【作者】 操安喜；

【导师】 崔维成；

【作者基本信息】 上海交通大学 ， 船舶与海洋结构物设计制造， 2008， 博士

【摘要】 载人潜水器是一种技术密度高、涉及学科面广的复杂工程系统,涵盖多个学科内容,不仅包括传统的水动力学、结构力学、推进理论、操纵和控制理论等,也包括现代的人机工程学等,每个学科又包含很多分支,载人潜水器的设计具有明显的“多学科”特点。随着结构、水动力、控制等学科理论不断完善,计算机技术的飞速发展,人们能够采用更高精度的模型进行学科分析与设计,各学科设计均取得了长足的进步。然而,与各学科或子系统设计蓬勃发展形成强烈对比的是,长期以来,潜水器总体设计方法的发展一直停滞不前,其理论落后、方法陈旧,虽然这些理论和方法在过去数十年发挥了极其重要的作用,为潜水器技术的发展做出了巨大的贡献,但是它们忽视工程系统中各学科之间的耦合效应、不能充分利用学科的发展成果、低效、耗时且成本高昂,这一系列固有缺陷决定了它们已无法适应现代潜水器发展的迫切需要。在这种情况下,本文引入在航空领域迅速发展起来的解决复杂系统设计与优化的多学科设计优化方法(Multidisciplinary Design Optimization, MDO),探讨多学科优化方法在载人潜水器设计中应用的可行性和适用性。论文以多学科设计优化算法及其在载人潜水器设计中的应用为核心。论文的主要工作包括以下几个方面:1.系统回顾和总结了现有多学科设计优化方法分析了载人潜水器设计过程中所涉及的优化问题及其可能的解决方法。对现有多学科设计优化方法进行了综述。重点描述了三种分布式多学科设计优化方法的产生、发展及其在实际工程中的应用情况,介绍了各方法的计算框架,并分析了它们的优缺点和适用情况。2.系统深入地研究了多学科协同优化方法协同优化方法是一种有效的多学科优化方法,其设计思想、计算框架与现有载人潜水器的设计组织形式相吻合,在载人潜水器设计中最具潜力。(1)详细介绍了协同优化方法的设计思想、计算框架、数学描述形式和求解步骤,指出协同优化方法存在的计算困难问题,并分析了产生计算困难的原因。(2)通过一个典型的数值算例,对现有几种改进型协同优化方法的计算性能进行了比较研究,研究结果表明,基于现代优化算法——遗传算法的协同优化计算性能最为稳定,可靠性最高。3.针对载人潜水器优化设计中的复杂多目标优化问题,发展了能够得到Pareto解集的协同优化方法。针对潜水器设计中涉及多个学科的耦合以及数据信息量大、数据关系复杂的问题,本文发展了一种新的多目标协同优化算法。该方法将Pareto遗传算法(PGA)引入协同优化框架。在PGA与协同优化框架结合的过程中,采用目标函数的归一化处理、分级罚函数法、浮点数编码、群体分级和Pareto解集过滤器等技术提高算法的计算效率和可靠性。二者的有机结合充分发挥各自的优势,该方法利用协同方法的分解协调机制将复杂系统的设计问题分解为一个系统级优化问题和几个学科级优化问题。采用PGA作为系统级优化器,不仅可以得到能够反映多目标优化问题实质的、客观的Pareto解集,而且,由于PGA是无需梯度信息的直接搜索算法,从而从根本上消除了协同优化由系统层一致性约束条件引起的收敛困难问题。最后通过一个数学算例证实了本文发展的多目标协同优化方法的有效性。4.载人潜水器总体多学科模型的建立根据载人潜水器总体设计的特点,将载人潜水器系统划分为外形/水动阻力、推进、能源、结构、重量与容积共5个相对独立的学科。在学科分析的基础上,建立了各学科数学分析模型,提取了设计参数,明确了各学科的输入输出以及它们之间的耦合关系,并编制了相应的分析设计程序。完成了载人潜水器的总体设计的建模工作。5.多学科协同优化方法在载人潜水器设计中的应用研究(1)利用基于遗传算法的单目标协同优化方法,实现了载人潜水器概念设计阶段的总体优化设计。(2)利用本文提出的基于Parato遗传算法的多目标协同优化方法(PGA-CO),实现了载人潜水器概念设计的多目标优化,优化结果表明该方法能得到稳态和均匀的高性能Pareto解集,与单目标优化相比,获得的Pareto解集能使设计者对可能的设计方案有全面认识,更好地进行权衡、折衷和决策。应用研究表明,该方法在载人潜水器总体设计中具有广阔的应用前景。6.响应面近似技术在载人潜水器耐压结构设计中的应用研究对多学科设计优化中的近似模型技术进行了研究,将二次响应面近似模型应用于潜水器载人耐压球壳结构的优化设计中,载人球壳结构采用ABAQUS软件进行有限元分析,采用中心组合试验设计方法获得初始样本点数据信息,通过对数据点的拟合构造了设计参数和设计目标的二阶响应面近似模型,在响应面模型基础上利用PGA算法进行多目标优化求解,最后得到耐压球壳结构的优化设计,在最优设计点处,近似模型达到了较高的精度。这一研究表明:在载人潜水器学科优化设计中,采用响应面近似模型替代原有复杂的、高精度分析模块进行优化迭代计算,极大地减少了计算量,提高优化计算效率,解决了详细设计阶段学科优化中的计算瓶颈问题,具有较强的工程实用性。本文的创新性工作主要体现在以下四个方面:(1)本文结合Pareto多目标遗传算法(PGA)和协同优化框架(CO),首次提出了基于PGA的多目标协同优化算法(PGA-CO),通过数值算例的验证和实际工程的应用表明:该方法能有效解决多目标的多学科优化问题,在类似于载人潜水器等复杂工程系统的总体优化设计中极具应用潜力。(2)将航空航天领域新近发展起来的多学科设计优化技术运用于载人潜水器的总体优化设计中。首次建立了包含水动力外形、推进、能源、结构及重量容积共5个学科在内的载人潜水器总体多学科优化数学模型,并采用多目标多学科协同优化算法(PGA)求解,获得了优化结果。(3)首次将多学科近似模型技术应用于载人潜水器的耐压结构优化设计中,基于响应面近似模型的耐压结构优化设计方法,能显著减少计算量,优化效率高,能够满足工程设计的精度要求,该方法工程实用性强。(4)通过一个典型MDO数学算例对现有的多种协同优化方法进行了比较研究,得到了一些有益的结论,为协同优化算法理论和应用的进一步研究提供参考。通过本文研究,表明多学科设计优化方法在提高载人潜水器总体优化设计技术方面具有巨大潜力,本文工作可作为进一步研究多学科设计优化方法的工程应用基础。更多还原

【Abstract】 Human Occupied Vehicle (HOV) is a complex system involving many different disciplines such as Hydrodynamics, structure, propulsion, weight/volume and cruise control, etc. HOV design is characterized by multidisciplinary interactions in which participating disciplines are intrinsically linked to one another. HOV design is also a complicated multistage process. In conceptual design phase and preliminary design phase, multidisciplinary optimization is especially significant for improving integration performance of HOV. However, for HOV design, it is really difficult to realize multidisciplinary optimization by conventional optimization. The difficulties are that such an integrated implementation is but dealing with the complex couple relationship among the disciplines, also subjected to complexities introduces as a result of a large number of design variables and constraints. The conventional optimization methods for general design of HOV are not capable of solving these problems.Under this circumstance, it is necessary to find new way for optimization design of HOV. Multidisciplinary design optimization (MDO) method has been emerged from aeronautics and astronautics fields, especially for such complicated engineering integrated optimization problems.The objective of this work is to explore MDO method and its application in HOV design. The main contents and contributions of this thesis may be summarized as follows:(1) Exiting MDO methods are reviewed and analyzed. This thesis reviews some of MDO approaches and focuses mainly on solution strategies, characters and recent advances of distributed MDO approaches. The advantages and disadvantages of these methods are discussed and analyzed.(2) Collaborative Optimization (CO) is a potential MDO method. In our work, CO is systematically investigated. Firstly, the motivation, architecture, mathematical description of CO is introduced in detail. Secondly, several varieties of CO are investigated and analyzed. Through numerical examples, the CO based on GA (GA-CO) method is proven to have better convergence performance and higher robust.(3) In order to deal with complicated multi-objective optimization problem in HOV design, multi-objective CO is investigated and developed in our study. We describe the novel integration of Pareto Genetic Algorithm (PGA), one of multi-objective optimization methods within the collaborative optimization framework, which remain the main metrics of CO architecture and ability of PGA to seeking non-inferior solution set. Introduction of PGA which is a direct search algorithm to CO can relieve the convergence difficulties in system-level. At the same time, the PGA enables the designer to select the fittest solution among the Pareto optimal set in according with their preference and the nature of the design problem. We have used some strategies such as regularization of objectives, graded penalized function technique to remove constraints, float code, Pareto rank of population and Pareto set filter of objective in the integration of PGA within CO. Through a numerical examp1e, our developed method is proven to be correct and effective.(4) Establishing the disciplinary analysis model According to the vehicle’s characteristics, HOV system is decomposed into five disciplines such as shape/hydrodynamics, structure, propulsion, energy and weight/volume. And then, each of these disciplines is analyzed, the mathematical models for all disciplines are established, and the inputs and outputs of disciplinary models are defined. These models are proven to be correct and efficient by system analysis, and can be used in the optimization design.(5) The developed PGA-CO in our study is applied to solve the HOV design problem. The PGA-CO is successfully used in the conceptual design of the HOV. A robust and well-distributed noninferior set is obtained, which can help the designers to understand the project and make decisions. Through this application, the presented methods are proven to be applicable and have the potential for multidisciplinary design optimization of HOV.(6) Approximation is one of the most important critical techniques in MDO. In this study, the structure multi-objective optimal design of the pressure spherical hull in the HOV is completed with a combined optimal method and this method is based on Response Surface Method (RSM) and Pareto Genetic Algorithm (PGA). The FEM model of the Pressure Spherical Hull is built firstly by ABAQUS. With the Design of Experiment (DOE), the response property of design objects can be obtained. The response surface model is fitted with these samples. PGA is used in subsequent optimal design. Finally, the optimal design of the pressure spherical hull is obtained. Optimization design of structure based on the response surface model is proven to be efficient and effective.更多还原