Optimization for robust design: Integrating model-based systems engineering with multi-criteria decision-making support in a distributed framework

复杂的制造产品，如汽车，涉及大量工程部件的设计和集成，这些部件相互作用以产生产品的整体质量。例如，一辆汽车的总体二氧化碳排放量取决于发动机、传动系统和排气系统的设计，所有这些系统都在一起工作。通常，不同的设计师团队负责飞行器的不同部分，可能因不同的物理位置、不同的设计时间尺度、不同科学学科的专业知识和语言而分离。这些差异有时意味着需要昂贵的返工才能成功解决设计中特定部分的不兼容或不希望的选择。该项目旨在开发方法，以支持这些设计团队为整体先进产品（例如整车）开发高质量的设计，并选择在实践中表现良好的设计-一旦它们被制造并交付给客户。这些方法将以捷豹路虎（Jaguar Land Rover）汽车设计的两个案例进行试点，项目期间开发的软件将作为开源软件提供给所有人使用。该研究将以三个相关学科的现有知识为基础：多目标优化（使用搜索方法识别并向设计师揭示产品性能不同方面之间的权衡），多标准决策（帮助设计师从权衡选项中选择单个首选设计）和多学科优化（帮助组织和整合不同设计团队之间的良好解决方案的搜索）。这项工作还将利用并进一步发展一个正在进行的相关研究项目的结果，该项目旨在了解复杂制造产品不同部分之间的关系，并确定可用于寻找良好解决方案的计算机模型。寻找在实践中表现良好的良好设计被称为“稳健优化”。该研究将开发在设计“节点”层面的稳健优化方法，即在整体产品中的单个组件或子系统层面。该方法将有效地利用可用的计算资源来搜索良好的设计，并将同时优化产品性能的多个方面。将从搜索过程中生成的数据中识别不同设计选择和不同可能的性能权衡之间的关系中的任何模式，并将其呈现给设计人员，以帮助他们选择单个首选设计。该研究将捕捉一个节点上的备选设计选择对其他节点上的设计活动的影响。研究的目的是利用这些信息在不同的节点上进行稳健的优化，同时，对整个产品进行有利的设计。将使用系统工程学科的方法开发协议，当在整个产品设计过程中在其他地方采取的决策影响到特定的产品需求或解决对与团队相关的性能方面不利的权衡时，可以向设计团队强调。研究目标的成功交付将代表优化和决策支持工具能力的逐步变化，以支持设计团队在复杂的工程设计过程中工作，这是现代先进制造产品的关键特征。对捷豹路虎战略利益的实际汽车设计问题的方法的成功测试将有助于证明这些方法在英国更普遍的先进制造业的价值，跨越广泛的行业。

Complex manufactured products, such as automotive vehicles, involve the design and integration of a large number of engineered components that interact to produce the overall qualities of the product. For example, the overall CO2 emissions from a vehicle depend on the design of the engine, transmission and exhaust systems, all working together. Usually different teams of designers work on the different parts of the vehicle, separated potentially by different physical locations, different design timescales, and the expertise and language of different scientific disciplines. These differences sometimes mean that costly re-work is required to successfully resolve incompatible or undesirable choices made for particular parts of the design. This project aims to develop methods that support these design teams in developing high quality designs for the overall advanced product (e.g. the full vehicle), and in choosing designs that will actually perform well in practice - once they have been manufactured and delivered to customers. The methods will be piloted using two case studies of vehicle design for Jaguar Land Rover and the software that is developed during the project will be made available as open-source, for everyone to use.The research will build on existing knowledge in three related academic disciplines: multi-objective optimization (which uses search methods to identify and reveal to designers the trade-offs between different aspects of product performance), multiple criteria decision-making (which helps designers to choose a single preferred design from amongst the trade-off options) and multidisciplinary optimization (which helps organize and integrate the search for good solutions between different design teams). The work will also use results from, and develop further, an on-going related research project that aims to understand the relationships between the different parts of a complex manufactured product and to identify computer models that can be used to inform the search for good solutions.The search for good designs that perform well in practice is known as "robust optimization". The research will develop methods for robust optimization at the level of a design "node" - that is, at the level of an individual component or sub-system within the overall product. The methods will make efficient use of the computing resources available for searching for good designs and will optimize multiple aspects of product performance simultaneously. Any patterns in the relationship between different design choices and different possible performance trade-offs will be identified from the data generated during the search and presented to the designer to help with choosing a single preferred design.The research will capture the impact of alternative design choices at one node on design activities at other nodes. The research then aims to use this information to steer robust optimization at different nodes, simultaneously, towards a favourable design for the overall product. Protocols will be developed, using methods from the academic discipline of systems engineering, that can highlight to design teams when decisions being taken elsewhere in the overall product design process are impacting on particular product requirements or resolving trade-offs unfavourably for aspects of performance relevant to the team.Successful delivery of the research objectives will represent a step-change in the ability of optimization and decision support tools to support the design teams working within the complex arrangement of engineering design processes that are a crucial feature of modern advanced manufactured products. Successful testing of the methods on real automotive design problems of strategic interest to Jaguar Land Rover will help to demonstrate the value of these methods to advanced manufacturing more generally in the UK, across a wide range of sectors.