Designing bioinspired brick-and-mortar composites using machine learning and statistical learning

Seyedreza Morsali,Majid Minary-Jolandan,Dong Qian

doi:10.1038/s43246-020-0012-7

Seyedreza Morsali, Majid Minary-Jolandan + Show 1 more

Open Access

https://doi.org/10.1038/s43246-020-0012-7

Copy DOI

Journal: Communications Materials	Publication Date: Mar 16, 2020
Citations: 17	License type: open-access

Affiliation: The University of Texas at Dallas

Abstract

The brick-and-mortar structure inspired by nature, such as in nacre, is considered one of the most optimal designs for structural composites. Given the large number of design possibilities, extensive computational work is required to guide their manufacturing. Here, we propose a computational framework that combines statistical analysis and machine learning with finite element analysis to establish structure–property design strategies for brick-and-mortar composites. Approximately 20,000 models with different geometrical designs were categorized into good and bad based on their failure modes, with statistical analysis of the results used to find the importance of each feature. Aspect ratio of the bricks and horizontal mortar thickness were identified as the main influencing features. A decision tree machine learning model was then established to draw the boundaries of good design space. This approach might be used for the design of brick-and-mortar composites with improved mechanical properties.

Highlights

The brick-and-mortar structure inspired by nature, such as in nacre, is considered one of the most optimal designs for structural composites
D was randomly partitioned into three non-intersecting sets: training data set (Dtraining) = 70% of D, validation data set (Dv) = 10% of D, and test data set (Dtest) = 20% of D for machine learning (ML) calculation
In order to test the dependency of the results on the size of Dtraining, the ML models were trained with three different sizes of the training data densities, that is, 50%, 60%, and 70%, while the Dv (10%) and Dtest (20%) were kept fixed

Summary

Introduction

The brick-and-mortar structure inspired by nature, such as in nacre, is considered one of the most optimal designs for structural composites. A decision tree machine learning model was established to draw the boundaries of good design space This approach might be used for the design of brick-and-mortar composites with improved mechanical properties. Gu et al used a perceptron and convolutional neural network (machine/deep learning model) to demonstrate their capacity to accurately and efficiently predict the toughness and strength of a two-dimensional (2D) composite[25]. The deep learning approach did not perform very well for some mid-range values of effective stiffness This manuscript presents a ML statistical learning approach for the analysis and design of brick-and-mortar composite architecture with a large data set (>20,000 samples). If we are able to improve the fracture toughness of synthetic composites by 40-folds through bioinspired design principles (e.g., brick-and-mortar microstructure), we will obtain composite materials that have the strength, stiffness, and low weight of ceramics, while their fracture toughness and work-to-failure matches or surpasses those of metals and alloys

Methods

Results

Conclusion