A substantial body of literature on global design optimization techniques is based on an implicit assumption that a system under optimization is designed by a perfectly centralized design team with all team members committed to a global optimum solution. This assumption enables the system engineers to minimize all the objective functions simultaneously while controlling all design decisions at the same time. Recent research suggested that this inherent assumption may need to be revisited. This paper proposes a novel framework for design optimization using game theory, which represents a design process as a normal-form game and unifies two fundamental decisions of discipline designers, namely a system-global or a discipline-local optimum solution. This paper provides a general mathematical formulation of the framework for a design process with any number of discipline designers. The framework introduces the notion of the Nash front, which is a set of design solutions consisting of the Nash equilibria derived through the reallocation of design authority between discipline designers. This paper then tests the framework on a numerical optimization case study of a Stirling micro combined heat and power plant with two discipline designers. The result indicates that for the system capital cost and the system power output the difference between global Pareto-optimum designs and the Nash front designs on average across five different scenarios of design authority allocation was 22% and 16%, correspondingly. The application of the novel framework in different fields of engineering could help to build models that are more realistic and prevent impractical technical objectives.

中文翻译：

---

使用博弈论进行设计优化

关于全局设计优化技术的大量文献都基于一个隐含的假设，即正在优化的系统是由一个完全集中的设计团队设计的，所有团队成员都致力于全局优化解决方案。这种假设使系统工程师能够同时最小化所有目标函数，同时控制所有设计决策。最近的研究表明，可能需要重新审视这种固有的假设。本文提出了一种使用博弈论进行设计优化的新框架，该框架将设计过程表示为范式博弈，并统一了学科设计者的两个基本决策，即系统全局或学科局部最优解。本文为具有任意数量学科设计师的设计过程提供了框架的一般数学公式。该框架引入了纳什前沿的概念，它是一组设计解决方案，由通过学科设计者之间设计权威的重新分配得出的纳什均衡组成。然后，本文与两名学科设计人员在斯特林微型热电联产电厂的数值优化案例研究中测试该框架。结果表明，对于系统资本成本和系统功率输出，全局帕累托最优设计和纳什前沿设计在五种不同设计权限分配场景中的平均差异分别为 22% 和 16%。