Fault signal feature extraction method
A feature extraction and fault signal technology, which is applied in pattern recognition in signals, testing of machine/structural components, testing of mechanical components, etc., can solve the problem of grinding of dynamic and static components, misalignment of unit shafting, and lack of analysis results etc. to improve the signal-to-noise ratio and filter out noise
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0058] This embodiment provides a fault signal feature extraction method, such as figure 1 Shown:
[0059] Include the following steps:
[0060] S1. Collect the original signal x(t);
[0061] S2. Perform local mean value decomposition on the original signal x(t) to obtain the PF component;
[0062] S3. Select the main PF component for reconstruction to obtain a reconstructed signal;
[0063] S4. Perform mathematical morphology filtering on the reconstructed signal to extract fault characteristic signals;
[0064] S5. Using the Hilbert demodulation method to extract the amplitude and frequency information of the fault characteristic signal.
[0065] In the specific implementation, the main PF component and the residual term are obtained by performing local mean value decomposition on the original signal x(t); then the main PF component is reconstructed and mathematically morphologically filtered to obtain the fault characteristic signal; finally, by analyzing the fault char...
Embodiment 2
[0070] As an optimization of the above embodiment, in step S2, when performing local mean decomposition on the original signal x(t), through the local mean function m k，i (t) and the envelope estimation function a k，j (t) At the same time, the original signal x(t) is smoothed, and the original signal x(t) is decomposed into a group of PF components whose frequency values are automatically arranged from high to low. The instantaneous frequency of the PF component can be obtained through the pure frequency modulation signal, and the instantaneous amplitude of the PF component can be obtained through the envelope signal. Combining the instantaneous frequency value and the instantaneous amplitude of the PF component, the original The complete time-frequency distribution of the signal, the time-frequency characteristics of the signal can effectively and accurately display the characteristics of the original signal. In step S2, the process of performing local mean value decomposi...
Embodiment 3
[0083] As an optimization of the above-mentioned embodiments, mathematical morphological filtering is a new filtering method based on traditional morphological filtering combined with genetic algorithm, with kurtosis value as the optimization target. Considering the characteristics of random background noise, pulse interference and filtering effect, a disc-shaped structural element is selected. The kurtosis index is used to measure the adaptive optimization of the filter effect structural element scale. The kurtosis is a dimensionless parameter describing the peak degree of the waveform, which is defined as:
[0084]
[0085] In the formula, E(x-μ) 4 is the fourth-order mathematical expectation, μ means the mean, and σ means the standard deviation. Kurtosis processes the signal amplitude to the 4th power, thereby highlighting high amplitudes and suppressing low amplitudes. It is most sensitive to the kurtosis value of the high-amplitude pulse signal. When the pulse signal...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More