Radial Basis Function and Wavelet Neural Networks for Feathering the Solar Arrays

Abstract

The current baseline solution for minimizing the potential of dynamic loads being Imparted onto the solar arrays of the Space Station during proximity maneuvera by the Space Shuttle is to lock the position of the solar arrays throughout the approach/departure of the Space Shuttle However, If the apriors locked position does not optimally feather the solar arrays throughout the Shuttle's approach/departure, excessive dynamic loads could result, producing a structural failure in the solar arrays. A more robust solution would Involve feathering the Position of the solar arrays throughout proximity maneuvers by the Space Shuttle. The robust solution will be implemented using a neural network acting as an open loop predictor/controller. Radial basis function and wavelet neural networks will be designed and evaluated based on their resultant performance, and the strong similarities between these two networks will be highlighted.