Title page for ETD etd-09182008-063616
|Type of Document
||Fuller, Russell M.
||Neural network estimation of disturbance growth and flow field structure of spatially excited jets
||Master of Science
|VanLandingham, Hugh F.
- neural networks
- non-linear identification
- jet shear flows
|Date of Defense
Neural networks were applied to the estimation problem consisting of identifying
both nearfield and quasi-farfield flow structures of a jet undergoing spatial mode
excitation. The evolution of disturbances introduced by a spatially excited jet spans a
linear and nonlinear regime in the downstream flow field. For the linear portion, the
neural network was trained to identify critical flow field parameters using numerical data
generated from linear stability analysis code. It was shown that the neural network could
function as a multiple-input adaptive linear combiner over the linear nearfield of the jet
flowfield. Beyond the nearfield (2.0 ≤ z/D ≤ 6.0), a back propagation neural network was
trained using experimental data captured during different modal excitation patterns.
Constant velocity contours for mode 0, mode 1, mode ±1, and mode ±2 jet excitations
were accurately estimated using a low-order neural network filter with conditioned inputs.
Moderate success was also demonstrated when the network was used to extrapolate flow
field parameters outside the initial training set. This demonstration of using neural
networks to predict flowfield structure in non-reacting flows is expected to be directly
applicable to estimation and control of reacting flows in combustors.
|| Approximate Download Time
| 28.8 Modem
|| 56K Modem
|| ISDN (64 Kb)
|| ISDN (128 Kb)
|| Higher-speed Access
next to an author's name indicates that all
files or directories associated with their ETD
are accessible from the Virginia Tech campus network only.
If you have questions or technical
problems, please Contact DLA.