CAREER: Integrated Design of Intelligent Structures with Tailored Distributed Damping

本学院早期职业发展（Career）计划研究项目旨在研究设计振动和运动控制智能结构的新方法，创造新的数值设计策略，并对如何最好地设计这类新型智能结构产生基本的理解。现有的智能结构使用集成传感器和执行器，如压电材料，分布在柔性弹性材料的表面或内部来控制动态行为。本文介绍了嵌入式粘弹性材料（VEMs），以克服目前的性能限制。整合VEM具有挑战性，因为1)工程师不能依靠过去的经验来设计这种前所未有的系统，2)精确的VEM模型在计算上是昂贵的。在这里，新的集成设计优化策略将被用于加速这种新型智能结构的设计知识的生成，并减少计算费用。数值和物理实验将以应用于天基望远镜的精确指向为中心。更精确的定位有可能显著增强科学数据收集，包括寻找系外行星。这些进步也有可能推动其他领域的发展，其中超安静结构稳定性或精密运动控制至关重要（例如，制造业，机器人和国防）。这个CAREER项目的教育和推广部分包括创造独特的实践活动，让K-12和本科生体验设计自动化工具的价值。通过与奥菲姆儿童博物馆的“女孩做科学”项目以及与美国宇航局喷气推进实验室的立方体卫星项目的合作，这些项目得到了加强。智能结构（IS）已经得到了广泛的研究，但主要是针对主动阻尼而不是运动控制，并且在很大程度上避免了将空间分布的VEMs用于阻尼。将虚拟机和扩展到运动控制中可以帮助实现新的性能水平，但由于没有设计历史、经过验证的设计指南或专家直觉，因此引入了深刻的设计挑战。当前的IS技术利用空间分布式控制驱动来定制动态行为。最近的研究将控制裁剪与分布式几何弹性子结构设计相结合。剩下的障碍是在实际实现中对高阶结构模式的控制，这里通过战略性分布的vm来被动地抑制高阶结构模式。VEM的分布可以在空间上变化，与弹性材料协同作用，控制分布。这增强了设计的灵活性，但也增加了新的复杂性。构造状态空间导数函数的代理模型并对其进行自适应改进，提出了集成动态系统设计优化的新思路。这利用了动态系统的内在特性来提高数值效率。使用高阶常微分方程（ODE）近似可以精确地建模ems，但计算开销很大。提出了一种新的导数函数代理建模（DFSM）策略，利用具有状态相关参数的高效低阶ode生成高精度的VEM建模。DFSM自适应调整参数映射，提高精度，支持高效的低阶计算。在深入研究阻尼IS设计的DFSM后，DFSM将用于快速的设计探索和系统的设计数据生成。然后将使用机器学习策略来识别这些数据中的模式和关系。迭代归纳过程将用于从这些结果中确定可能的设计指导方针，例如首选的分布式形状关系或传感器/执行器/VEM放置指导方针。额外的数值设计研究将用于验证/完善设计指南。

This Faculty Early Career Development (CAREER) Program research project aims to investigate a fundamentally new approach for designing intelligent structures for vibration and motion control, create new numerical design strategies, and to generate a foundational understanding for how best to design this new class of intelligent structures. Existing intelligent structures use integrated sensors and actuators, such as piezoelectric materials, distributed across the surface or interior of a flexible elastic material to control dynamic behavior. Here embedded viscoelastic materials (VEMs) are introduced to overcome current performance limitations. Incorporating VEMs is challenging because 1) engineers cannot rely on past experience to design this unprecedented system, and 2) accurate VEM models are computationally expensive. Here new integrated design optimization strategies will be used to accelerate generation of design knowledge for this new type of intelligent structure, and to reduce computational expense. Numerical and physical experiments will center on application to precision pointing for space-based telescopes. More precise pointing has the potential to enhance significantly scientific data gathering, including search for exoplanets. These advances also have potential to advance other domains where ultra-quiet structural stability or precision motion control is critical (e.g., manufacturing, robotics, and defense). Education and outreach components of this CAREER project involve the creation of unique hands-on activities that allow K-12 and undergraduate students to experience the value of design automation tools. These are enhanced by collaborations with the "Girls do Science" program at the Orpheum Children's Museum and a CubeSat project with NASA's Jet Propulsion Laboratory.Intelligent structures (IS) have been studied extensively, but primarily for active damping as opposed to motion control, and have largely avoided incorporation of spatially distributed VEMs for damping. Inclusion of VEMs and extension to motion control could help achieve new performance levels, but introduces a profound design challenge as no design history, validated design guidelines, or expert intuition exist. Current IS technology utilizes spatially distributed control actuation to tailor dynamic behavior. Recent work has combined control tailoring with distributed geometric elastic substructure design. A remaining obstacle is control of high-order structural modes in practical implementations, and is addressed here via strategically distributed VEMs to damp high-order modes passively. VEM distribution can be varied spatially to synergize with elastic material and control distribution. This enhances design flexibility, but this adds a new level of complexity. A new concept for integrated dynamic system design optimization where surrogate models of the state space derivative function are constructed and improved adaptively is planned. This capitalizes on the intrinsic properties of dynamic systems for numerical efficiency. VEMs can be modeled accurately, but with great computational expense, using high-order ordinary differential equation (ODE) approximations. The new derivative function surrogate modeling (DFSM) strategy is posited to produce high-accuracy VEM modeling using efficient low-order ODEs with state-dependent parameters. DFSM adjusts parameter mappings adaptively to improve accuracy, which supports efficient low-order computation. After thorough study of DFSM for damped IS design, DFSM will then be used for rapid design exploration and systematic generation of design data. Machine learning strategies will then be used to identify patterns and relationships within this data. An iterative inductive process will be used to identify possible design guidelines from these results, such as preferred distributed shape relationships or sensor/actuator/VEM placement guidelines. Additional numerical design studies will be used to validate/refine design guidelines.