Prediction of Delamination Size in Composite Material Using Machine Learning

Composite materials are famously known for providing safety and endurance to required components in the aerospace industry. Delamination of such materials is caused by rotating loads and decaying job settings which eventually pose a threat to the health of the composite layer structure and hence compromises safety of human life. Data was collected using an acoustic extraction method to assess the health of the composite material by disclosing the region of the damage, type, and size on the sample material. This paper proposes a predictive method using machine learning to measure the region of delamination over the composite sheet by comparing different machine learning models. By critically assessing and comparing various machine learning methods, an optimum solution through classification and regression models is suggested. The region of the delamination may be estimated by the length of the predicted route based on the observed conditions, changes, routes and the image of the structural health of the composite material. This approach will prevent the data overfitting of the model training and thus avoid the problem of overuse in the training process. The outcomes of the experimented load cycles on the composite sample coupons show that the suggested regression model surpasses other similar approaches in predicting accurately and efficiently.