Automatic picking of P and S arrivals using a minimal uncertainty wavelet approach

TL;DR

This paper introduces a novel mu-wavelet based algorithm enhanced with a signal/noise weighting function to accurately detect P and S wave arrivals in seismic data, improving automation and reliability.

Contribution

It presents a new method combining mu-wavelet decomposition with a signal/noise ratio to automatically identify seismic wave arrivals with high precision.

Findings

Accurately detects P and S wave arrivals with less than four discrete differences from manual detection.

Effectively isolates significant seismic arrivals using a signal/noise relationship function.

Improves automation in seismic event analysis for well observation systems.

Abstract

The so-called "minimum uncertainty" or mu-wavelet approach has been shown to be very effective. Prior work has shown that using special algorithms based on a mu-wavelet decomposition of a seismic signal allows for arrival detection of various waves on a single-component seismogram with good precision. However, the question of interpretation of the type of wave arriving has not been addressed. The problem lies in the fact that the algorithm detects more than two arrivals even on seismograms with little noise, the peaks corresponding to P and S wave arrivals on the indicator function are not always the strongest, and the peak corresponding to the P wave arrival is not always the first peak detected. We propose to use a signal/noise relationship function in a running window as a weight coefficient of the indicator function obtained using the mu-wavelet transformation, which allows to isolate specifically those arrivals which correspond to significant changes in the signal amplitude. To interpret wave arrivals on linear registration systems (well observation systems for monitoring hydrolic fracturing, ultrasound investigations on models and samples of geological material), an algorithm was also developed to isolate arrivals that best correspond to arrivals on all registration points that have observation systems. As a result, we propose an algorithm that allows to automatically isolate arrivals of S and P waves on "frac" event recordings of a well observation system, and provides agreement with manual detection that doesn't exceed four discretes.