An adjustable and controllable a priori design of lattice structure based on deep learning

Abstract

Lattice structures are widely used in aerospace, automotive manufacturing and biomedical engineering because of their excellent properties. Most of the existing design methods for lattice structures are the a posteriori design. Based on the basic structure of rods, beams, plates, shells, etc., the lattice cell is constructed first, and then the performance of the lattice materials is verified. It is difficult to realize the suitable design for the performance requirements of the lattice materials. To solve this problem, the adjustable and controllable a priori design method of lattice structures is proposed. By taking the expected performance goal of the lattice structure as the design starting point, lattice structures with different dominant performance are topologically associated to construct the design space, so that the adjustability of lattice structure performance can be ensured. A layered and progressive strategy is proposed to optimize the design space and ensure the controllability of lattice structure performance. Moreover, an Artificial Neural Network (ANN) is introduced to learn the law between the configurations and the corresponding mechanical performance. Then, through the optimization algorithm, the lattice structure is obtained oriented to the requirement of expected performance index. Through the above process, the adjustable and controllable a priori design of lattice structure is constructed. In addition, the example is given to verify the correctness and effectiveness of the adjustable and controllable a priori design method of lattice structures.