluence rate distribution models, including Multiple Point Source Summation (MPSS), Multiple Segment Source Summation (MSSS), Line Source Integration (LSI), Modified LSI (RAD-LSI), Discrete Ordinate (DO) and View Factor models, has been performed. As part of the evaluation, a complete MSSS model, which accounts for the quartz sleeve thickness while calculating refraction angles, was developed. In addition, a simple attenuation factor approach was introduced to integrate the physics of reflection, refraction, and absorption effects into the LSI model. In the second part, six turbulence models, including standard k-e, k-e RNG, k-w(88), revised k-w(98), Reynolds Stress Transport (RST), and Two-Fluid (TF), were investigated and applied to the flow field simulation of a closed conduit polychromatic UV reactor. Predicted flow field, kinetic turbulence energy, and turbulence energy dissipation rate were compared with the experimental data from a Particle Image Velocimetry (PIV) analysis at four locations close to the UV lamps. All of the predicted flow fields were combined with a MSSS fluence rate model and three different microbial inactivation kinetic models to simulate the disinfection process at two UV lamp power conditions. Microbial transport was simulated using the Lagrangian based particle tracking method. As part of the turbulence model analysis, the fluence distribution and the effluent microbial inactivation predicted by all turbulence models were reported. In the fluence rate distribution study, the results showed that models, neglecting the effects of refraction, deviated significantly from the experimental data. In addition, the MSSS approach or models that incorporated the MSSS concept were found to agree well with the experimentally measured fluence rate distribution. Moreover, little difference was found between the results of the MSSS model with quartz sleeve thickness and UVCalc_3D, which does not model the quartz sleeve thickness in the refraction angle calculation but uses a factor to account for the effects of the quartz sleeve on the fluence rate. The attenuation factor combined with LSI model was found to match the MSSS model predictions, while reducing the computational cost. The results show that the fluence distributions and the effluent inactivation level were sensitive to the turbulence model selection. The level of sensitivity was a function of the operating conditions and the UV response kinetics of the microorganisms. In the turbulence model study, the results show that operating conditions that produce higher log inactivation or microorganisms with higher sensitivity to UV will show greater sensitivity to the turbulence model selection. In addition, a broader fluence distribution was found with turbulence models that predicted a larger wake region behind the lamps and greater turbulent mixing characteristics. Overall, the results of this study suggest that numerical simulations of UV processes using CFD should be initially validated with experimental bioassay tests prior to its use as a tool for understanding and evaluating the UV disinfection performance for a specific reactor design and target microorganism.