Abstract:The identification of internal structural flaws is an important research topic in structural health monitoring. At present, structural safety inspections based on nondestructive testing procedures mainly focus on qualitative analysis; hence, it is difficult to identify the scale of flaws quantitatively. In this article, an inversion model that can realize quantitative detection is proposed by combing the scaled boundary finite element method (SBFEM) with deep learning. First, the lamb wave propagation processe… Show more
“…Some interesting works have been reported using physical signals, such as sound wave, electromagnetic wave, and temperature, which are adapted to identify structural defect and damage. [18][19][20] Here, because machining-induced plasticity behavior is highly nonlinear and has multifield coupling characteristics, traditional machine learning methods are difficult to accurately describe or predict its complex evolution process. Hence, considering its powerful data-driven and physical constraint ability, PIML strategy is adopted to investigate the machining-induced plasticity behavior for achieving fast and accurate prediction of dislocation behaviors and optimizing machining parameters with good validity and interpretability.…”
The evolution of the dislocation density induced by the nanomachining process dominates the plastic deformation behaviors of materials, thus affecting the mechanical properties significantly. However, a challenging topic related to how to establish an accurate model for predicting the dislocation density based on the limited simulations and experiments arises due to the complicated thermal–mechanical coupling mechanism during the machining process. Herein, a multistage method integrating machine learning, physics, and high‐throughput atomic simulation is proposed to investigate the effect of cutting speed on the dislocation behavior in polycrystal copper. Compared with the traditional one‐step machine learning method, the constraint of physical features effectively improves the accuracy and generalization ability of the model. The results indicate that the dislocation behaviors depend on the competition between the cutting force and temperature. In the low‐cutting speed, the predominated role of the cutting temperature leads to a rapid decline of the dislocation density. In contrast, the dislocation density tends to be stable under a high‐speed cutting process due to the dynamic balance between the effects of the cutting force and temperature. Notably, the proposed strategy provides a new and universal framework to design the machining parameters to obtain high‐quality products.
“…Some interesting works have been reported using physical signals, such as sound wave, electromagnetic wave, and temperature, which are adapted to identify structural defect and damage. [18][19][20] Here, because machining-induced plasticity behavior is highly nonlinear and has multifield coupling characteristics, traditional machine learning methods are difficult to accurately describe or predict its complex evolution process. Hence, considering its powerful data-driven and physical constraint ability, PIML strategy is adopted to investigate the machining-induced plasticity behavior for achieving fast and accurate prediction of dislocation behaviors and optimizing machining parameters with good validity and interpretability.…”
The evolution of the dislocation density induced by the nanomachining process dominates the plastic deformation behaviors of materials, thus affecting the mechanical properties significantly. However, a challenging topic related to how to establish an accurate model for predicting the dislocation density based on the limited simulations and experiments arises due to the complicated thermal–mechanical coupling mechanism during the machining process. Herein, a multistage method integrating machine learning, physics, and high‐throughput atomic simulation is proposed to investigate the effect of cutting speed on the dislocation behavior in polycrystal copper. Compared with the traditional one‐step machine learning method, the constraint of physical features effectively improves the accuracy and generalization ability of the model. The results indicate that the dislocation behaviors depend on the competition between the cutting force and temperature. In the low‐cutting speed, the predominated role of the cutting temperature leads to a rapid decline of the dislocation density. In contrast, the dislocation density tends to be stable under a high‐speed cutting process due to the dynamic balance between the effects of the cutting force and temperature. Notably, the proposed strategy provides a new and universal framework to design the machining parameters to obtain high‐quality products.
“…The scaled boundary finite element method has been developed into a generalpurpose numerical method for the solution of PDE problems [5][6][7][8][9][10][11][12][13][14][15][16][17] over the last few years. This paper aims to present a scaled boundary finite element framework that automates mesh generation and is suitable to high-performance computing.…”
This paper presents the development of the scaled boundary finite element method to benefit from modern technologies for geometrical modelling and high-performance computing. The scaled boundary finite element method allows the use of arbitrarily shaped star-convex polyhedral elements. The greater flexibility in spatial discretization than standard finite elements facilitates automatic mesh generation. A simple and efficient octree algorithm is developed to mesh geometric models given in common formats such as conventional CAD, STL, digital images, and point clouds. By identifying suitable transformations of the octree cells, a mesh can be deconstructed into a limited number of unique cell patterns. A pattern-by-pattern method for computing matrix-vector products in explicit dynamics and iterative solvers is developed. The operations grouping elements of the same pattern reduce the memory requirement and improve the parallel computation efficiency. Numerical examples of large-scale problems with complex geometries are presented. A significant speedup is observed for these examples with up to 1 billion degrees of freedom and running on up to 16,384 computing cores.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.