Arrival-time picking methodology using fuzzy c-means and Akaike information criterion for downhole microseismic data

Student thesis: Master's Thesis


Microseismic monitoring is a valuable technique to locate and characterize frac- tures in unconventional reservoirs. The monitoring is usually carried out from a large surface array of vertical-component receivers or a short downhole array of three- component receivers. For a downhole array, P- and S-wave arrival-time picking is typically required to process the microseismic data. Furthermore, arrival-time pick- ing is done automatically considering the large volumes of microseismic data. In this work, I propose a new methodology for automatic picking of P- and S- wave arrivals of microseismic events recorded by downhole arrays. The methodology consists of three steps: (1) For a single trace, intervals of possible arrivals are detected using the conditional fuzzy c-means (CFCM) method. (2) These intervals are further classi_ed into representing a P wave or an S wave using the information obtained from polarization analysis. (3) The Akaike information criterion (AIC) picker is then used on the P- and S-wave intervals to pick the corresponding arrival times. To automatically validate the arrival picks, I test the Random-sampling-based Arrival Time Event Clustering (RATEC) method. The proposed methodology was tested on a real downhole microseismic data set and was compared using fuzzy c-means (FCM) and with the short-term average over long-term average (STA/LTA) method. To evaluate the automatic picking, manual picks were used as a reference. For a time tolerance of ±5 ms, the percentage of correct P- and S-wave arrival picks was 81% and 82% for the FCM methodology, and 77% and 75% for the CFCM methodology. The STA/LTA was used to pick only P-wave arrivals; it obtained 60% of correct picks. The RATEC method was used to vali- date the arrival picks obtained by the FCM methodology. The percentage of correct classi_cations was 93% and 87% for the P- and S-wave arrival picks respectively. Based on the real data results, the best picking performance of the proposed methodology is achieved using FCM. The FCM methodology is more robust to de- tect and pick arrivals than the STA/LTA method. Additionally, the straightforward implementation of the FCM method and AIC picker make the FCM methodology implementation relatively simple.
Date of AwardMay 2019
Original languageEnglish (US)
Awarding Institution
  • Physical Sciences and Engineering
SupervisorDaniel Peter (Supervisor)

Cite this