Perancangan Sistem Penilaian Otomatis Dokumen Fire Safety Plan Berbasis Rule-Based Reasoning

Anoraga, Mahesa (2024) Perancangan Sistem Penilaian Otomatis Dokumen Fire Safety Plan Berbasis Rule-Based Reasoning. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 5018201069-Undergraduate_Thesis.pdf] Text
5018201069-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2026.

Download (36MB) | Request a copy

Abstract

Kecelakaan kapal merupakan kejadian yang tidak diinginkan, namun kerap terjadi dalam dunia
maritim. Di antara banyaknya penyebab kecelakaan kapal yang mungkin terjadi, kecelakaan
akibat kebakaran dan ledakan perlu mendapatkan perhatian khusus. Kebakaran dan ledakan
merupakan penyebab kecelakaan yang paling mungkin menyebabkan total loss pada sebuah
kecelakaan. Untuk mencegah terjadinya kebakaran pada kapal, fire safety plan disusun untuk
merinci posisi fire safety appliances pada kapal. Studi terkait penilaian otomatis terhadap
dokumen fire safety plan disusun untuk meningkatkan efektivitas waktu dalam proses
perancangan dokumen fire safety plan. Studi ini juga merupakan upaya untuk meningkatkan
tingkat keselamatan dari kebakaran pada kapal dengan memastikan dokumen fire safety plan
yang dirancang sudah lengkap. Algoritma YOLOv5 digunakan sebagai model deteksi karena
YOLOv5 dapat bekerja dengan cepat tanpa memberatkan perangkat yang mengoperasikannya.
Dataset yang digunakan object detection adalah dataset susunan sendiri yang terdiri dari 41
kelas fire safety symbols dengan variasi noise. Hasil deteksi dinilai menggunakan metode belief
rule-base untuk menentukan apakah jumlah fire safety appliances sudah memenuhi standar
minimum sebagai parameter kelengkapan. Dengan 1593 data dan 450 kali iterasi, proses
dataset training menghasilkan nilai precision sebesar 95,5%, recall sebesar 96,8%, mAP50
sebesar 97,5%, dan mAP50-95 sebesar 73,8%. Model tersebut diimplementasikan pada
program object detection dengan akurasi sebesar 95,12% dalam waktu 58,1 detik. Hasil object
detection dinilai oleh penilaian otomatis berbasis belief rule-base dengan tingkat akurasi
95,12% dalam waktu 0,5 detik. Dengan demikian, sistem penilaian otomatis dokumen fire
safety plan ini berhasil mendeteksi fire safety symbols dan melakukan penilaian dengan waktu
yang lebih singkat dibandingkan dengan metode manual.
============================================================
Ship accident is an undesirable event, yet often occur in the maritime sector. Among the number
of possible causes of ship accident, accident due to fire and explosion require a special attention.
Fire and explosion are the most potent cause of accident that would result in a total loss of the
ship. To prevent a fire accident on board, fire safety plan is designed to detail the position of
fire safety appliances onboard. A study about an automatic assessment for fire safety plan is
conducted to increase the time efficiency during the designing process. This study also serves
as an effort to increase the level of fire safety onboard by ensuring that the designed fire safety
plan document is complete. YOLOv5 algorithm is employed as the detection model due to its
fast nature without burdening the operating device. The dataset used for the object detection
program is a self-assembled dataset with 41 fire safety symbols classes with noise variation.
The object detection result is assessed via a belief rule-based method to determine whether the
amount of fire safety appliances meets the minimum standard as a completeness parameter.
With 1593 data in the dataset and 450 itteration, the data training process yields a precision
value of 95,5%, recall value of 96,8%, mAP50 value of 97,5%, and mAP50-95 value of 73,8%.
This model is implemented on the object detection program with the accuracy of 95,12% in
58,1 seconds. The detected symbols are assessed by an automatic assessment system based on
a belief rule-base with the accuracy of 95,12% in 0,5 second. Thus, this fire safety plan
automatic assessment system is successful in detecting fire safety symbols and carrying out
assessment in a shorter time compared to manual methods.

Item Type: Thesis (Other)
Uncontrolled Keywords: fire safety plan, object detection, YOLOv5, belief rule-base, penilaian otomatis.
Subjects: T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering > VM297 Ships Designs and drawings
Divisions: Faculty of Marine Technology (MARTECH) > Naval Architecture and Shipbuilding Engineering > 36201-(S1) Undergraduate Thesis
Depositing User: Mahesa Athala Anoraga
Date Deposited: 05 Aug 2024 05:53
Last Modified: 24 Sep 2024 06:51
URI: http://repository.its.ac.id/id/eprint/112134

Actions (login required)

View Item View Item