Classification of OLTC defects based on AE signals measured by two different transducers

摘要

The subject of the article is related to diagnosis of on-load tap-changer (OLTC) based on acoustic emission (AE) signals measured by two different transducers. The main advantage of the AE method is the possibility of its application during normal operation without having to turn off the device under investigation. Based on chosen parameters of the AE signal various defects of the OLTC may be recognized. A number of signals, gathered from laboratory tests, in which four types of typical OLTC defects were simulated, was applied for classification studies with the use of artificial intelligence methods. In particular seventeen different supervised learning algorithms were investigated, while their effectiveness was compared by using common measures. Based on the performed studies the best algorithm for each of the two applied transducers was determined. Results of the works in form of the chosen algorithm may be applied in an expert system for diagnosis of OLTC devices when using different AE measuring sensors.