Events

The Development of Parallel Level Set Method for Large-Scale Structure Topology Optimization

Speaker :
Mr. Ye TIAN
Department of Mechanical and Aerospace Engineering, HKUST
Date : 30 Sep 2019 (Mon)
Time : 10:30 am - 12:00 pm
Venue : Room 2548, HKUST (2/F., Lift #27/28)

Abstract

Nowadays, topology optimization has developed as a powerful design instrument for engineering process. An increasing number of manufactures begin to employ the topology optimization method to improve the structural performance during the design procedure. At present, CAD software manufacturers such as AutoCAD, Dassault, Altair and others have insert topology optimization algorithms into their commercial software. As one of structural optimization methods, topology optimization methods mainly include homogenization-based approach, the solid isotropic material with penalization (SIMP) approach, the evolutionary structural optimization approach, the level set approach and the newly moving morphable components/voids (MMC/MMV) approach. As topology optimization method is gradually applied into practical engineering projects, the increasing computational complexity has become the biggest obstacle. Therefore, how to effectively improve the computational efficiency of topology optimization is an urgent problem to be solved. The SIMP method implement the parallel computing to improve the computing efficiency. As for the level set method, the introduction of parallel computing also becomes the preferred solution.

In this thesis, a high-performance computing method will be utilized in level-set topology optimization method for generating large-scale structure. Here, the structure based on the compactly supported radial basis functions (CSRBGs) as the example is to exhibit how parallel computing employed in whole computation process, including mesh generation, calculation and assembly of finite element stiffness matrices, sensitivity analysis, solution of the structural state equation, updating and evolution of level set function and out-steaming of final results. Through the optimized structure and time-consuming analysis, there are several valuable discoveries: (1) the speed of optimization computing has extremely increased especially for large-scale structures, (2) the optimized structures of the truss-like components are gradually replaced by thin-sheet-like parts while refining the mesh. These results described above identify that the parallel computing has a significant contribution to level-set topology optimization.

(Supervisor: Prof. Michael Yu WANG)