FLORENTINO, M. T. B.; http://lattes.cnpq.br/7745353831423629; FLORENTINO, Marcus Tulius Barros.
Resumen:
In this work, a sensitivity analysis applied to a monitoring and predicting technique of the
operational condition polymeric insulators is proposed. The technique is based on the
acoustic inspection of ultrasonic noises coming from the electrical discharges on the
surface of insulators. For this purpose, tests were carried out in the laboratory with 230
kV polymeric insulators, extracted from a transmission line and with different
degradation levels. The Spectral Subband Centroid Energy Vectors algorithm was used
to extract attributes and process the ultrasonic noise signals. This algorithm splits the
frequency spectrum into a number of overlapping subbands, locates the centroids of each
subband and calculates the energy in the proximity of each centroid. Aiming to support
the decision-making, was employed an artificial neural network capable of separating into
classes the insulators degradation levels. Thus, it made possible to identify and distinguish
accurately the energy vectors from cleaned insulators, polluted insulators and damaged
insulators with success rates always above 80%. With these results, a sensitivity analysis
was performed to choose which parameters and topologies of computational processes
best suited to the proposed estimation of ultrasonic noise signals. The results form the
basis for implementation of the method with a larger amount of samples insulators,
making it possible to statistically evaluate its applicability in the field.