Vibration Analysis Prediction of Ship Engine Room Panels Using Neural Network Algorithm

Mardyanto, Enrico Edward (2025) Vibration Analysis Prediction of Ship Engine Room Panels Using Neural Network Algorithm. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5018211077-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (21MB) | Request a copy

Abstract

Struktur kapal sering kali terdiri dari panel lokal yang terbuat dari pelat datar yang ditopang oleh berbagai pengaku, yang membentuk bagian utama kerangka struktural kapal. Panel-panel ini, terutama yang berada di area seperti panel ruang mesin, rentan terhadap resonansi yang disebabkan oleh getaran dari mesin utama dan baling-baling. Metode konvensional untuk menganalisis getaran, termasuk analisis elemen hingga (FEA) dan alat berbasis teori pelat datar, banyak digunakan untuk menentukan frekuensi alami panel ini. Namun, karena banyaknya panel, pendekatan ini dapat memakan waktu dan menuntut komputasi. Studi ini memperkenalkan solusi pembelajaran mesin menggunakan jaringan saraf untuk memperkirakan frekuensi alami panel kapal lokal dengan cepat dan akurat. Sebuah kapal tanker minyak dipilih sebagai studi kasus, dengan FEA digunakan untuk menentukan frekuensi alami primer di zona struktural utama. Jaringan saraf dilatih berdasarkan parameter masukan seperti luas pelat, ketebalan pelat, jumlah pengaku, luas pengaku, massa jenis fluida (berlaku untuk panel tangki), sisi kontak fluida, dan kekakuan panel. Keluaran model mencakup frekuensi alami dasar panel dan massa total yang sesuai. Untuk memastikan akurasi, performa model dievaluasi dengan menyesuaikan parameter seperti jumlah neuron tersembunyi. Hasilnya menunjukkan bahwa metode ini dapat berfungsi sebagai alat yang cepat dan andal untuk memprediksi getaran dalam desain struktur kapal.
==========================================================================================================================================
Ship structures often consist of localised panels made from flat plates supported by various stiffeners, forming a major part of the vessel's structural framework. These panels, especially those in areas like the engine room panels are susceptible to resonance caused by vibrations from the main engine and propeller. Conventional methods for analysing vibration, including finite element analysis (FEA) and flat plate theory-based tools, are widely used to determine the natural frequencies of these panels. However, due to the large number of panels, these approaches can be both time-consuming and computationally demanding. This study introduces a machine learning solution using a neural network to quickly and accurately estimate the natural frequencies of local ship panels. An oil tanker is selected as the case study, with FEA employed to determine the primary natural frequencies in key structural zones. The neural network is trained on input parameters such as the area of the plate, the thickness of the plate, the number of stiffeners, the area of stiffeners, the density of fluid (applies for the tank panel), the fluid contact side, and stiffness of the panel. The model outputs include the panels' fundamental natural frequencies and corresponding total mass. To ensure accuracy, model performance is evaluated by adjusting parameters like the number of hidden neurons. The results indicate that this method can serve as a fast and reliable tool for vibration prediction in ship structural design.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Ship Engine Room Panel Vibration, Finite Element Analysis (FEA), Neural Network Algorithm, Backpropagation Neural Network, Natural Frequency Estimation, Total Mass Estimation
Subjects:	V Naval Science > VM Naval architecture. Shipbuilding. Marine engineering
Divisions:	Faculty of Marine Technology (MARTECH) > Naval Architecture and Shipbuilding Engineering > 36201-(S1) Undergraduate Thesis
Depositing User:	Enrico Edward Mardyanto
Date Deposited:	05 Aug 2025 03:39
Last Modified:	05 Aug 2025 03:39
URI:	http://repository.its.ac.id/id/eprint/125829

Actions (login required)

View Item