Prediction of functional longevity and mechanical stability of bio-printed scaffolds using Physics-Informed Neural Networks (PINNs)

Prediction of functional longevity and mechanical stability of bio-printed scaffolds using Physics-Informed Neural Networks (PINNs)

FATEMEH SADAT MIRBAGHERI FIROUZABAD1

1) Iran University of science and Technology (IUST), Tehran, Iran

IOI Code : ABEB-ACZGCK
Publication : 4th International Conference on Recent Advances in Engineering, Innovation and Technology - Belgium(eitconf.com)
Abstract :
    Predicting the functional longevity and mechanical stability of bio-printed scaffolds is crucial for success in tissue engineering applications. Traditional data-driven modeling approaches, although powerful, often act as black boxes that are unable to satisfy fundamental physical laws such as mass conservation and mechanical equilibrium, which leads to unreliable extrapolations. To address this limitation, this study proposes a Physics-Informed Neural Network (PINN) framework for modeling the coupled degradation and mechanical behavior of bio-printed scaffolds. By encoding the differential equations governing degradation kinetics and mechanical stress equilibrium directly into the network’s loss function, this model ensures physical consistency throughout the process. The proposed PINN model was validated against synthetic datasets and demonstrated superior accuracy with R2 > 0.95 and higher robustness compared to traditional regression models. The results indicate that incorporating physics-based constraints effectively enables the model to predict long-term mechanical degradation profiles with high accuracy. This framework provides a reliable computational tool for optimizing scaffold design and facilitates advancements in assessing structural integrity and the development of sustainable bioprinting within the context of the circular economy.

Keywords : physics informed neural networks (pinns) bio printed scaffolds scaffold degradation mechanical stability tissue engineering computational modeling