期刊
AIAA JOURNAL
卷 -, 期 -, 页码 -出版社
AMER INST AERONAUTICS ASTRONAUTICS
DOI: 10.2514/1.J062959
关键词
Digital Engineering; Reduced Order Modelling; Linear Elastic Fracture Mechanics; Rotorcrafts; Structural Integrity; Structural Analysis; Monte Carlo Simulation; Digital Twins; Crack Growth; Machine Learning
This study introduces a reduced-order modeling (ROM) method for predicting nonplanar crack growth in structural digital twins. The method represents the entire crack surface morphology using a B-spline surface, which captures its impact on crack growth. The proposed method demonstrates superior fracture mechanics parameter prediction and is three orders of magnitude more efficient than full-order simulation.
Nonplanar crack growth holds a critical role in aeronautical structures, necessitating effective analysis under mixed fatigue loading to assess structural integrity. This study introduces a reduced-order modeling (ROM) method for predicting nonplanar crack growth in structural digital twins. The method's advantage lies in its representation of the entire crack surface morphology using a B-spline surface, which better captures its impact on crack growth. The symmetric Galerkin boundary element method-finite element method coupling method is adopted as a full-order method to generate the crack database. Isoparametric coordinates of the crack surface and stress intensity factor serve as input and output, respectively, for training the ROM, which integrates K-means clustering, principal component analysis, and Gaussian process regression. The proposed approach is demonstrated using a rotorcraft-mast-like component. Results reveal superior fracture mechanics parameter prediction compared to the crack-front-based ROM. Furthermore, the method boasts three orders of magnitude greater efficiency than full-order simulation, enabling its coupling with approaches like Monte Carlo for probabilistic crack growth analysis. Future work entails integrating our method into the probabilistic framework of digital twins.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据