Computational methods for modeling and design of complex engineering systems under uncertainty

基于计算的设计开始成为许多工程领域的规范，高保真模型现在越来越多地用于支持工业设计实践中的决策制定。尽管在这一领域已经取得了重大进展，但为了充分实现基于计算的设计工作流的潜力，加速复杂系统的设计，仍有许多挑战有待克服。其中一个主要的挑战是计算成本。随着对高保真模型的不断推动，即使在大规模并行计算机上，对各种设计方案/概念的评估在计算上也是不可行的。另一个挑战来自这样一个事实，即不确定性在工程系统的数学建模和表征中无处不在。当处理不确定性时，现有的方法可能在计算上令人难以接受，例如，当工程师希望使用高保真模型预测性能统计数据和/或决定可以帮助减轻不确定性影响的参数设置时。现实世界中复杂工程系统的计算机模型通常是根据大量的设计变量和不确定参数进行参数化的，这一事实加剧了这两种挑战。在计算建模和设计优化中的高维问题，特别是在存在不确定性的情况下，迫切需要可扩展和高效的算法。

Computation-based design is starting to become the norm in many engineering sectors and high-fidelity models are now increasingly used to support decision making in industrial design practice. Despite the significant progress that has been made in this area, a number of challenges remain to be overcome in order to fully realize the potential of computation-based design workflows to accelerate complex systems design. One of the main challenges is computational cost. With the increasing push towards high-fidelity models, evaluation of a wide range of design alternatives/concepts can be computationally infeasible even on massively parallel computers. Another challenge arises from the fact that uncertainty is ubiquitous in the mathematical modeling and characterization of engineering systems. Existing approaches can be computationally prohibitive when dealing with uncertainty, e.g., when engineers wish to predict performance statistics using a high-fidelity model and/or decide on parameter settings that can help mitigate the impact of uncertainty. These two challenges are exacerbated by the fact that computer models of complex real-world engineering systems are often parametrized in terms of a huge number of design variables and uncertain parameters. There is a pressing need for scalable and efficient algorithms to tackle high-dimensional problems in computational modeling and design optimization, especially in the presence of uncertainty.