102 - STRENGTH PREDICTION OF NOTCHED COMPOSITES WITH MACHINE LEARNING (CompTest 2025)
In recent years, there has been a growing interest in using machine learning (ML) to study progressive damage and predict strength of notched composite laminates [1] In the literature, training data for ML is often generated from simplified progressive damage models by excluding anisotropy and certain failure mechanisms such as delamination in order to increase efficiency of data generation. However, this may result in inaccurate predictions, especially in cases where delamination is not insignificant [2]. The objective of this research is to use ML to predict open-hole tension (OHT) strength and failure strain of IM7-8552 quasi-isotropic carbon-epoxy composites after training with data of laminates with various laminating sequences. A large set of training data for the ML model was generated using a recently developed explicit finite element (FE) model capable of modeling not only matrix cracking and fiber failure but also delamination [3]. It employs a discrete crack model (DCM) technique, known as the floating node method (FNM) [4, 5], to simulate progressive damage patterns and obtain the strength of an OHT specimen. The model has been validated with limited experimental data [3]. In contrast to implicit FE, the explicit FE model is more efficient (i.e. significantly shorter CPU times) and does not encounter convergence issues while maintaining accuracy.
Pieteikuma datums
21.05.2025.
Atslēgas vārdi
Notched composites, Strength, Machine Learning, Progressive Damage, Modeling
