Analyzing rocket plume spectral data with neural networksThe Optical Plume Anomaly Detection (OPAD) system is under development to provide early-warning failure detection in support of ground-level testing of the Space Shuttle Main Engine (SSME). Failure detection is to be achieved through the acquisition of spectrally resolved plume emissions and subsequent identification of abnormal levels indicative of engine corrosion or component failure. Two computer codes (one linear and the other non-linear) are used by the OPAD system to iteratively determine specific element concentrations in the SSME plume, given emission intensity and wavelength information. Since this analysis is extremely labor intensive, a study was initiated to develop neural networks that would model the 'inverse' of these computer codes. Optimally connected feed-forward networks with imperceptible prediction error have been developed for each element modeled by the linear code, SPECTRA4. Radial basis function networks were developed for the non-linear code, SPECTRA5, and predict combustion temperature in addition to element concentrations.
Document ID
19950059070
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Whitaker, Kevin W. (University of Alabama, Tuscaloosa, AL US, United States)
Krishnakumar, K. S. (University of Alabama, Tuscaloosa, AL US, United States)
Benzing, Daniel A. (University of Alabama, Tuscaloosa, AL US, United States)
Date Acquired
August 16, 2013
Publication Date
January 1, 1995
Publication Information
Publisher: American Institute of Aeronautics and Astronautics