CAREER: Using Metamodeling to Enable High-Fidelity Modeling in Risk-based Multi-hazard Structural Design

为了评估美国地震和风振建筑的风险和弹性，高保真计算模型的使用对于表征建筑性能至关重要。然而，在评估最先进的风险/弹性指标时，需要在系统中传播不确定性，这严重阻碍了（如果不是排除的话）这种模型的使用。在基于风险的决策中，这种困难变得更加严重，因为必须对多种建筑设计方案进行评估和比较。本学院早期职业发展计划（Career）奖的研究目标是通过研究一种基于低/中保真度元模型与高保真度结构模型的最佳融合的新模拟范式来克服这一基本限制。通过独立于领域的方法定义元模型，多灾害评估将自然地包含在内，并将使新方法能够快速识别基于多灾害风险的决策问题的最佳权衡解决方案。这些进步将提供模型和程序，以实现完全过渡到基于风险的最佳设计，同时通过严格的优化促进计算资源的合理使用。基于风险的设计将通过提高建筑环境对风和地震事件的安全性，在极端事件中更好地保护生命和财产，并在响应和恢复期间保持基本服务和业务的连续性，从而有利于国家的福利和繁荣。这个职业奖的教育目标是增加工程领域的女性和具有减少风力损失专长的专业人员的数量。这将通过一项高中推广计划来实现，该计划利用基于风险的工程与社会效益之间的联系，激励多样化的学生群体从事工程职业，在密歇根大学发展本科风力工程项目，并为本科生提供研究机会。为了实施高中推广计划，将创建基于项目的学习模块，通过基础科学将基于风险的工程和社会效益联系起来。这些材料的散发将通过教师培训讲习班来实现。该项目的数据将被保存在nsf支持的自然灾害工程研究基础设施（NHERI）数据仓库（https://DesignSafe-ci.org）中。本研究将通过识别仿真环境中每个高保真计算模型的正交子空间，创建一类新的参数元模型（替代模型）。这将提供一种设置，可以通过超约简和机器学习来定义基于物理和数据驱动的降阶参数元模型。高保真度和参数元模型的组合空间将提供一个丰富的仿真环境，其中可以定义多保真度不确定性传播模型，以快速估计高保真度概率风险/弹性指标。元模型的参数化特性将能够创建新的自适应多目标优化方案，这将允许快速识别高保真的多风险帕累托前沿，这是有效的基于风险的决策的核心。通过这项工作确定的模型将直接受益于许多其他学科，包括航空航天和生物医学工程，大气科学和汽车工业，在这些领域，快速高保真计算在科学发现中起着关键作用。这项研究将使用佛罗里达大学的NHERI风洞设备。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。

To assess the risk and resiliency of seismic and wind excited buildings in the United States, the use of high-fidelity computational models is paramount to characterizing the building performance. However, the need to propagate uncertainty through the system when estimating state-of-the-art risk/resiliency metrics significantly hinders, if not precludes, the use of such models. This difficulty becomes exasperated in risk-based decision-making where multiple building design solutions must be evaluated and compared over several hazards. The research goal of this Faculty Early Career Development Program (CAREER) award is to overcome this fundamental limitation through the investigation of a new simulation paradigm based on the optimal fusion of low-/intermediate-fidelity metamodels with high-fidelity structural models. By defining the metamodels through domain independent approaches, multi-hazard assessment will be naturally encompassed and will enable new approaches for rapidly identifying the optimal tradeoff solutions to multi-hazard risk-based decision problems. These advances will provide models and procedures for enabling a full transition to optimal risk-based design, while promoting the rational use of computational resources through rigorous optimization. Risk-based design will benefit national welfare and prosperity through enhancing the safety of the built environment against wind and seismic events to better protect life and property during extreme events and to maintain essential services and business continuities during response and recovery. The educational goals of this CAREER award are to increase the number of women in engineering and professionals with expertise in wind loss mitigation. This will be achieved through a high school outreach program that leverages the link between risk-based engineering and societal benefit to inspire a diverse student pool to pursue careers in engineering, the development of an undergraduate wind engineering program at the University of Michigan, and undergraduate student research opportunities. To implement the high school outreach program, project-based learning modules that connect risk-based engineering and societal benefit through basic science will be created. Dissemination of these materials will be achieved through a teacher training workshop. Data from this project will be archived in the NSF-supported Natural Hazards Engineering Research Infrastructure (NHERI) Data Depot (https://DesignSafe-ci.org). This research will create a new class of parametric metamodels (surrogate models) through identifying orthogonal subspaces for each high-fidelity computational model of the simulation environment. This will provide a setting in which both physics-based and data-driven reduced-order parametric metamodels can be defined through hyper-reduction and machine learning. The combined space of the high-fidelity and parametric metamodels will provide an enriched simulation environment in which multi-fidelity uncertainty propagation models can be defined for rapidly estimating high-fidelity probabilistic risk/resiliency metrics. The parametric nature of the metamodels will enable the creation of new adaptive multi-objective optimization schemes that will allow the rapid identification of high-fidelity multi-hazard Pareto fronts, which are central for effective risk-based decision-making. The models identified through this effort will directly benefit a number of other disciplines, including aerospace and biomedical engineering, atmospheric sciences, and the automotive industry, where rapid high-fidelity computation plays a key role in scientific discovery. The research will use the NHERI wind tunnel facility at the University of Florida.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.