A Machine Learning Technique to Predict Biaxial Failure Envelope of Unidirectional Composite Lamina

Abstract

A machine learning technique was used to predict static, failure envelopes of unidirectional composite laminas under combined normal (longitudinal or transverse) and shear loading at different biaxial ratios. An artificial neural network was chosen for this purpose due to their superior computational efficiency and ability to handle nonlinear relationships between inputs and outputs. Training and test data for the neural network were taken from the experimental composite failure data for glass-and carbonfiber reinforced epoxies provided by the worldwide failure exercise (WWFE) program. A quadratic, stress interactive Tsai-Wu failure theory was calibrated based on the reported strength values, as well as optimized from the experimental failure data points. The prediction made by the neural network was compared against the Tsai-Wu failure criterion predictions and it was observed that the trained neural network provides a better representation of the experimental data.

The intersection of damage evaluation of fiber-reinforced composite materials with machine learning: A review

“…Bhuiyan et al. 199 confirmed the results obtained in previous studies 196,197 concerning the performance of the Tsai-Wu, optimized Tsai-Wu, and ANN failure envelopes using biaxial loading experiments of CFRP plates. Experimental biaxial tests using transverse and shear loading were conducted with CFRP laminates to compare against predicted values and obtain the optimized Tsai-Wu criterion coefficients.…”

“…Jones 49 presented a model for a transversely isotropic material with a plane stress condition:where F i and F ij are defined by the monotonic strengths aswhere X T and X C are the longitudinal tensile and compressive strengths in the fiber direction, Y T and Y C are the transverse tensile and compressive strengths, and S is the in-plane shear strength. Two forms of the Tsai-Wu criterion were utilized in previous work 196–199,206 : the form presented in equation (35) and an optimized Tsai-Wu criterion. Rather than using the monotonic strength values of a material for the strength tensors given in equation (36), the optimized Tsai-Wu criterion employed an optimization process on the values of the strength tensor coefficients to minimize the error between experimental data and the analytical failure surface.…”

“…Figure27. The ANN predicted failure envelope minimizes the error between the actual (experimental) values and its predicted values, thus producing an unsymmetric, non-uniform failure envelope 199. …”

A machine learning based prediction of elasto-plastic response of a short fiber reinforced polymer (SFRP) composite

“…The database used for training is taken from the experimental data reported in worldwide failure exercise (WWFE) [15]. A similar literature on prediction of failure envelope of UD composites using the ANN has been reported by Bhuiyan et al [16], Mukherjee et al [17], and Lee et al [18]. Naderpour et al [19] have reported the determination of the compressive strength of an FRP-confined concrete specimen using ANN.…”

Iterative experimental design and identifiability analysis of composite material failure models

“…Machine learning has already been used in material model identification. The Tsai-Wu model was employed to simulate unidirectional composite lamina, which was optimized by a neural network [13] . Soft computing has also been used in material fatigue estimation [14] .…”