Neural Network Representation of External Tilt-Rotor Noise

Results from a neural network study of the noise data from a full-scale XV-15 tilt-rotor are presented. Specifically, this database was acquired during the 1998 NASA Ames 80by 120-foot wind tunnel test to estahlish the blade-vortex-interaction noisesignature. The present study has threeobjectives: 1) Toconduet anenral-net\vork-based quality assessment of tltenoise data; 2) To obtain neural network representations of the noise data and to demonstrate their sensitivity to test conditions; 3) To obtain neural-network-based noise predictions. Overall, neural networks are successfully used to assess the quality of the noise data and to represent the complete database as well a s to predict tilt-rotor noise using the minimal amount of input data. As major findings, the data quality is found to he acceptable, and accurate neural network representations are obtained for the test-condition-sensitivity cases.