Prediction of Delamination Area of Laminated Composite Under Low Velocity Impact Based on Experimentally Validated Finite Element Modeling and Machine Learning Methods

Abstract

Laminated composite materials are usually vulnerable to impact loading. Low velocity impact (LVI) events can cause extensive internal damage including fiber breakage, matrix splitting and delamination. The internal damage is detrimental to the post-impact response of the composite material. Predicting internal damage due to LVI numerically with sufficient accuracy and detail tends to take several days on computer clusters with tens of CPUs. This paper aims at establishing a predictive model for the macroscopic extent of delamination of a composite subjected to LVI based on an FE model and an artificial neural network (ANN), which is a machine arameters. The input for the ANN model is the stacking sequence of the laminate and the output is the total area of delaminations in the composite. First, an experimentally validated numerical model for LVI based on the finite element method (FEM) is established. Then, a database is built using the FEM model for various stacking sequences. Finally, an ANN model is trained based on the database and a test set is predicted with the ANN and compared against the results obtained with the FEM model. With the help of the Experiment-FEM-ANN technology, predictions of LVI damage, and a parametric study to the full extent, and optimization of the design of a laminated composite against LVI, are all made possible.