I am using dataset found online for SHM which uses accelerometer sensor to detect changes in the structure. I know that I can transform the time-domain data into a frequency-domain and into a time-frequency domain (Ex.: Wavelet). So, I have a few questions:

  1. What features I can extract from each domain transform? I know that I could use Statistical as features for Time-domain analysis. But, I would like to know other features that can be used for each transfer.
  2. Is there another kind of time-frequency domain transform? I know that wavelet and STFT.
  3. In the dataset, there are multiple files where several files have different damage scenarios and this can be found below:

    Folder May 28:

    1-105 Damage scenario 1:

    1-20: D1 21-43: D2 44-65: D3 66-85: D4 86-105: D5 106-128: undamaged 181-273: undamaged

Where each file has 15 acceleration sensor data. So, How can I obtain features and form a feature matrix to be used for pattern recognition and machine learning? I am would like to know what is the typical approach in order to form a feature matrix so that it can be used in the machine learning to observe the pattern.

link to the dataset: http://users.metropolia.fi/~kullj/JrkwXyZGkhF/wooden_bridge_time_histories/

Note: I converted the wav file into a txt file where I got the acceleration data and the sampling frequency


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.