APLIKASI DEEP LEARNING DALAM INTERPRETASI SEISMIK HORISON

• Main Author: Muhammad, Hilmy
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/78974
• Institution: Institut Teknologi Bandung

The rapid advancement of technology has led to the continuous development of geophysical methods, particularly in the field of exploration. Before carrying out exploration, seismic interpretation is carried out first. Seismic interpretation is carried out by picking horizons, horizons which are very important for structural analysis, inversion and seismic attribute analysis. However, current horizon seismic is often obtained through manual tracking methods that take time and have the potential to produce errors. Although various automatic tracking techniques have been developed to improve efficiency, there remains a challenge in selecting seismic horizons with complex seismicity. In this research, one of the methods used comes from the field of computer science known as artificial intelligence (AI). This research consists of 3 stages, namely independent seismic features (FSM), preparing a stratigraphic model (PMS), and creating a horizon model. Where at the FSM and PMS stages it is only initial training whose weights will later be used in the horizon model. The horizon model, initialized by FSM and PMS successfully combines previous seismic knowledge on the target seismic data. The use of CNN architecture in predicting sequence boundaries produces good results only by manually interpreting 2.5% without cutting sequence boundaries marked as discontinuities only and obtaining a CNN prediction accuracy of 98.09%.