The design of aircraft propulsion configurations must digress from the typical configurations that are utilized on the majority of aircraft in order to consider the effects of environmental issues as well as the noise that is generated from the engines. One unconventional approach under consideration involves rectangular jets near flat surfaces that are parallel to the jet axis. This type of configuration makes an attempt to muffle the noise that propagates to the ground, but previous experimental work showed that the noise generated by this configuration was actually increased due to the effect that the plate trailing edge exerts on the flow. In this thesis, a large eddy simulation study is conducted to determine whether wall deformations at the plate trailing edge could reduce the jet noise. A high aspect ratio rectangular nozzle is placed over a flat surface featuring sinusoidal deformations at the trailing edge. A range of amplitudes and wavenumbers, characterizing the deformations at the trailing edge, is considered to determine the parameter range that corresponds to noise reduction.

Within the design process, computational fluid dynamics is typically used to compute specific quantities that assess the performance of the apparatus under investigation. These quantities are usually integral output functions such as force and moment coefficients. However, to accurately model the configuration, the geometric features and the resulting physical phenomena must be adequately resolved. Due to limited computational resources a compromise must be made between the fidelity of the solution obtained and the available resources. This creates a degree of uncertainty about the error in the computed output functions. To this end, the current study attempts to address this problem for two-dimensional inviscid, incompressible flows on unstructured grids. The objective is to develop an error estimation and grid adaptive strategy for improving the accuracy of output functions from computational fluid dynamic codes. The present study employs a discrete adjoint formulation to arrive at the error estimates in which the global error in the output function is related to the local residual errors in the flow solution via adjoint variables as weighting functions. This procedure requires prolongation of the flow solution and adjoint solution from coarse to finer grids and, thus, different prolongation operators are studied to evaluate their influence on the accuracy of the error correction terms. Using this error correction procedure, two different adaptive strategies may be employed to enhance the accuracy of the chosen output to a prescribed tolerance. While both strategies strive to improve the accuracy of the computed output, the means by which the adaptation parameters are formed differ. The first strategy improves the computable error estimates by forming adaptation parameters based on the level of error in the computable error estimates. A grid adaptive scheme is then implemented that takes into account the error in both the primal and dual solutions. The second strategy uses the computable error estimates as indicators in an iterative grid adaptive scheme to generate grids that produce accurate estimates of the chosen output. Several test cases are provided to demonstrate the effectiveness and robustness of the error correction procedure and the grid adaptive methods.

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.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.