Can wavelet transform be used for detection of faulty system?
Please explain me how.
Yes, you can do it.
Once you have defined your "mother wavelet", you process every part of your signal with such a wavelet you have defined.
As a result you find a set of coefficients and based on their entity you can understand if the waveform you have modelled with your wavelet is present or not in your faulty signal.
In the literature you can find a lot of handbooks to master the Wavelet Transform and Compressive Sensing.
Hope it can help
Alberto
Yes, you can do it.
Once you have defined your "mother wavelet", you process every part of your signal with such a wavelet you have defined.
As a result you find a set of coefficients and based on their entity you can understand if the waveform you have modelled with your wavelet is present or not in your faulty signal.
In the literature you can find a lot of handbooks to master the Wavelet Transform and Compressive Sensing.
Hope it can help
Alberto
It is related to change detection problem (or streaming data).
https://turing.ac.uk/anomaly-change-detection-in-streaming-big-data/
You may look for "non-destructive tests (NDT) using acoustic emission (AE)" in the articles.
Yes, it is my proper technique especially when you do not have a mathematical model of the plant.
One approach is to create templates with the wavelet transform to various operating conditions of the plant , ie , a template for normal operating conditions and failure .
Then perform the transformation of the signal under study and compare with your templates, so, if the error is less than the Normal template , it is very likely that the plant is operating in normal condition , otherwise , is operating in failure .
Also, yo can download one of my papers.
I hope I've helped.
Yes,
I have used the Morlet function for mother wavelet. Depending on the type of failure (its wave form) you have to isolate the range of frequencies. It is also recommended to use other analysis technique, like the phase space, in combination with the wavelet analysis
Yes, very much possible. After you decompose the signals into approx. and detail coefficients, look for the possible band of frequency represented by a particular level of coefficients, you can look for the fault in that particular coefficients. Typically fault occurs as a sudden change in signal amplitude, so edge detection can also be employed to detect such rapid changes/edges. Usually detail coefficient will capture such rapid change as it is the high frequency part among the coefficients. Good luck
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing