Diversified metamaterial reverse topology optimization design method and system

Abstract

The invention belongs to the technical field related to structure optimization, and discloses a diversified metamaterial reverse topological optimization design method and system. The method comprises the steps: carrying out finite element analysis by changing values of a volume fraction, a penalty index, a filtering radius and a filtering mode of a metamaterial structure to obtain multiple groups of topological configurations of the metamaterial structure and a corresponding elastic tensor matrix; selecting characteristic values in the elastic tensor matrix and averaging the characteristic values with the difference value within a preset range as an attribute value; coding the attribute value into single-channel tensor information, and coding the multiple topological configurations into multi-channel tensor information; training a neural network by adopting the attribute value and the corresponding various topological configurations to obtain a trained neural network; and obtaining a to-be-designed attribute value, and inputting the to-be-designed attribute value into the trained neural network to obtain a plurality of corresponding topological configurations. According to the reverse design mode, repeated debugging and verification are not needed, diversified customization requirements can be met, and the design efficiency is greatly improved.

Description

technical field [0001] The invention belongs to the technical field related to structure optimization, and more specifically relates to a method and system for reverse topology optimization design of diverse metamaterials. Background technique [0002] Obtaining a metamaterial structure with specific properties is a hot issue in the current academic and industrial circles. There are two feasible methods for theoretical analysis. One is to change the material and replace the material with poor performance by finding a material with better performance. This method is expensive and economical. The second is to change the design, generally by changing the microstructure unit cell configuration of the porous structure to change the structural performance. This is a relatively cheap way and is widely favored by engineering designers. The structural topology optimization design method can find the most reasonable distribution method for materials in a given design domain, so that t...

AI Technical Summary

Problems solved by technology

The deep learning model it adopts is U-Net network, and the input is 6-channel input tensor. Improper selection of input parameters can easily lead to discontinuous structures. Discontinuous structures are not available in engineering, especially for diverse As far as metamaterials with special requirements are concerned, their attribute values ​​are limited and relatively simple. If U-Net is used, it will lead to overfitting phenomenon in network training, that is, the prediction accuracy of new material structures will be greatly reduced; at the same time, the input of this method is The multi-channel output is a single channel, which cannot meet the needs of diverse topological structures with the same attribute value

Images