Bayesian optimization of origami multi-cell tubes for energy absorption considering mixed categorical-continuous variables

Multi-cell structures have been widely utilized in energy-absorbing applications for their lightweight and superior mechanical properties. However, they may produce high peak forces that are harmful to occupant safety during a vehicle collision. In addition, the naturally formed folding and global bending modes under axial loads can significantly restrict their energy-absorbing capability. In this study, a novel origami multi-cell structure is proposed to avoid excessive peak loads and at the same time to guide the deformation mode to improve the energy-absorbing capability. Four cross-sectional types of origami multi-cell structures are investigated experimentally and numerically. The results indicated that the web-to-web (W2W) type origami multi-cell structures are most promising for energy absorption, while web-to-corner (W2C) structures yield low peak forces and excellent deformation modes in our study cases. To further exploit the potential of these structures, multi-objective optimization is required to consider the conflicting objectives of maximizing energy-absorbing capability and minimizing the peak force. However, it is challenging for traditional optimization to deal with continuous, integer, and categorical variables simultaneously. Therefore, the Bayesian optimization method based on the Hamming distance is utilized to handle mixed categorical-continuous variables to optimize origami multi-cell tubes. The optimization results indicated that the optimal origami W2C can decrease the peak force by 30% while keeping a similar level of specific energy-absorbing capability compared with the traditional W2C. Moreover, the optimal origami W2C structure can improve the energy absorption by 40% compared with the origami W2C baseline design.