Thesis (Ph.D.)--Memorial University of Newfoundland, 2005. / Bibliography: leaves 113-118.

published_or_final_version / abstract / toc / Mechanical Engineering / Doctoral / Doctor of Philosophy

Thesis(M.S.)--Auburn University, 2006. / Abstract. Vita. Includes bibliographic references.

Thesis (doctoral)--University of Melbourne, 2005.

Thesis(M.S.)--Auburn University, 2005. / Abstract. Includes bibliographic references.

Thesis (Ph. D.)--University of Hong Kong, 1999. / Includes bibliographical references (leaves 196-210).

A method in which the transport and evolution of the bubble population in a surface ship wake is numerically simulated is presented. The simulation is accomplished by constructing an advective-diffusive transport model for the scalar bubble field and solving this model for late times after ship passage. The bubble population model requires convection velocities and turbulent diffusion information that is supplied by solving the Reynolds-averaged parabolized Navier-Stokes equations with a <i>k</i> - ∊ turbulence model. The mean flow equations are solved by approximating the differential equations with a second order accurate finite difference scheme. The resulting large, sparse, banded matrix is solved by applying a version of the conjugate gradient method. The method has proven to be efficient and robust for the free shear flow problems of interest here. The simulation is initiated with given information in a plane at some point downstream of the ship from which the solution is propagated. The model is executed for a single and a twin propeller ship at 15 knots. The simulation shows that the development of the hydrodynamic and bubble near wake is dominated by ship geometry via strong advective transport. The far wake is dominated by diffusion and bubble rise and dissolution. Thus relatively large changes in geometry have a limited influence on the far wake. / Ph. D.

Turbulent flow measurements were performed in the wake of a slender axisymmetric body in the presence of a flat plate strut and an image plane representing the "rigid lid" approximation to a free surface. The tests were performed in a wind tunnel at a nominal Reynolds number of 6.0 x 10⁵. All turbulent flow parameters were measured at three streamwise stations. A Yawhead probe was used for the mean flow measurements, and a Constant Temperature Anemometer System with a 'x'-wire probe was used to obtain the turbulent flow characteristics. The presence of the image plane was found to increase the velocity defect and the static pressure as the image plane was approached. A redistribution among the various components of velocity fluctuations was noted near the "rigid lid" plane. The transverse component was enhanced at the expense of the normal component. The image plane also was found to influence the magnitudes and radial spread of turbulence intensities and Reynolds stresses. Some interactions between the wake of the axisymmetric body and that of the plate strut were observed. Overall, the mean velocities and the turbulence quantities indicated symmetry about the image plane throughout the wake. / M.S.

Hot-wire measurements taken in the turbulent wake of a flat plate are presented. Symmetrical and antisymmetrical perturbations at various amplitudes and frequencies were introduced into the wake by small flap oscillations. As predicted by linear stability theory, the sinuous (antisymmetric) mode was observed to be more significant than the varicose (symmetric) mode. When the amplitude of the perturbation was low, the spatial development of the introduced coherent perturbation was predicted well by linear stability theory. At high forcing levels, the wake spreading showed dramatic deviations from the well known square-root behavior of the unforced case. Measured coherent Reynolds stresses changed sign in the neighborhood of the neutral point of the perturbation, as predicted by the linear theory. However, the linear theory failed to predict the disturbance amplitude and transverse shapes close to the neutral point. Some nonlinear aspects of the evolution of instabilities in the wake are discussed. Theoretical predictions of the mean flow distortion and the generation of the first harmonic are compared to experimental measurements. Given the unforced flow and the amplitude of the fundamental wave, the mean flow distortion and the amplitude of the first harmonic are predicted remarkably well.

A low-order, dynamical systems approach is applied to the modeling of flow induced vibrations of flexible cables. By combining a coupled map lattice wake model with a linear wave equation cable model, both the free response of the cable as well as the resulting wake structures are examined. This represents an extension of earlier coupled map lattice models that only modeled the wake of forced cable vibration. The validity of the model is assessed through comparisons with both Computational Fluid Dynamics models (NEKTAR spectral element code) and wake experiments. The experimental wake data was collected through the use of hot-film anemometry techniques. Eight hot-film probes were placed along the span of a flexible cable mounted in the test section of a water tunnel. Through the use of frequency domain correlation algorithms, the phase of vortex shedding was calculated along the cable span from the hot-film velocity data. Results for an elastically mounted rigid cylinder showed that the freely vibrating CML model predicted behavior characteristic of a self-induced oscillator; the maximum amplitude of vibration was found to occur at a cylinder natural frequency that did not coincide identically with the natural shedding frequency of the cylinder. Furthermore, the variation of the frequency of cylinder vibration with its natural frequency was seen to be linear. For standing wave cable responses, the freely vibrating CML model predicted lace-like wake structures. This result is qualitatively consistent with both the NEKTAR simulations and experimental results. Little difference was found between the wakes of forced and freely vibrating cables at the Reynolds number of the study $Re=100$. Finally, it was found that the freely vibrating CML could match numerical predictions of cross-flow amplitude as the cable mass-damping parameter was varied over an order of magnitude (once the CML was tuned to match results at a specific mass-damping level). In addition to providing wake patterns for comparisons with the freely vibrating CML, experimental data was supplied to a self-learning CML scheme. This self-learning CML was able to estimate the experimental wake data with good accuracy. The self-learning CML is seen as the next extension of the freely-vibrating CML model, capable of estimating unmodeled wake dynamics through the use of experimental data.