Stiffness modification in compliant joints with the use of mechanical metamaterials and the aid of machine learning

Abstract

Compliant joints (CJs) corresponds of a type of mechanisms which are designed with differ ent types of flexure hinges (F Hs), causing a notorious variation in motion ranges. These F Hreacts towards external forces giving them certain movement limited by the material or design of them. These factors can be represented as the stiffness that they have. With the usage of certain techniques this stiffness can be improved. In this research, we propose the use of spe cific 2D lattice metamaterials with different unit cell geometries and orientations to change the resultant stiffness. The 2D lattices used were the square honeycomb lattice, the re-entrant honeycomb lattice and the hexagonal honeycomb lattice. For the mechanical tests, some of the lattices with a specific unit cell orientation but similar relative densities were evaluated. In addition the use of artificial intelligence (AI), specifically the machine learning (ML) field which helped us to predict desired mechanical parameters of the CJs designed. Various ML algorithms were tested and compared with the finite element analysis (F EA) simulations of the CJs, to evaluate the prediction accuracy between learning algorithms. Finally, with the predictions gathered of a small and a larger dataset based only in simulations, the development of an automated design process based on the use of latticed CJs was achieved.