Image logs can provide critical information to reduce drilling risk, as they allow for early detection of structures such as faults or fractures. The timeliness of such information is undermined by the time required to manually interpret the data and the subjectivity of interpretations. The proposed methodology is a supervised Deep Learning - based method built on U-Net architecture for segmentaton of image logs acquired while drilling, i.e., automated detection of geological edges in borehole images. The proposed network has been trained on synthetic data and tested on field data. Different learning strategies, namely Curriculum and standard learning (CL,SL), were compared to observe the impact of segmentation process on final results: CL shows greater potential in early fracture detection due to its superior performance in chaotic and heterogeneous intervals

LWD Image Logs Interpretation (LILI) : An Automated Approach Using Deep Learning

A. Molossi
2023-01-01

Abstract

Image logs can provide critical information to reduce drilling risk, as they allow for early detection of structures such as faults or fractures. The timeliness of such information is undermined by the time required to manually interpret the data and the subjectivity of interpretations. The proposed methodology is a supervised Deep Learning - based method built on U-Net architecture for segmentaton of image logs acquired while drilling, i.e., automated detection of geological edges in borehole images. The proposed network has been trained on synthetic data and tested on field data. Different learning strategies, namely Curriculum and standard learning (CL,SL), were compared to observe the impact of segmentation process on final results: CL shows greater potential in early fracture detection due to its superior performance in chaotic and heterogeneous intervals
File in questo prodotto:
File Dimensione Formato  
Paper_20_-_Attilio_Molossi.pdf

Accesso chiuso

Tipologia: Documento in Versione Editoriale
Licenza: Copyright Editore
Dimensione 443.01 kB
Formato Adobe PDF
443.01 kB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11368/3057978
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact