Bushing diagnostics using an ensemble of parallel neural networks

This paper presents an ensemble of parallel artificial neural networks (ANN) that were successfully able to diagnose the condition of bushings using California State and IEEE C57.104 criteria taking fourteen variables of dissolved gas analysis (DGA) data for each oil impregnated paper bushing. The work compares the speed, stability and accuracy of collective parallel networks to that of individual artificial neural networks (ANN) of radial basis function (RBF), support vector machines (SVM), multiple layer perceptron (MLP) and Bayesian (BNN) networks. The analysis on 1255 bushings concludes that collective network has a better solution than the neural networks individually. In deciding whether to remove or leave a bushing in service, the accuracy of the individual networks was 60% for RBF, 88% for SVM, and 99% for MLP and 94% for BNN. The committee of ANN produced an accuracy of 99%.