Optimasi Parameter Model Ground Motion Prediction Equation (GMPE) untuk Gempa Kerak Dangkal di Pulau Jawa dengan Algoritma Covariance Matrix Adaptation Evolution Strategy (CMA-ES)

Chabiburrohman, Achmad (2026) Optimasi Parameter Model Ground Motion Prediction Equation (GMPE) untuk Gempa Kerak Dangkal di Pulau Jawa dengan Algoritma Covariance Matrix Adaptation Evolution Strategy (CMA-ES). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5001221054-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (102MB) | Request a copy

Abstract

Pulau Jawa memiliki aktivitas seismik yang kompleks akibat keberadaan sesar aktif dan pengaruh zona subduksi di bagian selatan. Kondisi tersebut menyebabkan estimasi gerakan tanah perlu dilakukan secara tepat, terutama untuk mendukung analisis bahaya gempa regional. Namun, model Ground Motion Prediction Equation (GMPE) global tidak selalu sesuai ketika diterapkan langsung pada data regional karena perbedaan karakteristik sumber, lintasan rambat, dan kondisi setempat. Untuk mengatasi permasalahan tersebut, penelitian ini melakukan optimasi parameter GMPE untuk data Peak Ground Acceleration (PGA) gempa kerak dangkal Pulau Jawa menggunakan pendekatan mixed-effects model dan algoritma metaheuristik. Pendekatan mixed-effects model digunakan karena mampu memisahkan ketidakpastian antar-kejadian dan dalam-kejadian, sedangkan fungsi objektif Negative Log-Likelihood (NLLH) digunakan untuk mengukur tingkat kesesuaian antara PGA observasi dan PGA prediksi dengan mempertimbangkan sebaran residual dan ketidakpastian model. Model yang dievaluasi meliputi Boore et al. (2014), Campbell dan Bozorgnia (2014), Chiou dan Youngs (2014), serta Bindi et al. (2017), yang selanjutnya masing-masing disebut BSSA14, CB14, CY14, dan Bindi17. Optimasi dilakukan dengan menempatkan Covariance Matrix Adaptation Evolution Strategy (CMA-ES) sebagai algoritma utama, sedangkan Particle Swarm Optimization (PSO), Teaching-Learning-Based Optimization (TLBO), dan Grey Wolf Optimizer (GWO) digunakan sebagai algoritma pembanding. Evaluasi algoritma dilakukan melalui kurva konvergensi, nilai best–worst, boxplot multi-run, dan uji Wilcoxon signed-rank, sedangkan performa GMPE dinilai berdasarkan NLLH, metrik galat, hubungan observasi–prediksi, Euclidean Distance-Based Ranking (EDR), Bayesian Information Criterion (BIC), Deviance Information Criterion (DIC1 dan DIC2), residual, dan ketidakpastian residual. Hasil penelitian menunjukkan bahwa optimasi meningkatkan kesesuaian seluruh model dibandingkan kondisi default. CMA-ES memberikan kestabilan terbaik secara keseluruhan. CY14-CMA-ES menjadi konfigurasi paling representatif karena memberikan kecocokan probabilistik, kinerja galat, dan hubungan observasi–prediksi terbaik, serta menghasilkan EDR, DIC1, DIC2, dan ketidakpastian residual total terendah tanpa tren residual signifikan terhadap magnitudo, jarak, dan kondisi setempat.
============================================================================================================================================
Java Island has complex seismic activity due to the presence of active faults and the influence of the subduction zone along its southern region. These conditions require accurate ground-motion estimation, particularly to support regional seismic hazard analysis. However, global Ground Motion Prediction Equation (GMPE) models are not always directly applicable to regional data because of differences in source characteristics, wave propagation paths, and site conditions. To address this issue, this study optimizes GMPE parameters for Peak Ground Acceleration (PGA) data from shallow crustal earthquakes in Java using a mixed-effects model and metaheuristic algorithms. The mixed-effects model is employed to separate inter-event and intra-event uncertainties, while the Negative Log-Likelihood (NLLH) objective function is used to measure the agreement between observed and predicted PGA by accounting for the residual distribution and model uncertainty. The evaluated models include Boore et al. (2014), Campbell and Bozorgnia (2014), Chiou and Youngs (2014), and Bindi et al. (2017), hereafter referred to as BSSA14, CB14, CY14, and Bindi17, respectively. The optimization is conducted using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) as the primary algorithm, while Particle Swarm Optimization (PSO), Teaching-Learning-Based Optimization (TLBO), and Grey Wolf Optimizer (GWO) are used as comparison algorithms. Algorithm performance is evaluated using convergence curves, best–worst values, multi-run boxplots, and the Wilcoxon signed-rank test. GMPE performance is assessed using NLLH, prediction-error metrics, observed–predicted relationships, Euclidean Distance-Based Ranking (EDR), Bayesian Information Criterion (BIC), Deviance Information Criterion (DIC1 and DIC2), residual analysis, and residual uncertainty. The results show that optimization improves the agreement of all models with the regional data compared with their default conditions. CMA-ES provides the best overall stability. CY14-CMA-ES is the most representative configuration because it provides the best probabilistic fit, prediction-error performance, and observed–predicted relationship, as well as the lowest EDR, DIC1, DIC2, and total residual uncertainty, without significant residual trends with magnitude, distance, or site conditions.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Ground Motion Prediction Equation (GMPE), Gempa Kerak Dangkal, Mixed-Effect Model, Optimasi Metaheuristik, One Stage Maximum Log-Likelihood., Ground Motion Prediction Equation (GMPE), Shallow Crustal Earthquake, Mixed-Effect Model, Metaheuristic Optimization, One Stage Maximum Log-Likelihood.
Subjects:	G Geography. Anthropology. Recreation > GB Physical geography Q Science > QA Mathematics > QA9.58 Algorithms Q Science > QC Physics Q Science > QE Geology > QE538.8 Earthquakes. Seismology Q Science > QE Geology > QE539.2.S4 Seismic models
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Physics > 45201-(S1) Undergraduate Thesis
Depositing User:	Achmad Chabiburrohman
Date Deposited:	04 Aug 2026 00:57
Last Modified:	04 Aug 2026 00:57
URI:	http://repository.its.ac.id/id/eprint/142671

Actions (login required)

View Item