- Dona Paula, Goa, India.
- +91-0832- 2450327
- iiosc2020[at]nio[dot]org
Abstract Submission No. | ABS-2022-13-0390 |
Title of Abstract | Detection of Faults and high amplitude reflections in Krishna Godavari basin using Machine learning |
Authors | Surbhi Nath*, Dibakar Ghosal, A. Sai Kumar |
Organisation | Indian Institute of Technology Kanpur |
Address | H 356 NEHRU COLONY, DHARAMPUR Dehradun, Uttarakhand, India Pincode: 248001 Mobile: 7042067835 E-mail: surbhi27goswami@gmail.com |
Country | India |
Presentation | Oral |
Abstract | Faults play a major role in hydrocarbon accumulation and its migration. Knowing the location of faults allows the assessment of reservoir quality and helps in identifying the traps. In this paper faults are identified using U-Net Convolution Neural Network where deep learning workflow used for the seismic interpretation for fault identification from the Krishna-Godavari basin, which is a petroliferous basin and was formed as a result of rifting along the eastern continental margin of Indian Craton in early Mesozoic. Statistical filter like DSMF (Dip Steered Median Filter) was applied in the 3D seismic cube and then Pretrained U-Net model was used as an autoencoder to predict fault probability map. Faults are identified between Inlines 50065 and 50200; growth faults and Blind faults were identified and then differentiated by plotting Throw vs Two way Travel Time. Fault throw were calculated by measuring the cutoff distance between the footwall and hanging wall in seismic profile perpendicular to the fault strike. Fault throws further used to compute fault contours as well as Expansion index (EI) which identifies whether the fault is active or inactive. Variation in gradient (K) of T-Z plot was computed to identify growth and non-growth periods, null-slopes (K = 0) representing periods of no-growth, positive slopes (K > 0) indicating growth periods and negative slopes (K < 0) reflecting the reversal of movement in faults, or fault overlap and linkage during the propagation of distinct segments. We demarcated the different stages of the faults where initial stage shows that normal slope sediments were least affected by the tectonic activities whereas later stages indicate that there were some neo tectonic activity that influenced the sedimentary strata and resulted in faulting which lead to the formation of pathways for fluid migration from the fault planes. Keywords- T-Z (Time v/s Depth), Convolution Neural Network, Throw |