Efficient forward uncertainty propagation strategies in complex systems

由于计算能力的最新进展，计算机辅助工程(CAE)工具今天可以用于支持大规模工程设计。CAE工具可以通过构建虚拟原型来快速探索设计的性能，而不是构建多个昂贵的物理原型，并评估无法通过实验重现的操作条件。然而，这些虚拟原型产生的性能结果可能与通过测试实际结构获得的结果显著不同。这可能是由于制造变异性或缺乏知识而导致的模型参数不确定(例如几何图形、材料特性和联接)造成的。因此，近年来，人们致力于开发能够考虑不确定性的分析方法，以开发对参数不确定性相对不敏感的稳健设计。然而，通过大规模模型(通常包括大量不确定参数)开发高效和准确的正向不确定性传播策略仍然是该研究领域的主要挑战之一，特别是在使用混合的概率和非概率不确定性描述对不确定参数进行建模的情况下。这项海外旅行赠款申请旨在资助对美国阿尔伯克基桑迪亚国家实验室优化和不确定性量化部门的短期访问(2017年7月5日至2017年7月14日，10天-不包括旅行时间)。这笔小额的旅行赠款将使Cicirello博士和优化和不确定性量化团队之间能够密切合作，深入研究这个问题，并促进和加快未来的合作项目。

Thanks to recent advances in computing capabilities, Computer-Aided-Engineering (CAE) tools can be used today to support large-scale engineering design. CAE tools allow to rapidly explore the performance of a design by building virtual prototypes, rather than building multiple expensive physical prototypes, and to assess operating conditions not reproducible with experiments.However, the performance results yielded by these virtual prototypes might be significantly different from those obtained by testing the actual structure. This can be caused by uncertainties model parameters (such as geometry, material properties, and joints) arising from manufacturing variability or lack of knowledge. As a result, in recent years extensive research has been devoted to the development of analysis methods which can take into account uncertainties to develop robust designs which are relatively insensitive to parameter uncertainties. Nonetheless, developing efficient and accurate forward uncertainty propagation strategies through large scale models (which usually include a large number of uncertain parameters) is still one of the major challenges in this research area, especially when uncertain parameters are modelled using mixed probabilistic and non-probabilistic uncertainty descriptions.This overseas travel grant application seeks to fund a short visit (10 days, 5 July 2017 to 14 July 2017 - excluding travel time) to the Optimization and Uncertainty Quantification department of Sandia National Laboratories in Albuquerque (US). This small travel grant will enable close collaboration between Dr Cicirello and the Optimization and Uncertainty Quantification team to work intensively on the problem, and to facilitate and accelerate future collaborative projects.