Data Driven Condition Monitoring for Maintenance Decision Support of Ship Main Engine Using Random Forest

Yazuar, Mohammad Iqbal (2026) Data Driven Condition Monitoring for Maintenance Decision Support of Ship Main Engine Using Random Forest. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5019221135-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (59MB) | Request a copy

Abstract

Transportasi laut mengangkut lebih dari 80 persen perdagangan barang global. Kerusakan mesin masih jadi penyebab utama insiden pelayaran, dan sebagian besar operator kapal masih mengandalkan Planned Maintenance System berbasis waktu yang tidak mencerminkan kondisi aktual mesin. Penelitian ini membangun model Random Forest untuk mengklasifikasikan kondisi mesin diesel kapal dari data engine room logbook, memakai sepuluh fitur yang mencakup suhu gas buang, tekanan dan selisih suhu oli pelumas, tekanan udara bilas, putaran mesin, dan suhu air pendingin silinder per silinder. Label kondisi dibangun lewat dua lapis, ambang batas manual dari manual mesin dan baseline statistik bergulir yang menyesuaikan putaran mesin, diterapkan pada 534 watch dari satu kapal selama empat belas bulan. Model biner mencapai rata-rata macro F1 0,736 pada lima pembagian data kronologis. Angka ini naik dari 0,591 pada model enam fitur dengan skema label yang sama. Atribusi komponen pada model yang dilatih penuh menemukan bahwa keausan liner silinder dan ring piston menyumbang 55 dari 154 true positive. Empat puluh satu di antaranya tidak tumpang tindih dengan kategori kegagalan lain, tanda bahwa fitur cooling water per silinder yang ditambahkan memang menangkap kejadian baru, bukan sekadar menandai ulang yang sudah terdeteksi. Keluaran model dipetakan ke tabel rekomendasi tindakan yang melengkapi jadwal BRB-K12 yang sudah ada, dan didemonstrasikan lewat prototipe dashboard berbasis browser. Hasil ini berasal dari satu kapal dan satu model mesin, jadi belum bisa digeneralisasi tanpa validasi lebih lanjut.
=====================================================================================================================================
Maritime transport carries more than 80 percent of global merchandise trade. Machinery failure is still a leading cause of shipping incidents, and most ship operators rely on time-based Planned Maintenance Systems that don't reflect the actual condition of the engine. This study builds a Random Forest model to classify marine diesel engine condition from engine room logbook data, using ten features covering exhaust gas temperature, lubricating oil pressure and temperature difference, scavenge air pressure, engine speed, and per-cylinder cooling water temperature. Condition labels come from two layers, fixed thresholds from the engine manual and a rolling statistical baseline grouped by engine speed, applied to 534 watches from a single vessel over fourteen months. The binary model reaches a mean macro F1 of 0.736 across five chronological splits, up from 0.591 for a six-feature model on the same revised label. Component attribution on the fully trained model finds cylinder liner and piston ring scuffing behind 55 of 154 true positives. Forty-one of those overlap no other failure category, a sign that the added per-cylinder cooling water features catch genuinely new cases rather than re-flagging ones already detected. Model output is mapped to a recommended-action table that supplements the existing BRB-K12 schedule, and demonstrated through a browser-based dashboard prototype. These results come from one vessel and one engine model, and shouldn't be generalized without further validation.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Mesin Diesel Kapal, Maintenance Decision Support, Random Forest,Marine Diesel Engine, Maintenance Decision Support, Random Forest
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. T Technology > T Technology (General) > T58.62 Decision support systems V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering > VM731 Marine Engines
Divisions:	Faculty of Marine Technology (MARTECH) > Marine Engineering > 36202-(S1) Undergraduate Thesis
Depositing User:	Mohammad Iqbal Yazuar
Date Deposited:	04 Aug 2026 06:19
Last Modified:	04 Aug 2026 06:19
URI:	http://repository.its.ac.id/id/eprint/143127

Actions (login required)

View Item