2.8
A neural network solution to forecasting launch pad winds at Kennedy Space Center
PAPER WITHDRAWN
Kenneth P. Cloys, Air Force Institute of Technology, Wright-Patterson Air Force Base, OH; and M. K. Walters and W. P. Roeder
This research is motivated by the need to provide mission controllers at Kennedy Space Center with advance warning of wind speeds above the thresholds required for free-standing rockets and space shuttles on the launch pad. Accurately estimating winds one to two hours in advance is a Launch Weather Officer's greatest short-term forecast challenge. The disastrous consequences of unexpected high winds are obvious, but false alarms lead to costly and time-consuming mission cancellations.
Time series of wind speed data are highly non-linear and do not have analytical solutions. Using five years of wintertime data from 47 towers of Cape Canaveral’s Weather Information Network Display System, a neural network is trained to predict future values of wind speed at Kennedy Space Center launch pads. The neural net uses historical data to "learn" the underlying deterministic--but unrecognizable--pattern in the data.
The neural network’s predictions are validated against observed wind speeds 15 minutes to 2 hours in the future. The network’s accuracy is expected to significantly enhance Launch Weather Officers' forecast skill.
Session 2, Artificial Neural Networks
Monday, 10 January 2000, 1:30 PM-4:30 PM
Previous paper