An ANN-based identification of geological features using multi-attributes: a case study from Krishna-Godavari basin, India

Abstract

Artificial neural network (ANN)–based workflow has been applied to the multi-channel seismic reflection data from the Krishna-Godavari (K-G) basin on the Eastern continental margin of India for delineation of prominent structural features such as chimneys and faults. To eliminate the noise and enhance the data quality, the seismic data is conditioned and filtered by data-conditioning techniques such as dip-steered and structural filters. A single attribute is not sufficient for the delineation of structures such as faults and chimneys; hence, an amalgamation of multiple attributes is necessary for the enhancement of laterally continuous seismic events. Hence, multi-attributes such as dip variance, curvature, coherency, energy, similarity, instantaneous frequency, and frequency washout ratio are extracted from the conditioned data which were combined and fed as inputs for the ANN for the chimney and fault detection. The neural network is trained at the identified chimney/non-chimney and fault/non-fault locations, which generates a final output, namely chimney probability attribute and a fault probability attribute, revealing an improved visibility of faults and chimneys in the seismic data. This multi-attribute approach shows more reliability in comparison with individual attribute responses, which helps in better structural interpretation. The study presents a workflow for better visualization of the subsurface structural features which thereby helps in detailed structural interpretation of the seismic data.