Background: I am measuring the reflection signal from the pipe. First strong kick is called firing kick. The later small bumps are reflection from devices in the pipe and final one should be the end of the pipe. The data is gathered from multiple shots in order to make sure that the signal is not confused with noise in the pipe.

transducer.txt Transducer data figure

And here is the expected final answer Expected final answer

My current approach is using find_max() which is rely on static threshold of data. It will be fail when the survey is longer and signal is weaker. Therefore I would like to apply machine learning to this problem.


  1. Use sliding windows and let A.I. learn from mean, medium, mode, variance, ... harmonic mean, .... I found that the more it learn the more confusion it gets.
  2. Try object detection approach. It require big dataset and intensive computing power. I don't have them.

Any approach I can train A.I. to learn how to localized bumps?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.