3 min Research Talk: Hierarchical Material Optimization using Neural Networks

By Miguel Arcilla Cuaycong


Published on


Materials that occur in nature commonly consist of complex architectures arranged in a hierarchy that can be observed at different length scales. However, the design of hierarchical materials is often challenging due to the enormous size of their design space. In this presentation, we sought to use a neural network (NN) to identify optimal arrangements of four different constituents in a tape spring to be used as snapping mechanisms in phase transforming cellular material that can dissipate energy. Training data for the neural network was generated by running finite element simulations in which tapes (whose topologies were generated via a brute force algorithm) were bent under quasi-static conditions, from which the moment-angle plots were obtained and the energy dissipated calculated. With the results from finite element analysis, the top half of the simulated tapes were labeled as “good” topologies with a 1, the rest were labeled as “bad” topologies with a 0. After training, it is expected that the neural network will be able to discern between a good and bad tape spring ligament topology. The weights used in the neural network will be used to study important features in the good tapes versus the bad tapes. Afterwards, the top performing tapes will be fabricated for quasi-static experimental validation using 4-point bending.


Miguel Cuaycong is an undergraduate student studying Mechanical Engineering at Skyline College. During the summer of 2019, Miguel worked as an undergraduate researcher for NCN URE. Miguel helped develop the nanoHUB tool "Hierarchical Material Optimization" which is based on the research called Phase Transforming Cellular Materials(PXCM), under the supervision of Prof. Pablo Zavattieri. During regular semesters, he works as a PI leader (General Tutor) for the SMT division at Skyline College; This position facilitates daily review/tutoring sessions for students enrolled in Calculus and Classical Mechanics among other classes.

Sponsored by

Cite this work

Researchers should cite this work as follows:

  • Miguel Arcilla Cuaycong (2019), "3 min Research Talk: Hierarchical Material Optimization using Neural Networks," https://nanohub.org/resources/31599.

    BibTex | EndNote



Rawls 1062, Purdue University, West Lafayette, IN