An open-source parallel topology optimization framework based on unstructured 3D FEA using PETSc and Eigen

This paper presents an open-source framework based on unstructured meshes for three-dimensional large-scale parallel topology optimization using PETSc and Eigen, which is easy to use and expand. The framework supports both eight-node hexahedral and four-node tetrahedral meshes and solves the compliance and stress topology optimization problem. The method of moving asymptotes (MMA) is chosen as the optimization solver. The validity of the framework is demonstrated by a classical cantilever beam problem and two more examples of wheel rim and rotator illustrate the expansibility of the framework.