Stowage Planning Assistant Berbasis LLM dan GraphRAG untuk Penempatan Kontainer Sesuai Aturan Keselamatan

Rahmatillah, Faizcha Aisyah (2026) Stowage Planning Assistant Berbasis LLM dan GraphRAG untuk Penempatan Kontainer Sesuai Aturan Keselamatan. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002211161-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Perencanaan penempatan kontainer (stowage planning) merupakan proses krusial dalam operasional kapal kontainer karena berpengaruh langsung terhadap keselamatan struktur dan stabilitas kapal. Proses ini secara konvensional masih banyak dilakukan secara manual atau semi-otomatis, sehingga rentan terhadap kesalahan dan sulit dijelaskan secara transparan. Penelitian ini bertujuan untuk mengembangkan prototipe Stowage Planning Assistant berbasis Large Language Model (LLM) dan Graph Retrieval Augmented Generation (GraphRAG) yang mampu menghasilkan rencana penempatan kontainer sesuai aturan keselamatan dengan tingkat pelanggaran aturan ≤ 0% pada skenario pengujian simulasi, serta mengevaluasi performa sistem dari aspek akurasi, efisiensi waktu, dan utilisasi slot. Sistem yang dikembangkan mengintegrasikan Stowage Planning Engine deterministik, Rule Validation Engine, basis data graf, serta antarmuka pengguna berbasis web dan chatbot. Perhitungan stabilitas kapal dilakukan secara kuantitatif menggunakan parameter Longitudinal Center of Gravity (LCG), Transversal Center of Gravity (TCG), dan Vertical Center of Gravity (VCG). Pengujian dilakukan menggunakan dua skenario data kontainer berukuran 20 kaki dengan variasi rentang berat muatan. Hasil pengujian menunjukkan bahwa sistem mampu menghasilkan rencana penempatan kontainer yang patuh terhadap seluruh aturan keselamatan, menjaga parameter stabilitas kapal dalam batas operasional, serta mengidentifikasi kontainer yang tidak dapat ditempatkan secara konsisten dan aman. Integrasi LLM dan GraphRAG berperan sebagai lapisan interpretatif yang meningkatkan keterjelasan dan transparansi hasil perencanaan tanpa menggantikan proses komputasi inti. Dengan demikian, sistem yang dikembangkan berpotensi menjadi alat bantu pendukung keputusan stowage planning yang aman, terstruktur, dan dapat dijelaskan.
===================================================================================================================================
Container stowage planning is a critical process in container ship operations, as it directly affects structural safety and vessel stability. In practice, stowage planning is often performed manually or semi-automatically, which increases the risk of errors and limits transparency in decision-making. This research aims to develop a Stowage Planning Assistant based on a Large Language Model (LLM) and Graph Retrieval-Augmented Generation (GraphRAG) that is capable of generating container stowage plans compliant with safety regulations with a violation rate of ≤ 0% under simulation-based test scenarios, as well as evaluating system performance in terms of accuracy, computational efficiency, and slot utilization. The proposed system integrates a deterministic Stowage Planning Engine, a Rule Validation Engine, a graph database, and a web-based user interface with a chatbot component. Vessel stability is quantitatively evaluated using the Longitudinal Center of Gravity (LCG), Transversal Center of Gravity (TCG), and Vertical Center of Gravity (VCG) parameters. System evaluation is conducted using two test scenarios involving 20-feet containers with different weight distributions. The results show that the system is able to generate stowage plans that fully comply with safety constraints, maintain vessel stability parameters within operational limits, and consistently identify unplaced containers as a safe outcome when constraints cannot be satisfied. The integration of LLM and GraphRAG serves as an interpretative layer that enhances explainability and transparency of the stowage decisions without replacing the core computational processes. Therefore, the proposed system demonstrates strong potential as a safe, structured, and explainable decision support tool for container stowage planning.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	stowage planning, keselamatan kapal, stabilitas kapal, basis data graf, Large Language Model, Graph Retrieval-Augmented. stowage planning, ship safety, vessel stability, graph database, Large Language Model, Graph Retrieval-Augmented Generation Generation
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Faizcha Aisyah Rahmatillah
Date Deposited:	26 Jan 2026 07:49
Last Modified:	26 Jan 2026 07:49
URI:	http://repository.its.ac.id/id/eprint/130537

Actions (login required)

View Item