Mechanical Fault Diagnosis Based on Acoustic Features in Transformers

Audio method is useful for fault diagnosis as some mechanical failures in power transformers may cause abnormal sound. Some metal elements are loose or even fall in the transformer tank, which may cause them to collide with each other and make an abnormal sound. A model is developed to simulate the impact in transformers and the sound of the impact is recorded. The wavelet packet algorithm is used to extract the audio characteristics, and then the neural network algorithm is used for fault diagnosis. The characteristics of the sound are analyzed, and the sound of the impact is identified, so that it can be diagnosed whether there is an impact fault caused by loose elements in the transformer.