Wavelet transform based decomposition and reconstruction for on‐line PD detection and measurement. Part I: Narrow band components decomposition

Abstract

SUMMARYIn this paper, a novel multi-resolution filter for on-line decomposition and analysis of partial discharge (PD) signals based on discrete wavelet transform (DWT) is developed. By applying this filter, as soon as a new sample of the original signal is drawn, the new elements of detail and approximate components at any desired level is derived, and the successive procedure of signal processing in basic DWT is eliminated. Using this filter, the PD or any non-stationary (transient) signal is decomposed to different frequency band components, simultaneously. Therefore, the decomposition components of PD signals in all frequency bandwidths are simultaneously available. The frequency characteristic of this filter is verified, then in order to make the proposed filter more suitable for on-line applications, its order is reduced without losing the accuracy of the results. Results, for simulation and field test data, have shown the appropriateness of the proposed filter for on-line PD detection and measurement.

On-line fault detection method for induction machines based on signal convolution

“…In order to overcome this drawback improved MCSA have been proposed [13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] using a joined time-frequency transformation on the motor stator currents and WT, and featuring extraction and fault diagnosis, even under variable load conditions. However, when using WT it is not easy to define a simple algorithm to develop an automatic fault-detection system due to the predetermined frequency analysis bands associated with discrete filter banks of the transformation.…”

“…On the other hand, discrete WT (DWT) is commonly used in electric engineering to detect and diagnose disturbances occurring in three-phase IMs [15][16][17]. In Ref.…”

“…The proposal presented in Ref. [17] is based on the analysis of the wavelet packet transform coefficients, and a useful implementation is presented. Discrimination between faults, however, is not possible and faults harmonics under main harmonic is lost, because only two levels are considered in wavelet decomposition.…”

Real‐time condition monitoring on VSD‐fed induction motors through statistical analysis and synchronous speed observation

“…The introduced technique is able to detect distinct faults like BRB, misalignment (MAL), and unbalance (UNB) at different rotating speed, unlike previous approaches that detect just one single fault as bearings [6][7][8], rotor bars [9][10][11][12][13], and UNB [14][15][16][17][18] at a fixed rotating speed. Previous methods in reviewed literature use computationally complex signal processing techniques like WT [19][20][21][22][23], FT [24][25][26][27], and their combination with other techniques [28][29][30][31][32][33][34][35][36][37] for detecting multiple faults, which restrain them to off-line applications. In this regard, and as added contribution, the proposed technique implements the standard statistical analyses, the sensorless synchronous speed estimation, and the neural network classifier into a field programmable gate array (FPGA), as a standalone system-on-chip (SoC), which utilizes a custom embedded processor and proprietary hardware to offer a low-cost and portable solution for real-time multiple fault diagnosis on VSD-fed induction motors at different rotating speed.…”

“…Many of these techniques utilize diverse signal processing algorithms; amongst the most popular are: WT and Fourier transform (FT). For instance, a discrete WT (DWT)‐based multi‐resolution filter for the analysis of partial‐discharge signals is proposed in . In , a methodology is proposed for diagnosing the presence of rotor bar failures utilizing a combination of DWT with scale transform, for feature extraction, and correlation coefficient, for pattern recognition.…”