Prediksi Waktu Tiba Gelombang P dan S Data Microearthquake Pada Lapangan 'X' Dengan Algoritma Ensemble Learning (Conv-1D, Res-CNN, BiLSTM, LSTM)

Jabar, Omar Abdul (2024) Prediksi Waktu Tiba Gelombang P dan S Data Microearthquake Pada Lapangan 'X' Dengan Algoritma Ensemble Learning (Conv-1D, Res-CNN, BiLSTM, LSTM). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5017201031-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (7MB) | Request a copy

Abstract

Monitoring dan evaluasi di lapangan geothermal sangat penting untuk menjaga kestabilan dan mengoptimalkan efisiensi produksi. Salah satu metode geofisika yang digunakan untuk hal tersebut adalah metode Microearthquake (MEQ). Distribusi MEQ yang akurat dapat memberikan gambaran terkait dinamika bawah permukaan dan kondisi reservoir. Namun, terdapat kendala yaitu ketidakkonsistensian penentuan waktu tiba gelombang yang dapat menimbulkan bias pada pengolahan dan analisis lanjutan. Berdasarkan hal tersebut pada penelitian ini bertujuan untuk mengaplikasikan model prediksi waktu tiba gelombang P dan S yang bernama EQTransformer. EQTransformer merupakan pre-trained model dengan algoritma ensemble learning. Ensemble Learning merupakan gabungan dari beberapa algoritma untuk menigkatkan performa model secara maksimal. Data MEQ yang digunakan adalah waveform hasil perekaman seismogram komponen vertikal dan data phase report sheet gempa mikro yang berisi informasi tentang event MEQ. Model EQTransformer terdiri dari empat algoritma machine learning diantaranya algoritma Convolutional 1-D yang mampu mengekstraksi informasi penting dari data, Residual Convolutional Neural Network (Res-CNN) pengembangan dari algoritma konvolusi dengan fitur skip-connection untuk mencegah overfitting, Bidirectional Long Short-Term Memory (BiLSTM), dan Long Short-Term Memory (LSTM) yang merupakan algoritma untuk data berbasis sequential. Berdasarkan hasil evaluasi yang telah dilakukan Model EQTransformer tersebut mampu mengenali karakteristik gelombang P dengan sangat akurat. Namun, Model EQTransformer kurang akurat dalam mengenali karakteristik gelombang S. Hal tersebut ditunjukkan dengan nilai akurasi sebesar 0,86, recall senilai 0,89, dan F-1 Score sebesar 0,88 untuk hasil prediksi gelombang P. Sedangkan gelombang S memiliki nilai akurasi sebesar 0,17, recall sebesar 0,17, dan F-1 Score sebesar 0,29.
=================================================================================================================================
Monitoring and evaluation in geothermal fields is very important to maintain stability and optimize production efficiency. One of the geophysical methods used for this is the Microearthquake (MEQ) method. Accurate MEQ distribution can provide an overview of subsurface dynamics and reservoir conditions. However, there is an obstacle, namely the inconsistency in determining the wave arrival time which can cause bias in further processing and analysis. Based on this, this research aims to apply a P and S wave arrival time prediction model called EQTransformer. EQTransformer is a pre-trained model with an ensemble learning algorithm. Ensemble Learning is a combination of several algorithms to maximize model performance. The MEQ data used is the waveform resulting from recording the vertical component seismogram and the micro-earthquake phase report sheet data which contains information about the MEQ event. The EQTransformer model consists of four machine learning algorithms including the 1-D Convolutional algorithm which is able to extract important information from data, Residual Convolutional Neural Network (Res-CNN) a development of a convolution algorithm with a skip-connection feature to prevent overfitting, Bidirectional Long Short-Term Memory (BiLSTM), and Long Short-Term Memory (LSTM) which are algorithms for sequential-based data. Based on the evaluation results that have been carried out, the EQTransformer model is able to recognize the characteristics of P waves very accurately. However, the EQTransformer model is less accurate in recognizing the characteristics of the S wave. This is shown by an accuracy value of 0.86, a recall of 0.89, and an F-1 Score of 0.88 for the P wave prediction results. Meanwhile, the S wave has an accuracy value of 0.17, recall of 0.17, and F-1 Score of 0.29

Item Type:	Thesis (Other)
Uncontrolled Keywords:	EQTransformer, MEQ, Ensemble Learning, Geothermal
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QE Geology > QE538.8 Earthquakes. Seismology Q Science > QE Geology > QE539 Microseisms.
Divisions:	Faculty of Civil, Planning, and Geo Engineering (CIVPLAN) > Geophysics Engineering > 33201-(S1) Undergraduate Thesis
Depositing User:	Omar Abdul Jabar
Date Deposited:	22 Aug 2024 01:28
Last Modified:	22 Aug 2024 01:28
URI:	http://repository.its.ac.id/id/eprint/115163

Actions (login required)

View Item