Perancangan Sistem Deteksi Dini Gempa Bumi Dengan Model Efficient Net B0

Rahman, Mohammad Arfan Nur (2026) Perancangan Sistem Deteksi Dini Gempa Bumi Dengan Model Efficient Net B0. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009221073_Undergraduate_Thesis.pdf Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Gempa bumi merupakan salah satu bencana alam yang memiliki dampak besar terhadap keselamatan masyarakat dan infrastruktur. Oleh karena itu, pengembangan teknologi deteksi dini menjadi aspek penting dalam upaya mitigasi bencana. Penelitian ini mengembangkan sistem deteksi prekursor gempa bumi serta estimasi magnitudo berdasarkan analisis data geomagnetik pada rentang pulsasi Ultra Low Frequency (ULF) khususnya Pc3 dengan memanfaatkan periode pengamatan 30 hari sebelum kejadian gempa bumi. Pendekatan berbasis deep learning diterapkan menggunakan arsitektur EfficientNet-B0 sebagai backbone pada dua model yang dikembangkan, yaitu model klasifikasi deteksi prekursor dan model estimasi magnitudo gempa. Hasil penelitian menunjukkan bahwa model deteksi prekursor mampu mempelajari karakteristik pola geomagnetik yang berkaitan dengan keberadaan prekursor. Performa model pada data validasi menunjukkan nilai akurasi sebesar 0,97. Selain itu, interpretasi model menggunakan Gradient-weighted Class Activation Mapping (Grad-CAM) menunjukkan adanya pola aktivasi yang konsisten pada data dengan label serta stasiun pengamatan yang sama. Hal tersebut memberikan indikasi bahwa model mampu mengekstraksi fitur geomagnetik tertentu sebagai dasar pengambilan keputusan. Sementara itu, model estimasi magnitudo masih menunjukkan keterbatasan performa dengan nilai akurasi data validasi sebesar 0,53 serta belum menghasilkan pola aktivasi yang konsisten berdasarkan analisis Grad-CAM. Penelitian ini menunjukkan bahwa pemanfaatan deep learning pada analisis sinyal geomagnetik memiliki potensi untuk mendukung pengembangan teknologi mitigasi bencana. Selain itu, pendekatan yang dikembangkan sejalan dengan SDGs poin 9 melalui penerapan inovasi berbasis kecerdasan buatan, serta SDGs poin 13 dalam upaya meningkatkan ketangguhan masyarakat terhadap risiko bencana.
===================================================================================================================================
Earthquakes are among the most devastating natural disasters, posing significant risks to human safety and infrastructure. Therefore, the development of early detection technologies has become a crucial aspect of disaster mitigation efforts. This study develops an earthquake precursor detection and magnitude estimation system based on the analysis of geomagnetic data within the Ultra Low Frequency (ULF) pulsation range, particularly the Pc3 band, using a 30-day observation period prior to earthquake occurrences. A deep learning approach was employed using the EfficientNet-B0 architecture as the backbone for two developed models: an earthquake precursor detection model and an earthquake magnitude estimation model. The results indicate that the precursor detection model successfully learned geomagnetic pattern characteristics associated with the presence of earthquake precursors. The model achieved a validation accuracy of 0.97. Furthermore, model interpretation using Gradient-weighted Class Activation Mapping (Grad-CAM) revealed consistent activation patterns across samples sharing the same labels and observation stations. This finding suggests that the model is capable of extracting specific geomagnetic features that serve as the basis for its decision-making process. In contrast, the magnitude estimation model demonstrated limited performance, achieving a validation accuracy of 0.53 and failing to produce consistent activation patterns based on Grad-CAM analysis. Overall, this study demonstrates that applying deep learning to geomagnetic signal analysis has the potential to support the development of disaster mitigation technologies. Moreover, the proposed approach aligns with Sustainable Development Goal (SDG) 9 by promoting innovation through artificial intelligence and contributes to SDG 13 by enhancing community resilience to disaster risks.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Prekursor, Deteksi dini, Gempa Bumi, Geomagnetik, Pembelajaran Mendalam, Precursor, Early Detection, Earthquake, Geomagnetics, Deep Learning
Subjects:	Q Science > QE Geology > QE1.F557 Magnetic anomalies--Measurement. Q Science > QE Geology > QE538.8 Earthquakes. Seismology
Divisions:	Faculty of Industrial Technology > Physics Engineering
Depositing User:	Mohammad Arfan Nur Rahman
Date Deposited:	31 Jul 2026 07:13
Last Modified:	31 Jul 2026 07:13
URI:	http://repository.its.ac.id/id/eprint/139981

Actions (login required)

View Item