An efficient data-driven optimization framework for designing graded cellular structures

Cellular structures are increasingly applied in engineering practice since they have excellent mechanical and multifunctional properties, such as light weight, energy absorption and noise reduction. In this paper, an efficient data-driven M-VCUT (multiple variable cutting) optimization framework is developed for designing the graded cellular structures. Three main parts are included in this method, i.e., the establishment of problem-independent offline database, the coarse-scale optimization, and the geometry reconstruction of full-scale cellular structure. The advantage of this method is that: (1) it has high computational efficiency due to the establishment of a problem-independent database in an offline manner; (2) it directly adopts numerical derivation, which makes the calculation of sensitivity more convenient; (3) it has fewer design variables. Based on the proposed method, the designer can predefine a variety of different basic microstructure configurations, and then combine them into different microstructures according to different requirements, realizing the integrated design of microstructure and macrostructure. Numerical examples are carried out for verifying the validation and efficiency of the proposed method.