Role of insulating barriers during electrical treeing in composite dielectrics using Partial Discharge signature analysis based on Adaptive Probabilistic Neural Network

Electrical treeing is one of the significant causes for degradation of solid insulating materials, particularly in polymer and power cable insulation. It is an established fact that the extent of treeing in insulators is strongly related to the type of the dielectric material and the nature of flaw. In this research, detailed studies are carried out to investigate the electrical tree growth characteristics due to the effect of barrier in a composite polymeric insulation system, using a needle-plane electrode model. Laboratory studies have been conducted with a constant alternating voltage supply at power frequency (50 Hz) on composite virgin polymeric insulation specimen made up of two layers of Poly Methyl Meth Acrylate (PMMA) and an intermediate high permittivity layer of mica which acts as a barrier for tree growth. Electrical tree initiation and propagation is analyzed in relation to Partial Discharge (PD) signatures acquired from the PD measurement system and by correlating the physical aspects of degradation due to tree morphological structure. PD patterns during tree growth are analyzed in correlation with PD signatures with and without barrier. Phase resolved patterns of PD signatures obtained from the acquisition system are preprocessed and unique features are extracted based on statistical operators which in turn forms the input to the proposed Adaptive Probabilistic Neural Network (APNN) for discrimination of `tree´ and `no-tree´ patterns. For the purpose of comparison the original and adaptive versions of PNN have been taken up for detailed studies for discrimination of tree patterns. In addition, this study envisages ascertaining the role of the barrier in inhibiting the tree inception, nature of tree morphology and effect of space charge during tree propagation.