Collaborative Research: Enabling Large-scale Multidisciplinary Design Optimization with Unsteady Simulations: A Hybrid Pseudo-spectral Approach

该项目将开发一个突破性的多学科设计优化（MDO）框架，该框架使用非定常多物理场计算机模拟来自动优化系统性能。缺乏有效的数值算法来缩短具有非定常过程的大型工程系统（如航天器、飞机和风力涡轮机）的设计周期，是这项研究的动机。由于对系统性能和安全性的期望不断提高，这个问题进一步加剧。自动化MDO框架将大大缩短变革系统的设计周期，这些系统有望改善国家的经济繁荣，改变人们的生活和联系方式，例如城市空中出租车和支持太空旅行的系统。此外，该项目将推进大规模工程系统中复杂机制和相互作用的知识，否则仅凭人类直觉很难获得这些知识。该项目还将为代表性不足的少数族裔和K-12学生开展教育和外展活动，以鼓励STEM参与，促进多样性和包容性，并激发学生对工程设计和优化的兴趣。本课题的研究目标是实现基于梯度的大型非定常工程系统多学科设计优化（MDO）。该项目将开发一种新的混合伪谱（HPS）伴随算法，以有效地计算各种学科的非定常梯度。HPS算法的创新之处在于它有效地结合了时准分析的鲁棒性和伪谱伴随的快速性，实现了高维非定常梯度的高效计算。该项目将研究HPS算法的基本特征，并开发一个模块化架构，以耦合大规模非稳态MDO的任意数量学科。它将通过进行考虑流体力学、结构、传热和动力学之间非定常耦合的城市空中机动电动飞机和海上风力涡轮机MDO来展示框架。随着进一步的发展，该框架可以扩展到更多的学科，如控制和多相流。非定常MDO框架将向公众开放，以促进工程设计界的合作。HPS算法是通用的，有望使MDO之外的许多其他基础研究领域受益，包括代理建模、误差和不确定性分析以及机器学习。此外，预计该项目将在工程设计行业产生催化作用，将传统的人工监督设计过程转变为更加自动化的设计过程。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project will develop a breakthrough multidisciplinary design optimization (MDO) framework that uses unsteady multiphysics computer simulations to optimize system performance automatically. The research is motivated by the lack of effective numerical algorithms to shorten the design period for large-scale engineered systems with unsteady processes, such as spacecraft, aircraft, and wind turbines. This issue is further exacerbated by ever-increasing expectations for system performance and safety. The automated MDO framework will significantly reduce the design cycle time for transformative systems that are poised to improve the nation’s economic prosperity and change how people live and connect, such as urban air taxis and systems supporting space travel. Furthermore, this project will advance the knowledge of complex mechanisms and interactions in large-scale engineered systems, which would otherwise be hard to obtain solely by human intuition. This project will also conduct educational and outreach activities for underrepresented minority and K-12 students to encourage STEM engagement, promote diversity and inclusion, and stimulate students' interest in engineering design and optimization.The research objective of this project is to enable the gradient-based multidisciplinary design optimization (MDO) of large-scale engineered systems governed by unsteady processes. The project will develop a new hybrid pseudo-spectral (HPS) adjoint algorithm to compute unsteady gradients for a broad range of disciplines efficiently. The originality of the HPS algorithm is that it effectively combines the robustness of time-accurate analysis and the speed of pseudo-spectral adjoint to enable efficient computation of high-dimensional unsteady gradients. The project will investigate the fundamental characteristics of the HPS algorithm and develop a modular architecture to couple any number of disciplines for large-scale unsteady MDO. It will demonstrate the framework by conducting urban air mobility electric aircraft and offshore wind turbine MDO that considers the unsteady coupling between fluid mechanics, structures, heat transfer, and dynamics. With further development, the framework can be extended to more disciplines, such as control and multiphase flow. The unsteady MDO framework will be open to the public to promote collaborations in the engineering design community. The HPS algorithm is general and expected to benefit many other fundamental research areas beyond MDO, including surrogate modeling, error and uncertainty analyses, and machine learning. Moreover, this project is anticipated to create a catalytic effect in the engineering design industry to transform the traditional, human-supervised design process into a more automated one.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.