Rights statement: This is the author’s version of a work that was accepted for publication in Composite Structures. Changes resulting from the publishing process, such as peer review, editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. Changes may have been made to this work since it was submitted for publication. A definitive version was subsequently published in Composite Structures, 267, 2021 DOI: 10.1016/j.compstruct.2021.113876
Accepted author manuscript, 1.39 MB, PDF document
Available under license: CC BY-NC-ND
Final published version
Research output: Contribution to Journal/Magazine › Journal article › peer-review
Research output: Contribution to Journal/Magazine › Journal article › peer-review
}
TY - JOUR
T1 - A micromechanics and machine learning coupled approach for failure prediction of unidirectional CFRP composites under triaxial loading
T2 - A preliminary study
AU - Chen, J.
AU - Wan, L.
AU - Ismail, Y.
AU - Ye, J.
AU - Yang, D.
N1 - This is the author’s version of a work that was accepted for publication in Composite Structures. Changes resulting from the publishing process, such as peer review, editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. Changes may have been made to this work since it was submitted for publication. A definitive version was subsequently published in Composite Structures, 267, 2021 DOI: 10.1016/j.compstruct.2021.113876
PY - 2021/7/1
Y1 - 2021/7/1
N2 - This study presents a hybrid method based on artificial neural network (ANN) and micro-mechanics for the failure prediction of IM7/8552 unidirectional (UD) composite lamina under triaxial loading. The ANN is trained offline by numerical data from a high-fidelity micromechanics-based representative volume element (RVE) model using the finite element method (FEM). The RVE adopts identified constituent parameters from inverse analysis and calibrated interface strengths form uniaxial and biaxial tests. A hybrid loading strategy is proposed for the RVE under triaxial loading to obtain the failure points on sliced surfaces whilst maintaining the constant stress at different surfaces. It has been found that the ANN algorithm is robust in the failure prediction of the UD lamina when subjected to different triaxial loading conditions, with over 97.5% accuracy being achieved by the shallow ANN model, where only two hidden layers and 560 samples are used. The predicted 3D failure surface based on trained ANN model has an elliptical paraboloid shape and shows an extremely high strength in biaxial compression. The approach could be used to inform the modification of existing failure criteria and to propose ANN-based failure criteria.
AB - This study presents a hybrid method based on artificial neural network (ANN) and micro-mechanics for the failure prediction of IM7/8552 unidirectional (UD) composite lamina under triaxial loading. The ANN is trained offline by numerical data from a high-fidelity micromechanics-based representative volume element (RVE) model using the finite element method (FEM). The RVE adopts identified constituent parameters from inverse analysis and calibrated interface strengths form uniaxial and biaxial tests. A hybrid loading strategy is proposed for the RVE under triaxial loading to obtain the failure points on sliced surfaces whilst maintaining the constant stress at different surfaces. It has been found that the ANN algorithm is robust in the failure prediction of the UD lamina when subjected to different triaxial loading conditions, with over 97.5% accuracy being achieved by the shallow ANN model, where only two hidden layers and 560 samples are used. The predicted 3D failure surface based on trained ANN model has an elliptical paraboloid shape and shows an extremely high strength in biaxial compression. The approach could be used to inform the modification of existing failure criteria and to propose ANN-based failure criteria.
KW - Failure prediction
KW - Finite element modelling
KW - Machine learning
KW - Representativevolume element
KW - Triaxial loading
KW - UD lamina
KW - Failure (mechanical)
KW - Forecasting
KW - Neural networks
KW - Numerical methods
KW - Stress analysis
KW - Artificial neural network models
KW - Element models
KW - Failure criteria
KW - Failures prediction
KW - Machine-learning
KW - Neural-networks
KW - Representative volume elements
KW - Unidirectional lamina
KW - Finite element method
U2 - 10.1016/j.compstruct.2021.113876
DO - 10.1016/j.compstruct.2021.113876
M3 - Journal article
VL - 267
JO - Composite Structures
JF - Composite Structures
SN - 0263-8223
M1 - 113876
ER -