Sustainable Development Goals
Abstract/Objectives
This study explores the shortcomings of existing simulation technologies from the perspective of extreme engineering problems; develops new numerical methods and meshless methods to improve the accuracy and stability of extreme engineering simulations; establishes reliable alternative physical models to improve physics by introducing machine learning and data-driven mechanics Insufficient problems, improve the overall extreme engineering simulation technology; the technical framework developed in this research can be applied to the fields of manufacturing and processing, national defense engineering, man-made and natural disaster prevention, mechanical and civil structure manufacturing design, etc.
Results/Contributions

• This study proposes a variational multi-scale meshless method to solve the volumetric locking problem. Through the hybrid formula and the developed multiscale steady-state term, the volumetric locking and pressure oscillation problems in the large deformation problem can be effectively avoided. We extended this method to the material point method, and the stress oscillation is eliminated through the pressure projection technique, and the Cross-Cell Instability is solved by the RKPM method

• This study used the neural network to overcome the difficulty of simulating weak discontinuities and stress concentration problems with the meshless method in the advection dominated problem in the fluid flows; we also use the neural network coupling finite element method to overcome another common phenomenon in multi-scale extreme engineering problems: the curse of dimensionality.

We have published several top-ranked journal papers in this field.

Keywords
Extreme EventsMeshfree MethodsVariational Multiscale FrameworkMachine LearningData Drive MechanicsThermal-Mechanical Analysis
Contact Information
黃琮暉教授
thhuang@mx.nthu.edu.tw