Bioinspired hierarchical composite design using machine learning: simulation, additive manufacturing, and experiment

Abstract

a Biomimicry, adapting and implementing nature's designs provides an adequate first-order solution to achieving superior mechanical properties. However, the design space is too vast even using biomimetic designs as prototypes for optimization. Here, we propose a new approach to design hierarchical materials using machine learning, trained with a database of hundreds of thousands of structures from finite element analysis, together with a self-learning algorithm for discovering high-performing materials where inferior designs are phased out for superior candidates. Results show that our approach can create microstructural patterns that lead to tougher and stronger materials, which are validated through additive manufacturing and testing. We further show that machine learning can be used as an alternative method of coarse-graining -analyzing and designing materials without the use of full microstructural data.This novel paradigm of smart additive manufacturing can aid in the discovery and fabrication of new material designs boasting orders of magnitude increase in computational efficacy over conventional methods.

Effect of Constituent Materials on Composite Performance: Exploring Design Strategies via Machine Learning

“…As demonstrated in our previous work, [9,10] ML techniques can be applied to generate high-performing designs of composites by learning the structure-property relationships from a small amount of training data. Here, we aim to understand whether the optimal designs identified at the continuum-scale can be transferred directly to a smaller length-scale composite system.…”

“…The performance of the CNN model on the highly ranked samples (Figure 4f) is also the best among these ML mod-els since the CNN model has a much higher learning capacity (model complexity) than the other two ML models. As demonstrated in our previous work, [10] where we used a CNN model together with a self-learning www.advancedsciencenews.com www.advtheorysimul.com algorithm to search for high-performing designs of hierarchical composites. As demonstrated in our previous work, [10] where we used a CNN model together with a self-learning www.advancedsciencenews.com www.advtheorysimul.com algorithm to search for high-performing designs of hierarchical composites.…”

“…[7] However, as composite design problems are usually nonconvex, the optimal solutions obtained from those optimization methods often depend on the initial geometry adopted in the optimization process. [9,10] Hanakata et al reported how ML techniques can be applied to the design of kirigami-inspired stretchable materials. On the other hand, machine learning (ML), a data-driven approach, is a promising alternative method to design composites.…”

Controlling the Properties of Additively Manufactured Cellular Structures Using Machine Learning Approaches

“…Accordingly, each layer is provided with data at a separate abstraction level and learns to represent the data in a distinct manner . One case‐study was found to use ML in the design of hierarchical materials of AM parts . The authors investigated optimized properties of hierarchical materials based on a trained model of finite element analysis database of hundreds of thousands of geometries without the use of full microstructural data.…”

“…[24] One case-study was found to use ML in the design of hierarchical materials of AM parts. [25] The authors investigated optimized properties of hierarchical materials based on a trained model of finite element analysis database of hundreds of thousands of geometries without the use of full microstructural data. However, the study did not provide details about the AM technique, used material, or process parameters.…”