Deteksi Sesar Otomatis pada Data Seismik 2D Post-Stack Menggunakan U-Net dengan Atribut Diskontinuitas Multi-Channel dan Fault Enhancement Filter di Pulau Madura

Putra, Muhammad Rizqy Arsa (2026) Deteksi Sesar Otomatis pada Data Seismik 2D Post-Stack Menggunakan U-Net dengan Atribut Diskontinuitas Multi-Channel dan Fault Enhancement Filter di Pulau Madura. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5017221046-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (11MB) | Request a copy

Abstract

Interpretasi sesar pada data seismik refleksi merupakan komponen kritis dalam pemodelan struktur bawah permukaan, namun proses manual yang konvensional bersifat subjektif dan tidak efisien untuk data berskala besar. Penelitian ini mengusulkan pendekatan deteksi sesar otomatis berbasis U-Net yang dilatih menggunakan data sintetis FaultSEG 3D, dikombinasikan dengan atribut diskontinuitas Multi-Channel dan Fault Enhancement Filter (FEF) sebagai tahap pra-pemrosesan, dan diterapkan pada tiga lintasan seismik 2D post-stack di Pulau Madura dalam Zona Sesar RMKS. Data prediksi bersumber dari Pusat Survei Geologi Bandung dengan kode 89-MDR, mencakup tiga lintasan berkualitas berbeda (good, moderate, bad). Model U-Net menerima input lima channel berupa atribut semblance, coherence, chaos, similarity, dan ant-tracking discontinuity yang masing-masing diekstraksi dari data seismik yang telah dikondisikan menggunakan FEF. Pelatihan dilakukan dengan 21.600 patch dari 200 volume sintetis menggunakan fungsi loss Focal Tversky (α=0.7, β=0.3, γ=0.75) dan optimizer AdamW. Evaluasi pada 2.160 patch validasi menunjukkan Dice Coefficient sebesar 0.7201, Precision 0.8102, Recall 0.6481, dan AUC-ROC 0.8587 pada threshold 0.80. FEF dapat meningkatkan kontras diskontinuitas pada zona sesar sekaligus mempertahankan kontinuitas reflektor melalui integrasi adaptif DSMF dan DSDF berbasis semblance. Hasil prediksi pada ketiga lintasan menunjukkan distribusi apparent dip yang konsisten dengan rata-rata 65.8°–69.3° dan proporsi segmen berdip di atas 60° mencapai 70.2–86.2%. Dominasi sesar sub-vertikal hingga vertikal ini secara kualitatif konsisten dengan karakter Zona Sesar RMKS yang dicirikan oleh sesar induk vertikal berakar dalam ke basement dan membentuk positive flower structure di Pulau Madura.
=================================================================================================================================
Fault interpretation in reflection seismic data is a critical component of subsurface structural modeling; however, conventional manual interpretation is subjective and inefficient for large-scale datasets. This study proposes an automatic fault detection approach using a U-Net architecture trained on synthetic FaultSEG 3D data, combined with multichannel discontinuity attributes and a Fault Enhancement Filter (FEF) as a preprocessing stage, applied to three 2D post-stack seismic lines across the RMKS Fault Zone on Madura Island. Prediction data were obtained from the Geological Survey Center Bandung (code 89-MDR), comprising three lines of varying quality (good, moderate, bad). The U-Net accepts five channel input consisting of semblance, coherence, chaos, similarity, and ant-tracking discontinuity attributes extracted from FEF-conditioned seismic data. Training was conducted on 21,600 patches from 200 synthetic volumes using Focal Tversky loss (α=0.7, β=0.3, γ=0.75) and the AdamW optimizer. Evaluation on 2,160 validation patches yielded a Dice Coefficient of 0.7201, Precision of 0.8102, Recall of 0.6481, and AUC-ROC of 0.8587 at threshold 0.80. FEF can enhance discontinuity contrast at fault zones while preserving reflector continuity through adaptive integration of DSMF and DSDF guided by semblance. Fault detection results across all three lines demonstrated consistent apparent dip distributions, with mean values of 65.8°–69.3 and 70.2–86.2% of segments exhibiting dips exceeding 60°. This dominance of sub-vertical to vertical faults is qualitatively consistent with the RMKS Fault Zone character, described as a deep-rooted vertical master fault system forming positive flower structures on Madura Island.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Deteksi Sesar Otomatis, Fault Enhancement Filter, U-Net Automatic fault detection, Fault Enhancement Filter, U-Net.
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing.
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geophysics Engineering > 33201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Rizqy Arsa Putra
Date Deposited:	20 Jul 2026 03:42
Last Modified:	20 Jul 2026 03:42
URI:	http://repository.its.ac.id/id/eprint/135557

Actions (login required)

View Item