Detecting and classifying flaws within insulating materials using ultrasound

Previous work has shown that flaws can be detected within insulating materials using ultrasound. An ultrasonic based system has been developed to detect, locate and characterise flaws both in the laboratory and on operating and production sites. The system comprises a standard non-destructive testing ultrasound device linked to a PC. Flaws in polymeric insulation such as cast resin bushings, may be detected via change in signal parameters. These include signal amplitude and depth which may be plotted as a different colour on the map. More detailed investigation of an area of interest is performed both in the time and frequency domains. The time domain signal provides information about the depth and extent of a flaw. The spectrum of the returned signal gives further information about the nature of the flaw. Signals from a number of different flaws have been taken and their spectra used to train a neural network. This network can then assist the operator in classifying flaws. Laboratory samples have been prepared to synthesize a variety of flaws in cast resin. The resulting ultrasound signals have been used to train and test both back-propagation and counter-propagation networks with a success rate of over 90%. Signals taken from industrial samples have been used to train and test networks in classifying voids, debonds and delaminations in medium voltage bushings, and in detecting voids at the semiconductor interface of medium voltage cables. Success is comparable with the synthetic samples. The system has also been used successfully on site.<<ETX>>