CAREER: Predictive Analysis of Stability-Critical Structures: an Uncertainty-Informed Path from Measurements to Theory

这个教师早期职业发展（Career）项目奖的基本主题是开发一个统一的预测分析框架，该框架将显著改善稳定性关键结构的基于分析的设计状态。该框架通过将数据科学中的数据挖掘、信息理论和统计推断技术以及基于嵌入式非线性预测器的高保真随机非线性解算器与计算科学中的增量迭代路径跟踪技术结合在一起，跨越了学科界限。薄壁结构构件，如冷弯型钢构件和薄壁圆柱形结构，对材料和制造缺陷极为敏感，因为它们在受到压应力时以屈曲模式失效。屈曲/稳定破坏模式使其难以预测坍塌荷载。为组件设计预测分析框架的挑战在于，稍微偏离完美就会极大地影响它们对负载的响应。结果是结果的大分散，特别是当对崩溃荷载的响应感兴趣时。适用于全球和局部尺度以及随机和确定性方法的可观测和不可观测不确定性的现实输入模型将是所提议框架的一个独特特征。目标是开发包括所有变量的分析非线性计算模型，将研究整合到课程中，并为社区大学和高中学生提供扩展服务。计算模型将用已有的实验数据进行验证。在这个项目中开发的方法将推动稳定关键结构的计算建模技术的发展，并将创造一个飞跃，以缓和设计这些结构时涉及的大安全因素。博士课程将开设一门新的结构稳定性的概率方法课程。

The underlying theme of this Faculty Early Career Development (CAREER) Program award is the development of a unified predictive analysis framework that will significantly improves the state of analysis-based design for stability-critical structures. The framework crosses disciplinary boundaries by bringing together data mining, information theory and statistical inference techniques from data sciences and high fidelity stochastic nonlinear solvers that are based on embedded nonlinear predictors into incremental-iterative path following techniques from computational sciences. Thin-wall structural components such as cold-formed steel members and thin wall cylindrical structures are extremely sensitive to material and fabrication imperfections because they fail in buckling mode when subjected to compressive stresses. The buckling/stability failure mode makes it difficult to predict collapse loads. The challenge in devising predictive analysis framework for components is that a slight deviation from perfection dramatically affects their response to loads. The result is a large scatter in results especially when the response up to collapse load is of interest. Realistic input models for observable and unobservable uncertainties that are adaptable to both global and local scales as well as to stochastic and deterministic methods will be a unique feature of the proposed framework. The goal is to develop analytical non-linear computational model that includes all variables, to integrate research in to curriculum and to provide outreach to community college and high school students. The computational model will be validated with experimental data that is already available.The methodologies developed in this project will advance the state of the art in computational modeling of stability-critical structures and will create a leap towards moderating large safety factors involved in designing these structures. A new course in probabilistic methods in structural stability will be developed for the PhD curriculum.