Evaluation of various transformations to extract characteristic parameters from vibration signal monitoring of power transformer

Abstract

Vibration Analysis has been beneficial for monitoring and assessment of High Voltage Transformer winding and core condition. A lot of research has been conducted to extract useful information from the vibration signals using various transformation methods. In this paper, several methods are evaluated to determine which method is better in providing consistent and reliable parameters to be used as a fingerprint for the transformer condition. The evaluation is conducted using a set of vibration signals taken sequentially with the same condition. The first evaluated method is Fast Fourier Transform (FFT) which is generally and effectively used to compute Discrete Fourier Transform. This method decomposes the vibration signal into various signals with different frequency and amplitude. The second evaluated method is Hilbert Huang Transform (HHT) which separated vibration signal into a finite and small number of Intrinsic Mode Functions (IMF) before applying Hilbert Transform. Since HHT has some intrinsic shortcomings a third method is evaluated which combine Wavelet Packet Transform (WPT) with HHT. WPT is used to decompose the vibration signal into a set of narrow band before being screened to have only the greatest energy signal. Furthermore HHT is applied to the signal after being recomposed. The evaluation result shows that only the main frequency signal can be used as a consistent and reliable parameter to construct fingerprints of the power transformer condition.