Predictions for hydrogel fracture experiment using data processing

• Main Author: Siak, Yen Kar
• Other Authors: Li Hua
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering
• Online Access: https://hdl.handle.net/10356/176519
• Institution: Nanyang Technological University

Hydrogels, intricate three-dimensional networks of hydrophilic polymers, are vital in biomedical engineering and regenerative medicine. Predicting their fracture behaviour remains challenging due to complex viscoelastic properties. While molecular modelling methods like High-Resolution Transmission Electron Microscopy (HRTEM) are conventional, they are computationally intensive. This study explores Convolutional Long Short-term Memory (Conv-LSTM) deep learning techniques for accurate hydrogel fracture prediction. Conv-LSTM, adept at handling spatiotemporal data, shows promise in modelling fracture mechanisms across materials. In this report, the research develops and optimizes a Conv-LSTM-based predictive model for enhanced hydrogel fracture prediction accuracy. Methodology involves data collection from tensile testing, preprocessing, model selection, training, and validation. Evaluation metrics include Mean Squared Error (MSE) and Structural Similarity Index (SSIM). Results demonstrate Conv-LSTM's high precision in capturing hydrogel fracture patterns, offering insights into fracture dynamics. Computational efficiency is assessed through GPU and CPU training comparisons. Future research will focus on expanding the dataset, exploring multi-scale prediction approaches, estimating prediction uncertainty, adapting the model to related domains, and optimizing for real-time prediction. Addressing these areas will advance predictive modelling for hydrogel fracture, benefiting biomedical engineering and materials science.