Machine learning-based determination of Mode II translaminar fracture toughness of composite laminates from simple V-notched shear tests

This paper presents a novel method for measuring the translaminar crack resistance curve of composite laminates under Mode II shear loading. A machine learning (ML)-based approach is utilized to extract the inapparent information of the crack resistance curve from the translaminar shear strength measurements obtained from simple V-notched shear tests. The entire campaign is built on the framework of the Finite Fracture Mechanics (FFM) combined with Finite Element Method (FEM). Special emphasis is made on the nonlinear mechanical behavior of composites under shear stress since the original FFM models are designed for quasi-brittle materials. With the well-trained recurrent neural network model, the Mode II R-curve of composite laminate can be obtained with un-notched and V-notched shear strength values as inputs. Experiments were conducted on carbon fiber-reinforced composites to validate the accuracy of the R-curve obtained by the proposed approach and that by the traditional compact shear test. The successful implementation of the method suggests a more convenient and low-cost way of obtaining this important damage-related parameter for composites.