当前位置: X-MOL 学术Integr. Mater. Manuf. Innov. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Uncertainty Quantification for Parameter Estimation and Response Prediction
Integrating Materials and Manufacturing Innovation ( IF 2.4 ) Pub Date : 2019-09-06 , DOI: 10.1007/s40192-019-00154-3
Denielle E. Ricciardi , Oksana A. Chkrebtii , Stephen R. Niezgoda

Integrated Computational Materials Engineering (ICME) is an engineering approach where the materials, manufacturing process, and component designs are optimized concurrently before an actual physical component is realized. This requires the integration of models across vast length and timescales. A key benefit of ICME is the ability to reduce the bulk of expensive and lengthy experiments via tailored simulation. However, ICME introduces new challenges and limitations as the statistical confidence in the final design and manufacturing process must be established from simulation rather than experimental observation and testing. The computational materials science community has not formally adopted tools for verification and validation or UQ for materials simulations. In this study, a Bayesian hierarchical model is considered which accounts for parameter uncertainty, the inherent variability in the properties of material samples tested, and measurement noise. The Bayesian inferential framework is used to calibrate model parameters given calibration data and also to make forward predictions with a confidence level established through the highest posterior density intervals. The generality of the framework is demonstrated through two case studies: (1) parameter estimation for a crystal plasticity model which provides key microstructural and grain-level deformation information to use within the ICME chain; (2) the estimation in uncertainty in thermodynamic phase stability from multiple databases for phase stability.

中文翻译:

参数估计和响应预测的不确定性量化

集成计算材料工程(ICME)是一种工程方法,可在实现实际物理组件之前同时优化材料,制造过程和组件设计。这就需要在广泛的长度和时间范围内集成模型。ICME的主要优势在于能够通过量身定制的仿真来减少大量昂贵且冗长的实验。但是,ICME带来了新的挑战和局限性,因为必须通过仿真而非实验观察和测试来建立最终设计和制造过程的统计可信度。计算材料科学界尚未正式采用用于验证和确认的工具或用于材料模拟的UQ。在这个研究中,考虑贝叶斯分层模型,该模型考虑了参数不确定性,所测试材料样本的特性固有的可变性以及测量噪声。贝叶斯推论框架用于在给定校准数据的情况下校准模型参数,并以通过最高后验密度区间建立的置信度进行前瞻性预测。通过两个案例研究证明了该框架的普遍性:(1)晶体可塑性模型的参数估计,该模型提供了可在ICME链中使用的关键微结构和晶粒级变形信息;(2)根据多个数据库的相稳定性估算热力学相稳定性的不确定性。贝叶斯推论框架用于在给定校准数据的情况下校准模型参数,并以通过最高后验密度区间建立的置信度进行前瞻性预测。通过两个案例研究证明了该框架的普遍性:(1)晶体可塑性模型的参数估计,该模型提供了可在ICME链中使用的关键微结构和晶粒级变形信息;(2)根据多个数据库的相稳定性估算热力学相稳定性的不确定性。贝叶斯推论框架用于在给定校准数据的情况下校准模型参数,并以通过最高后验密度区间建立的置信度进行前瞻性预测。通过两个案例研究证明了该框架的普遍性:(1)晶体可塑性模型的参数估计,该模型提供了可在ICME链中使用的关键微结构和晶粒级变形信息;(2)根据多个数据库的相稳定性,估算热力学相稳定性的不确定性。(1)对晶体可塑性模型的参数估计,该模型提供了在ICME链中使用的关键的微观组织和晶粒级变形信息;(2)根据多个数据库的相稳定性估算热力学相稳定性的不确定性。(1)对晶体可塑性模型的参数估计,该模型提供了在ICME链中使用的关键的微观组织和晶粒级变形信息;(2)根据多个数据库的相稳定性估算热力学相稳定性的不确定性。
更新日期:2019-09-06
down
wechat
bug