A BAYESIAN NETWORK ANALYSIS OF MARINE CAPSIZE ACCIDENT : STUDY CASE FOR INDONESIAN NON CONVENTION CARGO VESSEL

Marsa, Athallariq Kiya Dhia (2025) A BAYESIAN NETWORK ANALYSIS OF MARINE CAPSIZE ACCIDENT : STUDY CASE FOR INDONESIAN NON CONVENTION CARGO VESSEL. Other thesis, Sepuluh Nopember Institute of Technology.

Text Bachelor Thesis_Athallariq Kiya Dhia Marsa.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Maritime safety remains a critical concern in Indonesia, with cargo ship capsizing accidents posing significant risks to lives, property, and the environment. Despite existing regulatory frameworks, the persistence of such accidents highlights the urgent need for advanced risk assessment methodologies. This study employs a Bayesian Network (BN) integrated with the Human Factor Analysis and Classification System (HFACS) and Fishbone Diagram to analyze causal relationships and probabilistic dependencies among factors contributing to capsizing accidents involving Indonesian non-convention cargo vessels (NCVS). Using data from 12 capsizing incidents investigated by the National Transportation Safety Committee (NTSC) and Shipping Court reports, the study identifies key risk factors across four HFACS levels: Unsafe Acts decision errors (81.7% probability when present), skill-based errors (75% prior probability), and violations (91.7% prior probability); Preconditions adverse environmental factors (58.3%), ineffective crew resource management (83.3%), and technological failures (33.3%); Supervision inadequate oversight (41.7%) and planning deficiencies (75%); and Organizational Influences lax regulatory oversight (75%) and insufficient resource allocation (66.7%). The BN model, developed using GeNIe software, quantifies these relationships through conditional probability tables (CPTs) and sensitivity analysis, revealing Unsafe Acts as the most influential node (57.58% strength of influence), especially when violations, skill-based errors, and decision errors co-occur. Model validation achieved 73% accuracy, with a higher precision in predicting capsizing events (79.36%) compared to non-capsizing events (62.1%), and the ROC curve (AUC = 0.707) confirmed moderate predictive capability.Key recommendations include enhanced crew training on emergency response and stability management, stricter enforcement of safety regulations and certification audits, modernization of aging vessels and implementation of real-time monitoring systems, and organizational reforms to prioritize safety over profit-driven practices. This study contributes to the maritime safety literature by demonstrating the effectiveness of BN-HFACS integration for dynamic risk analysis and provides a decision-support tool for policymakers and operators to mitigate capsizing risks in Indonesia’s NCVS fleet.

========================================================================================================================

Keselamatan maritim tetap menjadi perhatian utama di Indonesia, di mana kecelakaan kapal kargo yang terbalik (capsizing) menimbulkan risiko besar terhadap nyawa, harta benda, dan lingkungan. Meskipun telah ada kerangka regulasi yang berlaku, masih sering terjadinya kecelakaan tersebut menegaskan perlunya metode penilaian risiko yang lebih canggih. Studi ini menggunakan pendekatan Bayesian Network (BN) yang diintegrasikan dengan Human Factor Analysis and Classification System (HFACS) serta Diagram Fishbone untuk menganalisis hubungan kausal dan ketergantungan probabilistik di antara faktor-faktor yang berkontribusi terhadap kecelakaan kapal terbalik pada kapal kargo non-konvensi Indonesia (NCVS). Berdasarkan data dari 12 insiden kapal terbalik yang diselidiki oleh Komite Nasional Keselamatan Transportasi (KNKT) dan laporan Pengadilan Pelayaran, studi ini mengidentifikasi faktor risiko utama berdasarkan empat level dalam HFACS,yaitu Tindakan Tidak Aman (Unsafe Acts): kesalahan pengambilan keputusan (probabilitas 81,7% saat terjadi), kesalahan berbasis keterampilan (probabilitas awal 75%), dan pelanggaran (probabilitas awal 91,7%),Prakondisi (Preconditions): faktor lingkungan yang merugikan (58,3%), manajemen sumber daya awak kapal yang tidak efektif (83,3%), dan kegagalan teknologi (33,3%).Pengawasan (Supervision): pengawasan yang tidak memadai (41,7%) dan kekurangan dalam perencanaan (75%),Pengaruh Organisasi (Organizational Influences): pengawasan regulasi yang lemah (75%) dan alokasi sumber daya yang tidak mencukupi (66,7%).Model BN yang dikembangkan menggunakan perangkat lunak GeNIe ini mengkuantifikasi hubungan-hubungan tersebut melalui Conditional Probability Tables (CPTs) dan analisis sensitivitas, yang menunjukkan bahwa Tindakan Tidak Aman merupakan node paling berpengaruh (dengan kekuatan pengaruh sebesar 57,58%), terutama saat pelanggaran, kesalahan berbasis keterampilan, dan kesalahan keputusan terjadi secara bersamaan. Validasi model mencapai akurasi 73%, dengan presisi yang lebih tinggi dalam memprediksi kejadian capsizing (79,36%) dibandingkan dengan kejadian non-capsizing (62,1%), dan kurva ROC (AUC = 0,707) menunjukkan kemampuan prediktif yang sedang.Rekomendasi utama dari studi ini mencakup peningkatan pelatihan awak kapal terkait tanggap darurat dan manajemen stabilitas kapal, penegakan regulasi keselamatan dan audit sertifikasi yang lebih ketat, modernisasi kapal-kapal tua serta penerapan sistem pemantauan secara real-time, serta reformasi organisasi untuk lebih mengutamakan keselamatan dibandingkan kepentingan profit semata. Studi ini memberikan kontribusi terhadap literatur keselamatan maritim dengan menunjukkan efektivitas integrasi BN-HFACS dalam analisis risiko dinamis, serta menyediakan alat pendukung keputusan bagi pembuat kebijakan dan operator dalam mengurangi risiko kapal terbalik pada armada NCVS di Indonesia.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Bayesian Network, HFACS, Ship Capsizing, Non-Convention Vessel, Maritime Safety, Risk Assessment,HFACS,Jaringan Bayesian,Kapal Terbalik, Kapal Non-Konvensional, Keselamatan Maritim, Penilaian Risiko.
Subjects:	Q Science > QA Mathematics > QA279.5 Bayesian statistical decision theory. T Technology > T Technology (General) > T174.5 Technology--Risk assessment. T Technology > T Technology (General) > T55.3.H3 Hazardous substances--Safety measures. T Technology > T Technology (General) > T58.62 Decision support systems V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering
Divisions:	Faculty of Marine Technology (MARTECH) > Marine Engineering > 36202-(S1) Undergraduate Thesis
Depositing User:	Athallariq Kiya Dhia Marsa
Date Deposited:	04 Aug 2025 08:26
Last Modified:	04 Aug 2025 08:26
URI:	http://repository.its.ac.id/id/eprint/126574

Actions (login required)

View Item