MULTI-RESOLUTION WAVELET ANALYSIS FOR FAULT DETECTION

MULTI-RESOLUTION WAVELET ANALYSIS FOR FAULT DETECTION

KARA, Zeynep
SEKER, Serhat

In this study, a multi-resolution wavelet analysis technique is applied to simulation data for fault detection. Data is simulated at the MATLAB environment. For this purpose, a sinusoidal wave form is generated at around 1 kHz sampling frequency and then a faulty case is simulated between 250- 500 Hz using a random process under the band-pass filtering. Hence data and its noisy form are used to show healthy and faulty cases of any physical system respectively. In order to show the fundamental properties of the data set, power spectral density variations are shown to indicate the availability of the data. After that Multi– Resolution Wavelet Analysis (MRWA) is applied to each case. In general, wavelet transform is a time-scale analysis technique which can be accepted as an alternative method to the Fourier transform. However, in this study, MRWA approach is considered. MRWA is a kind of the discrete wavelet transform and it uses filter banks approach. Hence, the time domain properties are shown in the sense of the statistical parameters. Also, calculating the power spectral densities, this comparison is done in frequency domain. With this way, a faulty case and its some properties can be determined at both of the time and frequency domains. Key Words: Wavelets, Filtering, Sub-band analysis, Fault detection

Article
PeerReviewed

ISSN 978-9958-834-36-3

International Burch University

2014-05-15

2533