Session: 12-06-02: Scientific Machine Learning (SciML) for Characterization, Modeling, and Design of Structures and Materials

Paper Number: 148556

148556 - Inverse Design of Non-Linear Mechanics of Nacre-Mimetic Materials Through Interface Engineering and Bayesian Optimization 

Nacre, or mother-of-pearl, is a biologically engineered composite material found lining the in-
ner shells of mollusks. Its intricate and hierarchical architecture features layers of aragonite
platelets arranged in a staggered, brick-and-mortar fashion, intertwined with an organic matrix
of chitin and proteins. Inspired by the brick-and-mortar structure of nacre, researchers have invented various
nacre-mimetic functional materials through different synthesis approaches. For exam-
ple, hierarchically structured fibre was fabricated by wet-spinning assembly technology, us-
ing graphene as tablets and hyperbranched polyglycerol (HPG) as adhesive, which achieved
excellent strength and corrosion-resistance. Nacre-inspired multi-layered films were
fabricated using an immersive layer-by-layer assembly techniques for fire-retardant coating
 and the separation of metallic ions in aqueous solution. Moreover, nacre-inspired
films exhibited high sensitivity and stretching range, qualifying them for wearable electronic
devices. To fully realize their functions, accurate knowledge of the non-linear mechan-
ical behaviors of the nacre-mimetic materials becomes increasingly critical, posing substantial
challenges during the design phase.

Various modeling and optimization techniques have been employed to maximize or balance
the strength and toughness of nacre-mimetic composites, where the design parameters gener-
ally include geometry and material properties of both tablets and interface. Beyond sole maximization, in real practice, a specified non-linear stress-strain relationship for the composite structure may be desirable. For such problems, the inverse design method, starting from desired outcomes to deduce optimal design parameters, proves effective espe-
cially in functional material design. Yet, its application to nacre structures remains scarcely investigated.

Recently, Bayesian optimization (BO), a sequential model-based approach, has been increasingly utilized in addressing inverse problems within material design. The BO framework comprises two principal components: a probabilistic surrogate model describing our hypotheses about the unknown objective function, and an acquisition function suggesting the
most promising point based on the mean value and uncertainty interval of the surrogate model. BO proves advantageous for complex material design inverse problems, owing to its reduced requirement for iterations and initial data points before commencing the optimization process. Typically, in real-world scenarios, designers may lack sufficient knowledge about where optimal solutions are located. Hence, BO with design space expansion feature is preferable for explor-
ing beyond the initial design space, an approach not yet widely adopted in material design problems.

In this study, we aim to bridge the gaps in the inverse design of the non-linear mechanics in composites whose performance is dominated by interface properties, such as nacre-mimetic structures. A BO design framework is proposed, incorporating an expandable design space with a diminishing expansion rate to facilitate exploration beyond initial design space. This
framework is applied to a nacre-mimetic structure, characterized by a five-parameter bi-linear traction-separation law that delineates the interface constitutive relation. The efficacy of the BO framework is validated through three case studies, where the ground truth is established for Case 1, but remains unknown for Cases 2 and 3. By reflecting the outcomes of optimization, the influence of the interface law on the performance of nacre-inspired composite is to be revealed.

 

Presenting Author: Wei Gao Texas A&M University

Presenting Author Biography: Dr. Gao is an Associate Professor of Mechanical Engineering at Texas A&M University. He earned a Bachelor degree in Engineering Mechanics from Sichuan University in 2003, and a Master degree in Solid Mechanics from Tsinghua University in 2006. Upon receiving his second Master in Mechanical Engineering from University of California, Irvine in 2007, Gao worked in industry for two years on the drive train design for electrical vehicles. Furthering his academic pursuit, he received Ph.D. in Engineering Mechanics from University of Texas at Austin in 2014 and conducted his postdoctoral training at Northwestern University. Before joining TAMU, Dr. Gao worked as an Assistant Professor at University of Texas at San Antonio from 2016 to 2022. Dr. Gao received the NSF CAREER award in 2021.

Authors:

Wei Zhang Texas A&M University
Wei Chen Texas A&M University
Wei Gao Texas A&M University