Acoustic-based particle detection in oil using artificial neural networks

This paper contributes to the detection of the presence of free conducting particles in oil-insulated apparatus based on particle-produced acoustics. Acoustic signals are generally produced-under AC-by particle collision against the tank walls. The work uses an inference engine to test for particle contamination in oil by deliberate application of an AC test voltage. The work proposes subjecting the oil-insulated systems to an intentional pre-calculated voltage magnitude for a pre-qualified duration. Using inference, the acoustic signal and pulse train and their statistics could uniquely disclose the characteristics of the contaminating particle.