Lecture Notes in Civil Engineering

Tetsuo OkadaKatsuyuki SuzukiYasumi Kawamura   Editors

Practical Design of Ships and Other Floating StructuresProceedings of the 14th International Symposium, PRADS 2019, September 22–26, 2019, Yokohama, Japan- Volume III

Lecture Notes in Civil Engineering

Volume 65

Series Editors

Marco di Prisco, Politecnico di Milano, Milano, Italy

Sheng-Hong Chen, School of Water Resources and Hydropower Engineering,Wuhan University, Wuhan, China

Ioannis Vayas, Institute of Steel Structures, National Technical University ofAthens, Athens, Greece

Sanjay Kumar Shukla, School of Engineering, Edith Cowan University, Joondalup,WA, Australia

Anuj Sharma, Iowa State University, Ames, IA, USA

Nagesh Kumar, Department of Civil Engineering, Indian Institute of ScienceBangalore, Bengaluru, Karnataka, India

Chien Ming Wang, School of Civil Engineering, The University of Queensland,Brisbane, QLD, Australia

Lecture Notes in Civil Engineering (LNCE) publishes the latest developments inCivil Engineering - quickly, informally and in top quality. Though original researchreported in proceedings and post-proceedings represents the core of LNCE, editedvolumes of exceptionally high quality and interest may also be considered forpublication. Volumes published in LNCE embrace all aspects and subfields of, aswell as new challenges in, Civil Engineering. Topics in the series include:

• Construction and Structural Mechanics• Building Materials• Concrete, Steel and Timber Structures• Geotechnical Engineering• Earthquake Engineering• Coastal Engineering• Ocean and Offshore Engineering; Ships and Floating Structures• Hydraulics, Hydrology and Water Resources Engineering• Environmental Engineering and Sustainability• Structural Health and Monitoring• Surveying and Geographical Information Systems• Indoor Environments• Transportation and Traffic• Risk Analysis• Safety and Security

To submit a proposal or request further information, please contact the appropriateSpringer Editor:

– Mr. Pierpaolo Riva at pierpaolo.riva@springer.com (Europe and Americas);– Ms. Swati Meherishi at swati.meherishi@springer.com (Asia - except China,

and Australia, New Zealand);– Dr. Mengchu Huang at mengchu.huang@springer.com (China).

All books in the series now indexed by Scopus and EI Compendex database!

More information about this series at http://www.springer.com/series/15087

Tetsuo Okada • Katsuyuki Suzuki •

Yasumi KawamuraEditors

Practical Design of Shipsand Other Floating StructuresProceedings of the 14th InternationalSymposium, PRADS 2019,September 22–26, 2019,Yokohama, Japan- Volume III

123

EditorsTetsuo OkadaFaculty of EngineeringYokohama National UniversityYokohama, Kanagawa, Japan

Yasumi KawamuraFaculty of EngineeringYokohama National UniversityYokohama, Japan

Katsuyuki SuzukiSchool of EngineeringThe University of TokyoBunkyo-ku, Tokyo, Japan

ISSN 2366-2557 ISSN 2366-2565 (electronic)Lecture Notes in Civil EngineeringISBN 978-981-15-4679-2 ISBN 978-981-15-4680-8 (eBook)https://doi.org/10.1007/978-981-15-4680-8

© Springer Nature Singapore Pte Ltd. 2021This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or partof the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmissionor information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilarmethodology now known or hereafter developed.The use of general descriptive names, registered names, trademarks, service marks, etc. in thispublication does not imply, even in the absence of a specific statement, that such names are exempt fromthe relevant protective laws and regulations and therefore free for general use.The publisher, the authors and the editors are safe to assume that the advice and information in thisbook are believed to be true and accurate at the date of publication. Neither the publisher nor theauthors or the editors give a warranty, expressed or implied, with respect to the material containedherein or for any errors or omissions that may have been made. The publisher remains neutral with regardto jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,Singapore

Preface

The first PRADS (International Symposium on Practical Design of Ships and OtherFloating Structures) was held in Tokyo in 1977, in celebration of the 80thanniversary of the Society of Naval Architects of Japan (later the Japan Society ofNaval Architects and Ocean Engineers), where 245 delegates from 26 countriesattended and 56 papers were presented. In the preface of the first PRADS pro-ceedings, the chairman advocated two important principles; one is the emphasis onthe practical application to ship design, and the other is the promotion of interna-tional mutual understanding through discussions and exchange of information onpractical problems in shipbuilding.

Since then, 42 years have passed, our circumstances and environmentexperienced tremendous changes, for example, rapid development of informationtechnology and computer science, expansion of offshore development, and moreand more emphasis on green shipping and renewable energy. Accordingly, thetopics covered by the subsequent PRADS symposia evolved, but we believe thatthe first two principles have been kept in mind throughout by all the participantslike a basso continuo.

Respecting the traditions of these past successful PRADS symposia, we areproud to organize the 14th PRADS Symposium in Yokohama. Yokohama, locatedjust south of Tokyo, is the second largest city in Japan by population. The city hasbeen Japan’s major gateway for international transportation and communication.Yokohama is also a center of excellence for shipbuilding and offshore engineering,with many leading universities, research institutes, and shipping companies.Yokohama is also a tourists’ destination with excellent accessibility from all overthe world.

More than 220 abstracts were accepted. The full papers were peer-reviewed bytwo reviewers for each paper, and about 170 papers were finally selected for pre-sentation. The topics cover hydrodynamics, ship dynamics, structures, machineryand equipment, design and construction, navigation and logistics, and oceanengineering.

v

On this opportunity, we would like to express our sincere appreciation to themain sponsor, ClassNK, and other sponsors, Mitsui O.S.K. Lines Ltd., MitsubishiShipbuilding Co., Ltd., Namura Shipbuilding Co., Ltd., Japan Marine UnitedCorporation, IHI Corporation, and Sumitomo Heavy Industries Marine EngineeringCo., Ltd. In addition, our gratitude is extended to the worldwide reviewers of thefull papers; without whose excellent voluntary works, our high-quality symposiumwas not even made possible. Special thanks are also extended to the local orga-nizing committee members, who devoted themselves to manage everything.

Last but not least, we would like to thank all the delegates for their participationand invite them to enjoy the symposium as well as this exciting city.

Tetsuo OkadaKatsuyuki Suzuki

Yasumi Kawamura

vi Preface

Organization

Standing Committee

Seizo Motora(Honorary Chairman)

Professor Emeritus, The University of Tokyo,Japan

Yasumi Kawamura(Chairman of StandingCommittee)

Yokohama National University, Japan

Tetsuo Okada(Chairman of PRADS2019)

Yokohama National University, Japan

Alan J. Murphy Newcastle University, UKBas Buchner MARIN (Maritime Research Institute

Netherlands), The NetherlandsEnrico Rizzuto University of Genoa, ItalyGe (George) Wang Seastel Marine System (USA) LLC, USAIlson P. Pasqualino Federal University of Rio de Janeiro, BrazilPatrick Kaeding University of Rostock, GermanyQuentin Derbanne Bureau Veritas Marine & Offshore, FranceSeung-Hee Lee Inha University, Republic of KoreaSverre Steen Norwegian University of Science

and Technology, NorwayUlrik D. Nielsen Technical University of Denmark, DenmarkXiaoming Cheng China Ship Scientific Research Center, China

Local Organizing Committee

Tetsuo Okada (Chairman) Yokohama National University, JapanKatsuyuki Suzuki

(Vice Chairman)The University of Tokyo, Japan

vii

Yasumi Kawamura(Vice Chairman)

Yokohama National University, Japan

Daisuke Yanagihara Kyushu University, JapanKazuhiro Aoyama The University of Tokyo, JapanKazuhiro Iijima Osaka University, JapanKentaro Kobayashi Japan Society of Naval Architects and Ocean

Engineers, JapanKoji Gotoh Kyushu University, JapanKunihiro Hamada Hiroshima University, JapanMotohiko Murai Yokohama National University, JapanTakanori Hino Yokohama National University, JapanTsutomu Fukui ClassNK, JapanYasuhira Yamada National Maritime Research Institute, JapanYasuyuki Toda Osaka University, JapanYoshiaki Hirakawa Yokohama National University, JapanYukitaka Yasuzawa Kyushu University, Japan

Honorary Committee

Shinjiro Mishima Japan Society of Naval Architects and OceanEngineers, Japan

Hiroyuki Yamato National Institute of Maritime, Port and AviationTechnology, Japan

Takanori Kunihiro IHI Corporation, JapanYukito Higaki Imabari Shipbuilding Co., Ltd., JapanTakashi Nakabe Onomichi Dockyard Co., Ltd., JapanYoshinori Mochida Kawasaki Heavy Industries, Ltd., JapanTakashi Ueda Sanoyas Shipbuilding Corporation, JapanKotaro Chiba Japan Marine United Corporation, JapanTetsushi Soga Shin Kurushima Dockyard Co., Ltd., JapanHideshi Shimamoto Sumitomo Heavy Industries Marine

& Engineering Co., Ltd., JapanSachio Okumura Tsuneishi Shipbuilding Co., Ltd., JapanKensuke Namura Namura Shipbuilding Co., Ltd., JapanTetsuro Koga Mitsui E&S Shipbuilding Co., Ltd., JapanKoji Okura Mitsubishi Shipbuilding Co., Ltd., JapanYasuhiko Katoh The Shipbuilders’ Association of Japan, JapanJunichiro Ikeda Mitsui O.S.K. Lines, Ltd., JapanKoichi Fujiwara ClassNK, JapanMasahiko Fujikubo Osaka University, JapanShotaro Uto National Maritime Research Institute, JapanTetsuo Okada Yokohama National University, JapanKatsuyuki Suzuki The University of Tokyo, Japan

viii Organization

Contents

Design and Construction

Design Methodology

A Bayesian Network Approach to Evaluate Shipboard Launchand Recovery Performance in Early-Stage Ship Design . . . . . . . . . . . . . 3James A. Coller and David J. Singer

Communication of Design Space Relationships Learnedby Bayesian Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24Claire Wincott and Matthew Collette

Optimizing Ships Using the Holistic Accelerated ConceptDesign Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38Roy de Winter, Jan Furustam, Thomas Bäck, and Thijs Muller

A “Fast Track to Approval” Process for InnovativeMaritime Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51Stéphane Paboeuf, Arnold de Bruijn, Franz Evegren, Matthias Krause,and Marcel Elenbaas

A Modified Design Framework Based on Markov Decision Processfor Operational Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Hao Yuan and David J. Singer

Ship Design and Optimization

IGC Code-Based LNG Filling Limits Computation Using 3DCAD Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80Seok-Ho Byun, Woo-Sung Kil, Myeong-Jo Son, Jeong-Youl Lee,and Mi-Ho Park

Developments of a Synthesis Model for FLNG Design . . . . . . . . . . . . . . 91Lariuss Zago, Daniel P. Vieira, Yu K. Kam, Raul Dotta,Rodrigo M. Amarante, and Claudio M. P. Sampaio

ix

The First LNG Fuel Train Ferry for St. Petersburg –

Kaliningrad Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106Gennadiy Egorov, Alexander Nilva, and Dmitriy Chernikov

Creation of River-Sea Cruise Passenger Vesselsof New Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120Alexander Egorov, Gennadiy Egorov, and Igor Ilnitskiy

Research on the Accessibility Evaluation of the Cruise PassageDesign Based on Passenger Movement Simulation . . . . . . . . . . . . . . . . . 137Zhan Chen, Wei Cai, Yazhong Wang, and Yigang Wu

Genetic Algorithm Selection for Ship Concept Design . . . . . . . . . . . . . . 156Adam Sobey, Przemyslaw Grudniewski, and Thomas Savasta

Multi Objective Design of Ships; A Pareto Procedure . . . . . . . . . . . . . . 173Sander Calisal

Optimal Arrangement Method of a Ship Consideringthe Performance Against Flooding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190Ki-Su Kim and Myung-Il Roh

Optimal Design of a Floating Raft Vibration Isolation Systemwith Multiple Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198Wenzhang Liu, Wenwei Wu, and Xuewen Yin

A Study of Multi-Objective Optimization for PropulsionPerformance and Cargo Capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209Yasuo Ichinose, Yusuke Tahara, Tomoki Takami, Azumi Kaneko,Takayoshi Masui, and Daisuke Arai

Hull Form Optimization to Fulfil Minimum Propulsion Powerby Using Frequency and Time Domain Potential Flow Solvers . . . . . . . 220Jaekyung Heo, Daehwan Park, Jan Henrik Berg-Jensen, Zhiyuan Pan,and Torgeir Kirkhorn Vada

Welding and Construction

Controller Design of a Gantry Crane for the Safe Erection of Blocksin Shipyards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237Hye-Won Lee, Myung-Il Roh, Seung-Ho Ham, and Do-Hyun Chun

A Study on Using Hot-Rolled Steel Sheet for Ship Superstructure . . . . . 243Hisao Narimatsu, Masakazu Kuwada, and Koji Gotoh

Overview of the Joint Industrial Project for Practical Applicationof Laser-Arc Hybrid Welding in Construction of General MerchantShips in Japan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258Koji Gotoh, Takamori Uemura, Issei Uchino, Hisao Narimatsu,Toshimitsu Maeda, Takumi Torigoe, and Atsuo Moriyama

x Contents

Control of Car Decks Welding Deformation in RO-RO Ships . . . . . . . . 273Junichi Toyoshima, Kiyoharu Hayashi, Mikikazu Murakami,Masanori Sano, Shogo Imasaki, Yasuo Katayama, Hidekazu Murakawa,Naoki Osawa, and Sherif Rashed

Big Data in Ship Design and Construction

Development of Basic Planning Support System UsingMarine Logistics Big Data and Its Application to ShipBasic Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287Kunihiro Hamada, Noritaka Hirata, Kai Ihara,Dimas Angga Fakhri Muzhoffar, and Mohammad Danil Arifin

Evaluation Method for the Maximum Wave Load Based on AISand Hindcast Wave Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308Masayoshi Oka, Tomoki Takami, and Chong Ma

“Monitoring Platform” of Monitoring and Visualizing Systemfor Shipyard: Application to Cutting and Subassembly Processes . . . . . 321Kazuhiro Aoyama, Takayuki Yotsuzuka, Yoshiki Tanaka,and Yoshikazu Tanabe

Monitoring Systems in Design of Ships . . . . . . . . . . . . . . . . . . . . . . . . . 338Serena Lim, Kayvan Pazouki, and Alan J. Murphy

Performance of Ships in Ice

Ice Resistance Calculation in Pack Ice Conditions . . . . . . . . . . . . . . . . . 349Guiyong Zhang, Biye Yang, Yuyan Jiang, and Zhe Sun

On the Development of the Arc7 LNG Carrier – HullForm Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363Ho-Jeong Lim, Ji-In Park, Sahng-Hyon Lee, Sung-Pyo Kim,and Joong Kyoo Kang

A New Technique for Prediction of Ship-Ice Floes InteractionWithout Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373Donghwa Han, Joonmo Choung, and Kuk-Jin Kang

Study on Estimation of Ice Resistance and Attainable Speed for Shipof Arbitrary Shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383Hyun-Soo Kim, Donghwa Han, and Ali Erinc Ozden

Estimation of Ice Conditions Along the Northern Sea Route . . . . . . . . . 397Wei Chai, Bernt J. Leira, Chana Sinsabvarodom, and Wei Shi

Effect of Ship Speed on Bow Impact Load Estimation for PolarClass Ship Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409Mahmud Sazidy, Claude Daley, Bruce Colbourne, and John MacKay

Contents xi

Navigation and Logistics

Safety vs. Sustainability – How Much Underpowering of ShipsIs Acceptable? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423Stefan Krüger, Michal Josten, and Leonidas Souflis

Sustainability Assessment of Small Scale Fishing Vessel Operations:A Case Study in Palabuhanratu, Indonesia . . . . . . . . . . . . . . . . . . . . . . 436Vita R. Kurniawati, Richard W. Birmingham, and Alan J. Murphy

Assessing Port Congestion Using Ship Movement Data:A Case Study of Tianjin Port . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462Guanlan Liu, Alessio Tei, and Alan J. Murphy

Numerical Simulation of Hybrid Platform Supply Vessel (PSV)Fuel Consumption for the Pre-Salt Layer in Brazil . . . . . . . . . . . . . . . . 478Kazuo Nishimoto, Claudio M. P. Sampaio, Rodrigo J. Vale,and Felipe Ruggeri

Optimization of River-Sea-Going Ship Type Scheme and OperationStrategy Under Emission Control Area in China . . . . . . . . . . . . . . . . . . 498Xinjing Tan, Wei Cai, and Chao Liu

Estimation of the Collision Risk on Planned Route . . . . . . . . . . . . . . . . 516Makiko Minami and Ruri Shoji

Autonomous Route Planning for Ships in Complex Waterways . . . . . . . 533Jinsong Xu, Chunxiao Hou, Rongwu Yang, and Ge (George) Wang

Ocean Engineering

Hydrodynamic Analysis of Offshore Aquaculture Platform and ItsMooring Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551Jun Yu, Xiaoming Cheng, and Xiaofeng Wu

Prediction of Extreme Nonlinear Hydrodynamic Responsesand Mooring Line Loads of Floating Offshore Structures . . . . . . . . . . . 564Dong-Hyun Lim, Yonghwan Kim, and Seung-Hoon Lee

Design Automation of Mooring Systems for Floating Structures . . . . . . 579Bo Wu, Xiaoming Cheng, Ying Chen, Xinyun Ni, and Kai Zhang

A Study of Riser Oscillations Caused by Slug Flows During SubseaPetroleum Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595Sergio Nascimento Bordalo and Celso Kazuyuki Morooka

A Simplified Nonlinear Analysis Procedure for the Flexible Risersin Subjected to Combined Loading Effects . . . . . . . . . . . . . . . . . . . . . . . 614Jeong Du Kim and Beom-Seon Jang

xii Contents

Modal Parameter Identification of an Electrical Submersible PumpInstalled in a Test Well Using Drop Tests . . . . . . . . . . . . . . . . . . . . . . . 631Ulisses A. Monteiro, Ricardo S. Minette, Ricardo H. R. Gutiérrez,and Luiz A. Vaz

Fatigue Analysis of Vortex-Induced Vibration for Marine Riserswith Top-End Platform Motion Excitations . . . . . . . . . . . . . . . . . . . . . . 639Yuchao Yuan, Hongxiang Xue, Wenyong Tang, and Jun Liu

Real Time Estimation of Local Wave Characteristics from ShipMotions Using Artificial Neural Networks . . . . . . . . . . . . . . . . . . . . . . . 657Bulent Duz, Bart Mak, Remco Hageman, and Nicola Grasso

Nonlinear Wave Loads on Deep-Water Semisubmersibles . . . . . . . . . . . 679Rahul Manohar, Xin Wang, Velizar Prampakov, and Arun Dev

Experimental and Numerical Comparisons of a Floating AbsorberWave Energy Converter in Regular Waves . . . . . . . . . . . . . . . . . . . . . . 697A. Novás-Cortés, L. Santiago-Caamaño, M. Miguez-González,and V. Díaz-Casas

Study on Motion Response of a Floating Offshore Turbines’ Unitwith OWC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 707Motohiko Murai and Ryohei Mochiduki

Development and Design of a Floating Type Ocean CurrentTurbine System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 732Yasushi Dodo, Shigeki Nagaya, Tetsuo Okada, Makoto Toyoda,and Akio Ito

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 757

Contents xiii

Design and Construction

A Bayesian Network Approach toEvaluate Shipboard Launch and RecoveryPerformance in Early-Stage Ship Design

James A. Coller(B) and David J. Singer

Department of Naval Architecture and Marine Engineering,University of Michigan, Ann Arbor, MI, USA

jcoller@umich.edu

https://ancr.engin.umich.edu/

Abstract. Vessel performance in shipboard launch and recovery oper-ations is a critical element of its overall mission effectiveness. Both com-mercial and military vessels are required to launch and recover a vari-ety of manned and unmanned platforms including submersibles, smallboats, aircraft, and amphibious vehicles. A fundamental attribute ofthese launch and recovery operations is their dependence on a numberof probabilistic factors such as environmental conditions and design fac-tors associated with the vessel and the platforms. To facilitate effectivelaunch and recovery operations, ship designers require a design methodthat takes into account these probabilistic factors to gain a better under-standing the implications of launch and recovery on a variety of earlystage design decisions. Recent advancements in design tools and method-ologies have pushed critical decisions earlier in the design process whichnecessitates evaluating mission critical systems sooner as well.

This paper presents an application of Bayesian networks for evaluatingthe performance of shipboard launch and recovery operations. Bayesiannetworks provide a framework for understanding system interdependen-cies by modeling and analyzing conditional probabilities between vari-able pairs. In previous ship design research, Bayesian networks have beenapplied to risk modeling and the examination of principal particulars.

This research utilizes Bayesian networks to model the environmen-tal and design factors that contribute to the performance of launch andrecovery operations, using a level of information that is consistent withearly-stage ship design. These networks are analyzed and examined toassess the impact of the overall design attributes on those factors. A shipdesign case study is presented to demonstrate the method and elucidatethe driving interdependencies in launch and recovery operations.

Keywords: Bayesian networks · Ship design · Launch and recovery ·Design tools · Unmanned vehicles · Causal dependency

c© Springer Nature Singapore Pte Ltd. 2021T. Okada et al. (Eds.): PRADS 2019, LNCE 65, pp. 3–23, 2021.https://doi.org/10.1007/978-981-15-4680-8_1

4 J. A. Coller and D. J. Singer

1 Introduction

1.1 Problem and Approach

Advancements in naval architecture continue to push the boundaries of howmuch of the design space can be explored for a given ship within time andbudgetary constraints. With methodologies that require large portions of thedesign space to be investigated, such as set based design (SBD), [1] the designspace needs to be explored in a computationally efficient manner while providingnovel information that enables decision making with limited information.

Early stage design requires efficient design tools to effectively examine a givendesign and evaluate the design performance [2]. These tools are used to evaluatethe potential ship designs and ensure that the key design requirements are met.While these tools have inherent biases and will have an impact on the decisionmaking process [3], they remain necessary in modern ship design.

Naval ships are often tasked with the launch and recovery of additionalsystems, both manned and unmanned, including boats, helicopters, unmannedunderwater vehicles (UUVs), unmanned aerial vehicles (UAVs), and unmannedsurface vehicles (USVs). The design of the ship can have a large impact on theeffectiveness of these operations, in addition to environmental conditions andoperating procedures. Design features such as well decks, stern ramps, davits,cranes, and helicopter pads can be complicated to integrate and evaluate.

As the use of unmanned vehicles increases, these operations will continue toincrease in complexity. The US Navy has stated it plans to continue rolling outunmanned vehicles to integrate with the fleet, across the skies [4], the surface[5], and subsea [6]. Unmanned and autonomous vehicles can be more difficultto launch and recover without human operator guidance and with an increasednumber of systems that must flawlessly execute.

This paper investigates a novel approach to this problem by using a Bayesiannetwork to evaluate the environmental and ship design factors for a shipboardlaunch and recovery system. The Bayesian Network will allow the designer toevaluate the interdependencies of the design decision and evaluate the expectedsatisfaction with the overall design. The network structure allows the designerto simulate satisfaction for a wide range of vehicles and complex systems.

The goal of this paper is not to prove the specific network structure is correct,but rather to show that the use of a Bayesian framework can be beneficial to earlystage designers. The specific network definition is dependent upon an individualdesign.

In the remainder of the paper, an introduction to launch and recovery sys-tems and Bayesian networks will be presented. The methodology is explained,and a case study is presented to highlight the network structure and examinevariable sensitivity. Additional opportunities for further work and implicationsare discussed.

Bayesian Networks in Ship Design 5

1.2 Launch and Recovery Design Space

In naval ship design, many mission profiles include launching and recovering anynumber of other vehicles from the ship platform. A given design may be taskedwith having several launch and recovery areas such as well decks, flight decks,and davits [7,8]. Multi-mission ships are tasked with flexible spaces that canserve a wide variety of mission capabilities depending on the call to service andthe changing geo-political climate [9,10].

Within this design space, the possible vehicles to be launched or recovered arenumerous. Categorically, they include aerial, surface, and subsea vehicles. Withinthese categories, dozens to thousands of variants exist. Figures 1 and 2 highlightsome of the variety a single ship may see. Additional unique missions may includerecovering non-propulsive vehicles such as a NASA landing module [11].

Fig. 1. Sailors launch the Rigid HullInflatable Boat (RHIB) aboard theguided-missile destroyer USS Russell(DDG 59). US Navy Released Photo,060619-N-4166B-009.

Fig. 2. A landing craft, air cushion(LCAC) is launched from the welldeck of the USS Boxer (LHD 4).US Navy Released Photo, 190116-N-SH168-1054.

The launch and recovery spaces of a ship can have large impacts on the overallship design. Some launch and recovery systems will dictate large portions of theship arrangements and will be primary design drivers such as flight decks, welldecks, and stern ramps. Other systems, such as davits and articulating cranes,will have minimal impact on the overall design. This makes the analysis of thedesign alternatives an important aspect of the overall design process.

1.3 Bayesian Networks

Bayesian networks are a type of directed acyclic graph (DAG) encoded withconditional probability distributions. The network allows the expression of rela-tionships which can result in simplified knowledge generation and transparencyto relationship dependencies. Bayesian networks allow for both top-down andbottom-up reasoning, which means that the outcome can be predicted, givena set of inputs, or given an outcome, the root cause of the outcome can beidentified [12].

6 J. A. Coller and D. J. Singer

Bayesian networks are directed graphs, meaning they show the direction-ality of relationships. Bayesian networks rely on the use of Baye’s Theorem forBayesian inference, shown in Eq. 1. The posterior distribution, P (θ|y), representsthe probability that is the basis for action, given collected knowledge, y. Thisutilizes information from a prior distribution, P (θ), and the likelihood function,P (y|θ).

P (θ|y) =P (y|θ)P (θ)

P (y)(1)

The marginal distribution, P (y), is a normalizing constant that is often disre-garded, as it can be difficult to determine [13], and the result is that in Bayesianinference we may only show the proportionality of the posterior to the likelihoodand prior, shown in Eq. 2.

P (θ|y) ∝ P (y|θ)P (θ) (2)

Bayesian networks rely on joint probability distributions to combine parentnodes for the child node. In this probabilistic model, each variable, Xi is repre-sented as a node on the DAG. The probability distribution P (X1,X2, ...,Xn) isrepresented as follows:

P (X1,X2, ...,Xn) =n∏

i=1

P (Xi|ΠXi) (3)

where ΠXirepresents the set of nodes that are the parents to Xi in the network

topology.Bayesian networks are widely used in engineering and defense research.

Bayesian networks have been used to predict drinking water distribution pipebreaks [14]. They have also been used for probabilistic assessments for nuclearwaste disposal [15], and predicting air defense threats [16]. In marine structuralengineering, they have been used in structural performance and reliability mod-els [17]. There have been some limited applications in marine design [18], and inmarine decision making [19].

In some of these cases, the network topology was learned from the data [14],while in others it was determined by the authors [16,17].

2 Methodology

The Bayesian framework examined in this paper can be adapted to any networkstructure. The network structure should be set based on the specific design sce-nario and structure of a given ship design project by the designers. The networkshould be representative of the ship design and factors under consideration.

For the purposes of examining the Bayesian network framework, we willuse an existing sample network. The network structure for this methodology isadapted from an influence diagram created by Wireman, 2019 [20]. This modelis intended to show the ability of Bayesian Networks to examine a ship design,

Bayesian Networks in Ship Design 7

and does not necessarily try to perfectly model individual variables such as craneoperating capacities.

In the network, the top level nodes, represented by squares, are the primaryinput variables. The middle level nodes, represented by circles, are intermediatevariables that are of importance and are dependent upon the input nodes. Lastly,the output nodes, represented by diamonds and a hexagon, are calculated metricsthat give probabilistic outputs based on the input and intermediate nodes. Inthe structure, the directed edges between nodes represent the dependencies.

The top level and intermediate nodes are established by the design team.These nodes should represent the high level inputs under consideration and thefunctional dependencies that map from them. These nodes will change betweenships, but this current format gives one example of how this can be done.

The previous influence diagram structure was slightly modified. The variablesincluded are described further in Tables 1 and 2. The full network structure isshown in Fig. 3, with the four primary subtrees shown separately in Figs. 4, 5, 6and 7.

All modeling was performed in Python 2.7. All graphing of the networks wasperformed using the NetworkX, [21], and Graphviz, [22], packages for Python.

Fig. 3. The full network tree developed. Closer views of the four primary subtrees areshown in Figs. 4, 5, 6, and 7.

8 J. A. Coller and D. J. Singer

The model takes in 14 primary input variables, listed in Table 1, which are thefirst nodes below each of the four variable categories. From here, the input nodesare passed to the first layer of calculated intermediate nodes. These intermediatenodes utilize joint conditional probability tables to determine the outcome of theinput node. These joint conditional probability tables are shown in Tables 3, 4,5, 6 and 7.

In instances where the input variable is a rating, Table 3 is used to determinethe probability that the outcome will be satisfactory. Table 4 is used for vari-ables where the weight of the launch or recovery vehicle is dependent. Table 5 isfor variables that are dependent on the vehicle volume, Table 6 is for variablesdependent on the ship freeboard, and Table 7 is used for variables dependent onthe crew size.

In this method, P (θ) is the probability of a satisfactory design outcome, givenobservations yi. There are several intermediate satisfaction metrics, includingFreeboard Satisfaction, θf , Arrangement Flexibility Satisfaction, θa, ManningSatisfaction, θm, Performance Satisfaction, θp, and Vehicle Interoperability Sat-isfaction, θi.

The probability of a satisfactory design outcome was the metric chosen asthe key performance indicator to evaluate the overall design effectiveness fromthe network. This metric could have been any number of things including costeffectiveness or mission effectiveness. This metric was chosen for its simplicity.

There are multiple intermediate variables which contribute to the knowledgebase. In most cases, these are calculated as:

P (θchild) =N∏

i=1

P (θparent,i|yi) (4)

where the child nodes are the dependent intermediate nodes, and the N parentnodes are the nodes above them in the hierarchy which have edges leading tothe child signifying dependence. The parent probabilities are calculated usingthe conditional probability tables previously described.

In the case of the probabilistic estimate of the actual capacity of the launchand recovery system, Ca, it is calculated as follows:

Ca = Cr − 0.5I (5)

where Cr is the rated capacity, and I is the impact of the other dependent nodes,which is calculated as:

I = Cr(1 − P (θ|ywind)P (θ|yss)) (6)

where P (θ|ywind) and P (θ|yss) are the conditional probabilities from the impactof the wind and the sea state, respectively.

Bayesian Networks in Ship Design 9

Table 1. Variables included in the Bayesian network.

Item Description

Host vessel The ship that the secondary vehicle is beinglaunched from or recovered onto

Arrangement flexibility* Rating of if the ship design’s arrangement is flexible

Crew size* Total number of crew members on the host vessel

Minimum freeboard* Minimum freeboard required for the ship, in meters

Available manning Probability there will be enough crew to man thelaunch/recovery

Vehicle The vehicle being launched or recovered

Vehicle volume* The volume of the vehicle, in cubic meters

Vehicle weight* The weight of the vehicle, in metric tons

Variability of vehicle type* Rating of the variability of other vehicles of the sametype

L/R flexibility* Rating of the flexibility the vehicle has for multiplelaunch/recovery methods

Vehicle complexity* Rating of how complex the vehicle is tolaunch/recover

L/R system The launch or recovery system

System requirements* Rating of the overall requirements of thelaunch/recovery system

Arrangement impact Probability that the system does not heavily impactthe ship design’s arrangements

Required manning Probability that the manning requirement of thesystem can be satisfied

Freeboard flexibility* Rating of the flexibility of the system to deal withvarious freeboard amounts

Relative height* Rating of how close to the water level or top of theship the system is

Simplicity of L/R Probability that the system will be simple to execute

Rated capacity* The rated capacity of the system, in metric tons

Actual capacity A probabilistically determined capacity of the system

Environment The environmental conditions

Sea state rating* Rating of the sea state condition and impact on thelaunch and recovery

Wind speeds* Rating of the wind conditions and impact on thelaunch and recovery

Freeboard available Probability of environmental impacts on thefreeboard

* Denotes input variables to the network

10 J. A. Coller and D. J. Singer

Table 2. The satisfaction variables included in the Bayesian network.

Item Description

Arrangement satisfaction Probability that the arrangement of the ship issatisfactory

Manning satisfaction Probability that the manning levels are satisfactory

Freeboard satisfaction Probability that the freeboard levels will besatisfactory

Performance satisfaction Probability that the performance of thelaunch/recovery will be satisfactory

Interoperability satisfaction Probability that the interoperability of multiplevehicles will be satisfactory

Overall satisfaction Overall probability of satisfaction

Table 3. The conditional proba-bility table for the rated variables.

Rating P (Satisfactory)

1 0.99

2 0.98

3 0.95

4 0.80

5 0.50

6 0.20

7 0.10

8 0.05

9 0.01

Table 4. The conditional proba-bility table for the mass variables.

Mass (mt) P (Satisfactory)

1 0.99

5 0.98

10 0.95

20 0.90

30 0.80

40 0.70

50 0.60

100 0.40

200 0.30

In the case of the Performance Satisfaction, the weight of the vehicle needsto be considered compared to the actual capacity of the launch and recoverysystem. In this case, the probability of satisfaction is calculated as follows:

P (θp) =

{0.01, if W > Ca

P (θ|ycmplx)((1 − W/Ca) + 0.25), otherwise(7)

where W is the weight of the vehicle, P (θ|ycmplx) is the probability of satisfactiongiven the complexity of the vehicle, and ((1 − W/Ca) + 0.25) is capped at 1.0.

The overall satisfaction, P (θ), is computed as the geometric mean ofP (θf,a,m,p,i):

P (θ) = [P (θf )P (θa)P (θm)P (θp)P (θi)]1/5 (8)

Bayesian Networks in Ship Design 11

Table 5. The conditional proba-bility table for the volumetric vari-ables.

Volume (m3) P (Satisfactory)

0.1 0.99

1.0 0.95

5.0 0.85

10.0 0.75

20.0 0.55

30.0 0.40

40.0 0.30

50.0 0.22

75.0 0.10

100.0 0.05

150.0 0.01

Table 6. The conditional proba-bility table for the freeboard vari-ables.

Fbd Min (m) P (Satisfactory)

0.5 0.99

1.0 0.95

2.0 0.85

3.0 0.75

4.0 0.65

5.0 0.55

6.0 0.45

7.0 0.35

8.0 0.25

9.0 0.15

10.0 0.05

Table 7. The conditional probability table for the crew-based variables.

Crew size P (Satisfactory)

5 0.01

10 0.08

15 0.15

20 0.20

30 0.30

40 0.40

50 0.50

60 0.60

80 0.72

100 0.80

150 0.90

200 0.95

400 0.99

12 J. A. Coller and D. J. Singer

Fig. 4. The Host Vessel portion of the network is shown.

Fig. 5. The Vehicle portion of the network is shown.

Bayesian Networks in Ship Design 13

Fig. 6. The Launch and Recovery System portion of the network is shown.

Fig. 7. The Environmental portion of the network is shown.

14 J. A. Coller and D. J. Singer

3 Model Sensitivity

To examine the effectiveness of the proposed model, a simple sensitivity analysisis examined. A sample hose vessel, vehicle, launch and recovery system, and envi-ronmental conditions are proposed. From this baseline, each variable is changedindependent of the rest between the minimum and maximum values assessed inthe conditional probability tables. The baseline input variables, as well as themaximum and minimum values, are shown in Table 8.

Table 8. The input variables for the case study. The baseline value is shown alongwith the minimum and maximum values used in the simulation.

Input variable Value

Baseline Min Max

Host ship Arrangement flexibility 3 1 9

Crew size 200 5 400

Freeboard minimum 1 1 9

Vehicle Volume [m3] 15 0.1 150

Weight [mt] 20 0.5 200

Variability 3 1 9

Flexibility 3 1 9

Complexity 2 1 9

L/R system System requirements 2 1 9

Freeboard flexibility 2 1 9

Relative height 2 1 9

Capacity [mt] 40 0.5 300

Environment Sea state 2 1 9

Wind 2 1 9

The baseline variable selection results in an overall probability of satisfactionof 0.76. This is reflected in Fig. 8 and Table 9. As shown, under the baseline

Table 9. The results of the baseline variables through the network.

Variable Value

P (θ) 0.76

P (θf ) 0.91

P (θa) 0.69

P (θm) 0.63

P (θp) 0.69

P (θi) 0.95

Bayesian Networks in Ship Design 15

conditions the overall satisfaction is being driven positively by the vehicle inter-operability and is being driven negatively by the manning requirement.

Fig. 8. The baseline performance data from the initial simulation. The overall sat-isfaction is 0.76, which is largely driven down by the poor probability of manningsatisfaction, arrangement flexibility, and performance.

Using this baseline data as a starting point, 134 test cases are simulated byvarying each input variable independently. From this data, the most useful datawas determined to be the maximum, minimum, and the difference between themaximum and minimum values for each of the satisfaction probabilities. Thisdata is shown in Table 10 and Table 11.

We can validate the model by examining the edge cases. Conceptually, ifthe sea state is extremely large, the probability of a positive outcome with highsatisfaction is low.

This is accurately depicted in Table 10, where min(P (θ)) = 0.02 correspondsto the worst sea state condition. Similarly, not having enough crew to performthe task, having extreme system requirements, having a very complex vehicle,or having a heavy vehicle with a medium capacity launch/recovery system allresult in low probabilities of satisfaction. This makes sense given our intuition.

Conversely, we can see improvements over the baseline across the board aswe loosen restrictions on the maximum value side of Table 10. Many variableswere already near their best case scenarios in the baseline, so only minimal gainswere observed. However in the cases of the variables where the baseline was ofmoderate difficulty (vehicle weight, launch system capacity), larger changes wereobserved, as expected. In no case do we expect a perfect situation, given the restof the baseline variables, which is why the overall satisfaction is never expectedabove 0.89.

16 J. A. Coller and D. J. Singer

Table 10. The minimum and maximum results from the simulation.

Input variable Minimum output Maximum output

P (θ) P (θf ) P (θa) P (θm) P (θp) P (θi) P (θ) P (θf ) P (θa) P (θm) P (θp) P (θi)

Arr flexibility 0.31 0.91 0.01 0.63 0.69 0.95 0.77 0.91 0.72 0.63 0.69 0.95

Crew size 0.31 0.91 0.69 0.01 0.69 0.95 0.77 0.91 0.69 0.66 0.69 0.95

Fbd minimum 0.53 0.14 0.69 0.63 0.69 0.95 0.76 0.91 0.69 0.63 0.69 0.95

Volume 0.75 0.91 0.61 0.63 0.69 0.95 0.78 0.91 0.76 0.63 0.69 0.95

Weight 0.26 0.91 0.69 0.21 0.01 0.95 0.83 0.91 0.69 0.69 0.92 0.95

Variability 0.56 0.91 0.69 0.63 0.69 0.20 0.77 0.91 0.69 0.63 0.69 0.98

Flexibility 0.56 0.91 0.69 0.63 0.69 0.20 0.77 0.91 0.69 0.63 0.69 0.98

Complexity 0.31 0.91 0.69 0.63 0.01 0.95 0.77 0.91 0.69 0.63 0.70 0.95

Sys requirements 0.12 0.91 0.01 0.01 0.69 0.95 0.85 0.91 0.94 0.78 0.69 0.95

Fbd flexibility 0.34 0.01 0.92 0.77 0.69 0.95 0.85 0.92 0.92 0.77 0.69 0.95

Relative height 0.34 0.91 0.92 0.77 0.01 0.95 0.85 0.91 0.92 0.77 0.70 0.95

Capacity 0.36 0.91 0.92 0.77 0.01 0.95 0.89 0.91 0.92 0.77 0.92 0.95

Sea state 0.02 0.01 0.69 0.00 0.01 0.95 0.77 0.92 0.69 0.64 0.70 0.95

Wind 0.05 0.91 0.69 0.00 0.01 0.95 0.77 0.91 0.69 0.64 0.70 0.95

Table 11. The difference between the minimum and maximum results of the simula-tion. A zero indicates that the input variable has no influence on the output.

Input variable Max-Min

P (θ) P (θf ) P (θa) P (θm) P (θp) P (θi)

Host ship Arrangement flexibility 0.46 0.00 0.71 0.00 0.00 0.00

Crew size 0.46 0.00 0.00 0.65 0.00 0.00

Freeboard minimum 0.24 0.77 0.00 0.00 0.00 0.00

Vehicle Volume 0.03 0.00 0.15 0.00 0.00 0.00

Weight 0.56 0.00 0.00 0.48 0.91 0.00

Variability 0.21 0.00 0.00 0.00 0.00 0.78

Flexibility 0.21 0.00 0.00 0.00 0.00 0.78

Complexity 0.46 0.00 0.00 0.00 0.69 0.00

L/R system System requirements 0.73 0.00 0.93 0.77 0.00 0.00

Freeboard flexibility 0.51 0.91 0.00 0.00 0.00 0.00

Relative height 0.51 0.00 0.00 0.00 0.69 0.00

Capacity 0.53 0.00 0.00 0.00 0.91 0.00

Environment Sea state 0.75 0.91 0.00 0.64 0.69 0.00

Wind 0.72 0.00 0.00 0.64 0.69 0.00

These data allow the identification of the key variables in the design byexamining the differences between their minimum and maximum values acrossthe various metrics, shown in Table 11. A result of zero indicates that the variablehas no influence on the predicted satisfaction. A low number indicates it has lowimpact, and a high number indicates that it has a large impact. In this case,the data show that the environment has the largest impact on the overall designsatisfaction outcome.

Bayesian Networks in Ship Design 17

The sea state influences more aspects of the satisfaction outcome than anyother element. This is due to the sea state’s influence across the board on oper-ations. As the sea state intensifies, ship motions increase and situations canpotentially become more dangerous. The safe working load for cranes can bediminished and higher manpower levels can be required for shipboard operations.While individual ships set manning requirements for given ship operations[23],these can often be increased in heavier sea conditions.

The volume of the vehicle on the other hand, influences the outcome satis-faction least, through its impact on the arrangement flexibility. While movinga small man-portable UUV can be simple and have no impact on the arrange-ment flexibility, moving a large helicopter will have a larger impact. However,this larger vehicle will also require a larger launch and recovery system, which isbeing accounted for in the system requirements side of the arrangement impactsas well. The influence paths of these variables are highlighted in Fig. 9.

Fig. 9. The impact of the sea state is (blue path) is high due to its high degree ofconnectivity. The vehicle volume (yellow path) has very little impact due to its linearpath.

The largest impact variable, the sea state, is examined in Fig. 10 and Fig. 11.This shows that the overall output function changes with the sea state’s con-ditional probability function, as expected. They do not match exactly becauseother variables are still influencing the outcome. We can also observe in the