Fracture prediction of hydrogel using machine learning and inhomogeneous multiscale network

Abstract

Hydrogels are soft polymeric materials with promising applications in biomedical fields. Understanding their fracture behavior is crucial for optimizing device design and performance. However, predicting hydrogel fracture is challenging due to the complex interplay between material properties and environmental factors. In this study, a machine learning (ML) approach to predict hydrogel fracture behavior is presented. A multiscale hydrogel fracture model is developed to generate simulation data, which is used to train a predictive neural network model. The ML model utilizes a hierarchical architecture of convolution long short-term memory units to capture spatial and temporal dependencies in the data. Model predictions are found to closely match simulation results with high accuracy, demonstrating the ability to learn complex fracture processes. Comparison of crack lengths shows the model can generalize across different material parameters. This work highlights the potential of ML for advancing the understanding of hydrogel fracture and soft matter failure. The presented approach provides an efficient framework for predicting fracture in complex materials and systems.