Machine learning-assisted strategies to enhance the mechanical properties of PVA hydrogels.
Journal:
Journal of the mechanical behavior of biomedical materials
Published Date:
Apr 19, 2025
Abstract
Polyvinyl alcohol (PVA) hydrogels have garnered increasing interest in the field of biomedical materials due to their excellent biocompatibility and controllable mechanical properties. Although various preparation strategies such as freeze-thaw cycles, solvent-exchange, salting-out and annealing treatments have been extensively employed in the preparation of PVA hydrogels, the current literature lacks systematic comparisons under the same PVA molecular weight and mass fraction conditions. It limited the in-depth understanding of the mechanism of optimizing the properties of PVA hydrogels and could not provide guidance for the construction of high strength pure PVA hydrogels. In this study, PVA with a molecular weight of 145,000 was utilized to prepare hydrogels with mass fraction of 10 wt%, 15 wt%, and 20 wt% using the aforementioned four preparation strategies. We thoroughly investigated the effects of preparation strategies and mass fraction on the mechanical properties of PVA hydrogels by employing the eXtreme Gradient Boosting (XGboost) machine learning model for precise data analysis and predictions. Additionally, we also investigated the effects of different preparation strategies and mass fraction on the microstructure and surface properties of PVA hydrogels. The results indicated that the choice of preparation strategies significantly influenced the mechanical properties of PVA hydrogels, surpassing the effects of PVA mass fraction. Notably, under the same preparation conditions, the 20 wt% annealing-PVA hydrogels exhibited the best tensile strength (3.96 ± 0.511 MPa), tensile modulus (4.36 ± 0.160 MPa), and compressive modulus (3.17 ± 0.644 MPa), representing increases of 10 times, 62 times, and 26 times, respectively, compared to freeze-thaw cycles-PVA, while also demonstrating the lowest friction coefficient (0.05). According to the XGboost machine learning model, it showed that the PVA mass fraction had 26.21 % of the effect on the variation of mechanical properties, while the preparation strategy accounted for the remaining 73.79 %. In summary, we successfully established the correlation between the mechanical properties of PVA hydrogels and preparation parameters, providing a solid technical foundation for the development of high strength pure PVA hydrogels.
