On exploiting machine learning for failure pattern driven strength enhancement of honeycomb lattices

Isanaka, B. R., Mukhopadhyay, T., Varma, R. K. and Kushvaha, V. (2022) On exploiting machine learning for failure pattern driven strength enhancement of honeycomb lattices. Acta Materialia, 239, [118226]. (doi:10.1016/j.actamat.2022.118226).

Abstract

Honeycomb lattices exhibit remarkable structural properties and novel functionalities, such as high specific energy absorption, excellent vibroacoustic properties, and tailorable specific strength and stiffness. A range of modern structural applications demands for maximizing the failure strength and energy absorption capacity simultaneously with the minimum additional weight of material to the structure. To this end, conventional approaches of designing the periodic microstructural geometry have possibly reached to a saturation point. This creates a strong rationale in this field to exploit the recent advances in artificial intelligence and machine learning for further enhancement in the mechanical performance of artificially engineered lattice structures. Here we propose to strengthen the lattice structure locally by identifying the failure pattern through the emerging capabilities of machine learning. We have developed a Gaussian Process Regression (GPR) assisted surrogate modelling algorithm, supported by finite element simulations, for the prediction of failure bands in lattice structures. Subsequently, we strengthen the identified failure bands locally instead of adopting a global strengthing approach to optimize the material utilization and lattice density. A range of sequential local strengthening schemes is explored logically, among which the schemes having localized gradation by increasing the elastoplastic properties and lowering Young's modulus of the intrinsic material lead to an increase up to 37.54% in the failure stress of the lattice structure along with 32.12% increase in energy absorption. The comprehensive numerical results presented in this paper convincingly demonstrate the abilities of machine learning in material microstructure design for enhancing failure strength and energy absorption capacity simultaneously when it is coupled with the physics-based understanding of material and structural behavior.

This record has no associated files available for download.

Contributors

Download statistics