In the present study neural networks are investigated for use in fluid dynamics simulations. These range from static simulations for a simple 2D geometry like an airfoil section to dynamic simulations for a complicated 3D geometry like a model submarine. A detailed analysis of the application of neural networks for the case of vehicle trajectory determination is provided. This involves identifying the physics of the problem and tailoring it to a neural network architecture. The learning process involves training the neural network on a variety of maneuvers and the prediction process involves applying new maneuvers to the neural network. The results are compared to both experimental data and CFD data for the training sets and the prediction sets. The need and scope for parallelization in neural networks is also examined and the performance of pattern partitioning and vertical partitioning algorithms is studied.

No description available.

"Compressible fluid dynamics of flow through plain-faced long orifices was investigated with the hope of gaining insight into the fluid dynamics of tip leakage flow. The Reynolds number range investigated was greater than 10*. Measurements were made of the discharge coefficient as a function of back pressure ratio for a sharp-edged orifice and long orifices with an l/d from 1/2 to 8. The discharge coefficient measurements indicate the mass flow rate in an orifice with an l/d of approximately 2 is the largest and the flow rate in a sharp-edged orifice is the smallest for pressure ratios greater than 0.27. The mass flow rate in a sharp-edged orifice is largest for pressure ratios below 0.27. To visualize the flow in a long orifice and model centerline pressure variation, a water table study was performed. The results demonstrate that the flow separates from the sharp corner at the orifice entrance, it accelerates to a maximum Mach number, and then the pressure increases. For back pressures above 0.50, a pressure decrease follows the initial pressure increase. If the maximum Mach number is supersonic, oblique shocks will exist. At the higher back pressures that produce supersonic maximum Mach numbers (0.50 P_B/P₀ < 0.70), the oblique shocks reflect from the centerline as ""Mach reflections"" and the flow is subsonic after the pressure increase. The maximum Mach number for a back pressure ratio of 0.50 is approximately 1.5. At lower back pressure ratios (P_B/P₀ <0.70), the oblique shocks reflect from the centerline in a ""regular"" manner and the flow remains supersonic on the centerline once supersonic speeds are reached. The flow in a long orifice is relatively constant for all back pressure ratios below approximately 0.30. The maximum Mach number for pressure ratios below 0.30 is approximately 1.8. One-dimensional analyses were used to model the flow in long orifices with maximum Mach numbers less than 1.3. Higher discharge coefficients of long orifices compared to sharp-edged orifices are due to pressure rises in the orifices caused by mixing and shock waves. These increases in the discharge coefficients are partly offset by friction and boundary layer blockage. For maximum Mach numbers greater than 1.3, the flow in long orifices is believed to become significantly two-dimensional because of supersonic effects." / Master of Science

Hydraulic energy absorbers may be described as high-loss centrifugal turbomachines arranged to operate as stalled torque converters. The device absorbs the kinetic energy of a vehicle in motion and dissipates the energy into water. A steady, single-phase, Computational Fluid Dynamics (CFD) simulation has been performed to investigate the flow field in a hydraulic energy absorber. It was determined that to better predict the performance of the energy absorber, more sophisticated modeling approaches may be needed. In this research, a steady, two-phase calculation with basic turbulence modeling was used as a first assessment. The two-phase model was used to investigate cavitation effects. Unsteady and advanced turbulence modeling techniques were then incorporated into single-phase calculations. The Multiple Reference Frame (MRF) Technique was used to model the interaction between the rotor and the stator. The calculations provided clearer details of the flow field without dramatically increasing the computational cost. It was found that unsteady modeling was necessary to correctly capture the close coupling between the rotor and the stator. The predicted torque in the unsteady calculations was 70% of the experimental value and twice of the result in the steady-state calculations. It was found that the inaccuracy of torque prediction was due to (1) high pressures in the regions with complicated geometrical boundaries and, (2) dynamic interactions between the rotor and the stator were not captured fully. It was also determined that the unrealistically low pressure values were not caused by the physical cavitation, but by the lack of proper boundary conditions for the model. Further integration of the modeling techniques studied would improve the CFD results for use in the design of the energy absorber. / Master of Science

Digitized by Kansas State University Libraries

An advanced understanding of particle entrainment is required to optimise the design and maintenance of numerous open channel hydraulic systems and structures placed in these systems; including river channels and canals. This study is on particle entrainment (defined as the movement of a particle from a stationary position to being mobile in the flow). Three aspects of particle entrainment were identified as the focus of this work: First, the waiting time for an exposed particle to entrain under constant flow conditions. Second, the flow features responsible for the entrainment of an individual exposed particle. Third, the motion of an entrained particle immediately after entrainment. Waiting time was found to be highly sensitive to protrusion, with a small increase in protrusion resulting in a significant decrease in waiting time. Contrary to previous suggestions the waiting time to entrainment was found to be poorly described by an exponential distribution; instead Weibull or gamma distributions provide an improved fit in both qualitative and quantitative terms. Ensemble averaged flow fields at the point of entrainment were computed to determine the features responsible for entrainment. The data from the transverse vertical PIV plane indicated the presence of two counter-rotating vortices, with the boundary between the vortices located directly over the entrainment particle. The streamwise vertical PIV measurements showed the presence of a structure extending for a considerable distance in the streamwise direction, the length of which appeared to be independent of submergence. Further, the inclination of the downstream end of the structure appeared to increase with submergence. From the point of entrainment particle dffusion in all three coordinate directions displays an exponent significantly greater than that of ballistic diffusion. From the point of entrainment particle diffusion in all three coordinate directions displays an exponent significantly greater than that of ballistic diffusion. The results highlight the clear difference in the local scale between the diffusion of an already mobile particle with one starting from a position of rest.

No description available.

published_or_final_version / Mechanical Engineering / Doctoral / Doctor of Philosophy

published_or_final_version / Mechanical Engineering / Doctoral / Doctor of Philosophy

published_or_final_version / abstract / toc / Mechanical Engineering / Master / Master of Philosophy