A neural network based method for leakage current prediction of polymeric insulators

This letter describes a neural network approach to the prediction of the leakage current (LC) on silicone rubber insulators exposed to salt-fog. The validity of the approach was examined by testing several insulators in a salt-fog chamber. Feed-forward back propagation was found as the best method among several training methods evaluated for the prediction of the LC. The predicted LC with this method has less than 12% error for the tested cases.