Introduction

Design for modular construction of ship hull can be defined as a design to reduce construction costs and time to a minimum, compatible with the requirements of the ship hull to fulfill its operational functions with acceptable safety, reliability and efficiency. It is important that the design for modular construction effort starts early in the design process. The designer has the greatest influence on the cost of the vessel during the early design stages. This influence drops off quite rapidly in the later design stages. But at the early design stages the uncertainties that affect the decision making process of the designer will be high. The management of these uncertainties is very important for the design for modular construction to be effective. In this thesis, a system to help the designer in design for modular construction by the effective management of various uncertainties in the early design stage is proposed.

Uncertainties in design for modular construction

Some of the key factors that affect the design for modular ship hull construction considered in this research are module division planning, assembly process planning, weld deformation and rework of each module and modular construction scheduling. These factors are related to each other. The management of these factors for the optimization of design for modular construction at the early design stage is difficult due to the uncertainties at that stage.

The factors affecting module division planning are many and uncertain at the early design stage. The weight and size of modules depend on the facility constraints, which are uncertain at the early design stages. Also, the accurate estimation of weight and size of modules, generated from module division planning at the early design stage, is difficult. Since the structure of the modules itself is uncertain at the early design stage, the amount of welding between modules at the erection stage is also uncertain. Similar is the case of work content of each module. Also there are uncertainties in the amount of modular outfitting done and the locations of outfit components. In order to develop a more flexible and optimum design for module division, the consideration and management of uncertainty at the early design stage is very important.

At the early design stages, the choice of assembly sequence from a range of options available is also difficult. Also the selection and allocation of the production resources is also difficult and uncertain. This makes the estimation of the assembly process time and cost for a module at the early design stage difficult. Also there are limitations in the tools available to the designer in analyzing and evaluating assembly strategies, assembly sequences and estimating assembly time and cost. Predictions of module inaccuracies like weld deformations and associated rework time and cost and its management are also difficult at the early design stage, due to the uncertainties in the weld parameters affecting weld deformations. The uncertainties in the intermediate tasks like the module division planning, assembly sequence, module assembly time and rework may affect the overall construction time and cost schedule.

The uncertainty factors that affect design for modular construction can be broadly classified as shown in Fig.1. These are uncertainties related to information, decision-making, estimation, analysis tools, physical processes, performance of worker or equipment and engineering change orders. These are related to each other and cannot be considered independently. This makes the management of uncertainties more difficult.

Aim of the research

As mentioned in the previous section, the design for modular construction should start at the early design stage because the designer will be able to control the overall cost and time of production more effectively at this stage than the later stages. But the uncertainty factors are high at this stage. As shown in Fig.2, as the design progresses to more advanced stages and as the production starts, the uncertainties reduces, but at the same time the cost reduction opportunities will also be reduced.

The aim of the research is to develop a system to help the designer involved in the design for modular construction in the process of decision-making at the early design stage by effective management of the uncertainties. The major factors addressed are module division planning, assembly planning of each module, accuracy management （mainly weld deformations and associated rework） of modules and the associated uncertainties, and the effect of uncertainty at the intermediate stages on the overall modular construction time schedule （Fig.3）.

Overview of the proposed system

Fig.4 shows the overview of the system proposed in this research. The system consists of four main parts: - Fuzzy logic based module division planning, module assembly analysis, module weld deformation variation and rework variation analysis and modular construction scheduling risk management systems. The details of each of the components are explained briefly in the following sections.

Module division planning system

The aim of the module division planning system is to help the designer in module division planning by generating various module division patterns considering the designer's built strategy or design intentions and evaluating each pattern considering various design and manufacturing constraints and associated uncertainties. The system is developed based on the concept of module division using candidate seams, graph theory and fuzzy logic as shown in Fig.5. At first, the designer will input 'candidate seams' in locations that are considered suitable for module division. The ship hull structure is divided to 'modules' based on the candidate seams. A liaison graph of modules is generated. In the liaison graph, a node represents a module and a link between nodes represents the connection between two modules. Module division patterns are generated by 'cutting' the module liaison graph using cut-set method. In the next step, module division patterns are evaluated and ranked using fuzzy logic. The uncertainties in the various evaluation factors are expressed as fuzzy sets. The candidate seams used to generate each pattern and the modules in each pattern are evaluated. Other evaluation criteria defined by the designer can also be considered. Module division patterns can be ranked based on this evaluation. The designer can carry out further analysis on the selected module division patterns.

Module assembly analysis system

The module assembly analysis system helps the designer in evaluating various assembly sequences of modules, generated from module division planning, by analyzing each assembly stage from individual parts to the final assembly considering designer's assembly strategy. The uncertainly in the assembly time at each stage is considered and final assembly time uncertainty is calculated. As shown in Fig.6, the system is based on graph model or liaison diagram of the module structure, assembly sequence generation from the liaison diagram considering the assembly preferences defined by the designer in the liaison diagram, analysis of each assembly stage from the assembly sequences, generation of assembly PERT chart from selected assembly sequence and Monte-Carlo simulation of assembly PERT chart for the estimation of risk of variations in assembly.

Module weld deformation analysis system

The module weld deformation analysis system helps the designer in calculating the weld deformation values of each modules as well as variations in deformation values considering the variations in weld parameters, mainly heat input variation. Considering a selected assembly sequence of the module and the weld postures, weld heat input and its variations at each weld line are identified. The force/moment equivalent of the hear input is calculated based on experimental values and defined on the FEM model generated from the module structure model. The weld deformation values of the module and its variations are calculated using FEM analysis of the model and Monte-Carlo simulations. Fig.7 shows an outline of the proposed weld deformation analysis system.

Module rework analysis system

The rework time associated with the weld deformations and its variations are calculated using the module rework analysis system. As shown in Fig.8, the calculation is done considering the deformation values at the module interface and its variations. The root gap and misalignment values at the module interface and the associated rework activities are identified. From the accuracy-labor functions defined by the designer, total rework time and its variations are calculated using Monte Carlo simulations. The optimization of the module positioning in order to reduce the rework can also be done using the system.

Modular construction schedule risk management system

The aim of the modular construction time scheduling risk management system is to help the designer in effectively managing the risk in modular construction time scheduling due to uncertainty in each intermediate activities, especially the risk due to the uncertainty in module assembly time and module rework time due to weld deformations. First, a module division plan is selected and analyses of the modules are done based on the systems explained in the previous sections and activity times at each stage are estimated. Then a modular construction schedule network is generated for the selected plan. Using the Monte-Carlo simulation of modular construction schedule network, total construction time and its variation are estimated. The risk is identified as the probability of the total estimated time exceeding the planned time. For the management of this risk by reducing the total modular construction time variation, alternate module division plan, alternate construction plan for each module and alternate selection of weld resources can be considered. Using the system, for given module division plan and construction plan for each module, risk management can be done by the optimization of the resources, by minimizing the total resource management cost and also considering the constraints in resources. Outline of the proposed system is as shown in Fig.9.

The system, proposed in this research, is implemented using VisualWorks Smalltalk Release 2.5.2J and based on SODAS （System Of Design and Assembly in Shipbuilding）, a CIM system developed by Manufacturing System Engineering Laboratory of Department of Environmental and Ocean Engineering, University of Tokyo.

Conclusion

The major contribution made by the research outlined in this thesis is the development of the system for the integration of some of the critical design for modular construction activities and the management of uncertainties in these activities at the initial design stages, considering the overall design for modular construction. The important design for modular construction activities considered are module division planning, assembly process planning of each module, module weld deformation and associated rework and modular construction time scheduling. Using the proposed system, designer will be able to make decisions in the design process aimed at modular construction by considering all the design activities and the associated uncertainties, both simultaneously.

Fig.1 Uncertainties in design for modular construction

Fig.2 Uncertainty and cost determination variation along design stages

Fig.3 Design for modular construction considering uncertainties

Fig.4 An overview of the proposed system

Fig.5 Outline of module division planning system

Fig.6 Outline of the assembly process planning system

Fig.7 An outline of the proposed weld deformation analysis

Fig.8 Outline of the module rework analysis system

Fig.9 Outline of the proposed risk management

第二次世界大戦後において高度に船舶建造を発展させる要因として「ブロック建造法：Modular Construction System」が存在する。この建造法は，大規模な鋼構造物である船舶を複数のブロックに分割することで，造船工場内でそれらのブロックを同時並行的に建造し，ドックや船台における効率的なブロックの搭載組立を実現化するものである。単に，工期短縮などの生産性向上のメリットだけでなく，船舶の溶接変形などの防止など，品質管理においても重要とされる建造方法である。したがって，コスト削減や品質確保などにおいて「良いブロック」を設計することは極めて重要である。しかしながら，現状の造船設計においては，過去の経験や熟練設計者のノウハウに依存する形態が通常であり，システマティックにブロックの設計である「ブロック分割」を実施することが待望されている状況にある。

本研究は，以上の要望に対して，造船設計における「ブロック分割」を高度なコンカレント・エンジニアリングの問題と捉え，船舶の上流設計における不確定な設計情報や建造工程における溶接変形などの不確定な要因を考慮する考え方を整理し，船舶設計の上流段階から「ブロック分割」を効果的に実行できる設計手法を提案するものである。本論文は，全9章から構成されている。

第1章では，「ブロック分割」の重要性および特徴などを整理し，造船設計および建造工程における不確定な因子を考慮する必要性も含め，本研究の背景と進め方に関して述べている。

第2章では，本研究の主題である「ブロック分割」に関する既存研究の概要を分類整理し，問題点を述べることにより，本研究の動機付けおよび重要性を示すと共に，位置付けと着眼点を明確に示している。

第3章では，船舶の初期設計段階を対象に，「ブロック分割」の特徴と目的を整理している。初期設計段階の「ブロック分割」の検討は，後続の詳細設計に大きく影響を与えるため，工場の設備制約や建造性を考慮した出来る限り良い分割案を確定する必要性がある。本研究では，初期設計段階における未確定な船舶の構造，艤装情報を対象とした「ブロック分割」を実現するために，情報の未確定に対してファジィ・モデルを適用し，ブロック分割案をファジィ推論を活用して定量的に評価することを提案している。

第4章では，搭載工程および建造工程の建造日程を考慮したブロック分割案の評価方法に関して議論している。初期設計段階における設計情報の未確定性などが起因して，建造時数を正確に見積もることは困難である。さらに，組立手順，溶接作業者などの作業者の割当も未確定である。本研究では，これらの未確定な情報から建造時数を考慮するために，作業時間のバラツキを導入し，そのバラツキを考慮した上での組立手順の導出方法を提案している。

第5章では，ブロックの溶接変形と変形ブロックの修正作業を考慮したブロック分割を評価する方法を議論している。本研究では，溶接変形量のバラツキを考慮するために，作業者のスキルに応じた入熱量のバラツキを考慮し，変形量とそのバラツキを簡易数値計算モデルで算出し，変形の修正に必要な作業量を求める方法を示している。これによって，ブロックの建造に対して溶接による変形量だけではなく変形量のバラツキを考慮した修正作業のバラツキを定量データとして示すことを提案している。さらに，第6章では，ブロックの変形量のバラツキを最小に制御するために，溶接部に対する適切な溶接作業者の割当方法の必要性を指摘し，その割当の最適化手法を提案している。

第7章では，第4章から第6章までのブロック分割と建造日程の議論を踏まえ，ブロック分割案の評価としての建造日程の遅延リスクの考え方について述べている。ブロック分割案から算出される建造工程の日程には，建造時数の不確定性，溶接変形の不確定性などの不確定性が含まれる。これらの不確定性が因子となる建造日程の遅延リスクを定義し，その遅延リスクのリスク・マネジメントに関して議論している。

第8章では，第3章から第7章の議論によって導かれたブロック分割案の生成と様々な不確定要因を考慮したブロック分割案の評価方法の計算機上への実装方法を示し，構築したプロトタイプ・システムの適用によって，本研究の提案の有用性を議論している。

最終章となる第9章では，本研究で得られた知見を整理し，今後の課題を議論している。

本研究は，船舶の建造コストを確定するといっても過言ではない「ブロック分割」という設計問題に対して，コンカレント・デザイン性に起因する不確定性，および溶接変形の不確定性などを考慮する設計手法を提案している。「ブロック分割」という熟練設計技術を，未確定な情報や不確定な物理現象に対してファジィ理論やバラツキを考慮したリスクマネジメント手法などの工学的手法を適用し，システマティックに意思決定できる設計方法を具体的に示している。

本研究が示す成果により，造船設計における設計の高度化，熟練設計者のリタイヤ問題に対する知識マネジメントなどを実現化することが期待でき，より，技術的かつ，コスト的に競争力が望まれる造船設計の実現に期待がもたれる。このように，本研究が示す方法論，成果の効果はきわめて大きいものと評価できる。

よって本論文は博士（工学）の学位請求論文として合格と認められる。