Penetuan Kualitas Site Seismic Di Wilayah Sumatra Utara Berdasarkan Machine Learning Spectral Density Dan Horizontal Vertical Spectral Ratio

Fachriyeni, Triya (2026) Penetuan Kualitas Site Seismic Di Wilayah Sumatra Utara Berdasarkan Machine Learning Spectral Density Dan Horizontal Vertical Spectral Ratio. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6009232002_Master_Tesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Akurasi parameter gempa bumi sangat bergantung pada kualitas sinyal seismik dan selalu dipengaruhi oleh model noise dari sumber alami dan antropogenik. Noise antropogenik mendominasi frekuensi tinggi (5-10 Hz) dan menurunkan rasio signal-to-noise sehingga sinyal gempa bumi sulit dideteksi secara visual. Penelitian ini bertujuan untuk mengevaluasi kualitas sensor seismik di wilayah Sumatera Utara berdasarkan karakteristik noise dan respon geologi bawah permukaan. Data yang digunakan berupa sinyal seismik dari 31 stasiun seismik BMKG, dengan 22 stasiun memenuhi kriteria analisis. Proses evaluasi dilaksanakan melalui pendekatan terpadu yang mengintegrasikan analisis Power Spectral Density (PSD), Horizontal to Vertical Spectral Ratio (HVSR), serta teknik unsupervised machine learning menggunakan algoritma Fuzzy C-Means (FCM). Hasil PSD menunjukkan rentang 101 - 103 counts untuk noise rendah hingga menengah, dan meningkat >103 counts akibat aktivitas antropogenik, khususnya frekuensi > 5 Hz. Hasil HVSR menghasilkan nilai frekuensi dominan rendah hingga menengah yang mengindikasikan variasi ketebalan sedimen dan kondisi geologi alluvial. Hasil clustering FCM memberikan pola noise ke dalam 5 cluster yang merepresentasikan noise rendah, menengah, dan tinggi, serta keterkaitan dengan kondisi geologi dan lingkungan sekitar stasiun. Sekitar 40% sinyal seismik masuk dalam cluster 0 dan 1 yang sangat baik dalam analisis sinyal gempa bumi, 25% pada cluster 2 yang dipengaruhi oleh noise alam dan 35% pada dipengaruhi oleh cluster 3 dan 4 yang merupakan noise antropogenik yang perlu dihilangkan. Integrasi metode ini memberikan pendekatan kuantitatif yang efektif untuk evaluasi kualitas site seismik guna mendukung peningkatan keandalan monitoring gempa bumi.
========================================================================================================================
The determination of earthquake parameters is fundamentally constrained by the quality of recorded seismic waveforms, which are susceptible to interference from both natural and human-induced noise sources. Noise of anthropogenic origin primarily dominates the high-frequency band (5–10 Hz), causing a substantial decrease in the signal-to-noise Ratio and thereby hindering the visual identification of seismic events. In this context, the present study focuses on assessing seismic site quality in North Sumatra through an analysis of ambient noise properties and subsurface geological responses. This study utilizes seismic waveform data collected from 31 BMKG seismic stations, of which 22 stations met the predefined selection criteria and were included in the analysis. Site evaluation was conducted using a comprehensive framework that integrates Power Spectral Density (PSD) analysis, Horizontal-to-Vertical Spectral Ratio (HVSR) estimation, and an unsupervised machine learning clustering approach based on the Fuzzy C-Means (FCM) algorithm. Results from the PSD analysis indicate that low to moderate noise environments are characterized by Spectral amplitudes between 101 and 103 counts, whereas values exceeding 103 counts are predominantly linked to anthropogenic disturbances, especially at frequencies above 5 Hz. The HVSR analysis reveals dominant frequencies within the low to moderate range, reflecting spatial variability in sediment thickness and the presence of alluvial geological formations beneath the stations. Based on FCM clustering, seismic noise characteristics are classified into five distinct clusters representing varying noise intensities and their associations with local geological and environmental conditions. Approximately 40% of the analyzed signals fall within clusters 0 and 1, which exhibit favorable noise conditions for earthquake analysis. Around 25% of the data are grouped in cluster 2, largely influenced by natural noise sources, while the remaining 35% correspond to clusters 3 and 4, where anthropogenic noise is dominant and mitigation measures are required. Overall, the combined application of PSD, HVSR, and FCM establishes a robust quantitative framework for evaluating seismic site quality and enhances the reliability of earthquake monitoring systems.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Seismic, Noise, Site, Earthquake, machine learning, Spectral Ratio,Keywords: seismic noise, site quality, earthquake monitoring, machine learning, Spectral Density, Spectral Ratio
Subjects:	Q Science
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30101-(S2) Master Thesis
Depositing User:	Triya Fachriyeni
Date Deposited:	05 Feb 2026 06:20
Last Modified:	05 Feb 2026 06:20
URI:	http://repository.its.ac.id/id/eprint/132156

Actions (login required)

View Item