Bibliographic Details
| Title: |
Machine learning prediction of mechanical properties of braided-textile reinforced tubular structures |
| Authors: |
Wenhao Wang, He Wang, Jiannan Zhou, Hualin Fan, Xiao Liu |
| Source: |
Materials & Design, Vol 212, Iss , Pp 110181- (2021) |
| Publisher Information: |
Elsevier, 2021. |
| Publication Year: |
2021 |
| Collection: |
LCC:Materials of engineering and construction. Mechanics of materials |
| Subject Terms: |
Braided-textile reinforced tubular structure, Mechanical property, Energy absorption, Machine learning, Materials of engineering and construction. Mechanics of materials, TA401-492 |
| More Details: |
Braided-textile composites have been applied in various engineering fields and used for high-efficiency energy absorbers to meet different loading requirements. However, it’s difficult to accurately forecast the mechanical properties of these braided-textile composites based on theoretical method, which is a big obstacle to the precise design of the energy-absorbing structure. In this research, carbon fiber reinforced composite (CFRC) tubular structures were fabricated by two-dimensional braided textiles, and 160 groups of orthogonal axial compression experiments were carried out.Axial compression behaviors of CFRC tubular structures, including peak force, mean crushing force, energy absorption, specific energy absorption, elastic modulus and other relevant parameters, were tested and collected in a database. The feed forward back propagation (FFBP) algorithm based on artificial neural network (ANN), as an algorithm with high convergence accuracy in machine learning (ML), was adapted to build a prediction model to forecast the overall axial compression properties of those CFRC tubular structures uncollected in the database. By a series of error analyses, the ML method was proven to have high accuracy in predicting the relevant mechanical properties within the range of orthogonal design groups. Although the relative error of marginal sample data (especially single-layer tubes) is large, the mean absolute error (MAE) of training set and test set is only about 5% and 10%, respectively. Our work demonstrates the potential of machine learning methods in predicting the overall mechanical properties and guiding the design of composite materials. |
| Document Type: |
article |
| File Description: |
electronic resource |
| Language: |
English |
| ISSN: |
0264-1275 |
| Relation: |
http://www.sciencedirect.com/science/article/pii/S026412752100736X; https://doaj.org/toc/0264-1275 |
| DOI: |
10.1016/j.matdes.2021.110181 |
| Access URL: |
https://doaj.org/article/e690032d001546519473d062fcaa0091 |
| Accession Number: |
edsdoj.690032d001546519473d062fcaa0091 |
| Database: |
Directory of Open Access Journals |
