Application of an Artificial Neural Networks Model for Prediction of Instability Phenomena in Low Current Vacuum Arc by Cathode Spot Model

For several years, the cathode spot model has been used as a mechanism for analysis of the instability phenomena in vacuum arcs with numerical methods. Artificial neural networks (ANNs) can accurately predict all disciplines. In this study, the ANN model was used for the prediction of instability phenomena in vacuum arc through the cathode spot model. The traditional method was experimentally used for numerical methods, to determine parameters which are related to the instability of phenomena analysis. The ANN model was used for TRAINBR of the training function. The transfer function was Tan-Sigmoid (Tag-Sig), while the input was two variables, the hidden layer was 39 units, and the output data were ten variables. As a result, the percentage error of mean absolute was 0.1635%. All parameters of the instability phenomena in vacuum arc were similar between the traditional method and the ANN model compared. However, the traditional method was only presented by approximation of the arc current in instability phenomena. Moreover, the advantage of the ANN model is that that the lowest arc current was immediately displayed before instability occurred. In addition, the ANN model can possibly be applied to predict the effects of the instability phenomena in vacuum arc from changing some parameters more quickly than the traditional method.