Page 81 - IJAMD-2-3

International Journal of AI for
Materials and Design Optimization of membrane shrinkage and stability

exhibited a slight increase. In Figure 8D, both %RD and broader applications in smart material design and precision
%TD elevated with increasing voltage, indicating voltage fabrication.
as a key parameter in shrinkage control. These results
demonstrate that meeting shrinkage targets alone is Acknowledgments
insufficient for ensuring process reliability. Incorporating None.
CIW as a stability metric allows selection of process
conditions that are both accurate and robust, providing a Funding
practical basis for optimizing electrospinning parameters This work was funded by the National Natural
in scalable production.
Science Foundation of China (grant no.: 52031005,
The machine learning models in this study were 52571227), Natural Science Foundation of Shanghai
developed and validated using experimental data from (grant no.: 24ZR1438200), Shanghai Academy of
controlled electrospinning trials of TPU membranes, Spaceflight Technology Joint Research Fund (grant no.:
with repeated measurements used to capture both mean USCAST2023-19), Equipment Development Department
shrinkage and variability via the CIW metric. Monte Huiyan Action (grant no.: 5D3D1365), and China
Carlo optimization was conducted on these validated Scholarship Council (grant no.:202406230025).
models within practical parameter ranges, ensuring
physically relevant predictions. The modeling framework Conflict of interest
is material-agnostic and, in principle, applicable to other The authors declare they have no competing interests.
polymers; extending it requires retraining with high-
quality data specific to the target material to capture Author contributions
its unique processing-structure-property relationships.
Direct experimental validation of our multi-objective Conceptualization: Wei Min Huang, Shiyu He
optimization framework was not performed due to the Formal analysis: Shiyu He, Li Cong Huang
high cost and time required. Future work will address Investigation: Shiyu He, Wei Min Huang, Fei Xiao
this to further enhance the model’s robustness and Methodology: Shiyu He, Chentong Gao, Runzhi Lu
generalizability. Writing–original draft: Shiyu He, Wei Min Huang
Writing–review & editing: Shiyu He, Wei Min Huang,
4. Conclusion Fei Xiao
This paper presents a data-driven framework that Ethics approval and consent to participate
combines machine learning with Monte Carlo simulation
to achieve accurate and stable control of biaxial shrinkage Not applicable.
in electrospun membranes, even under limited data Consent for publication
conditions.
Not applicable.
By modeling shrinkage ratios in both RD and TD
along with their stability, the framework enables multi- Availability of data
objective optimization of process parameters. Among the
evaluated models, SVR demonstrated the highest accuracy Data are available from the corresponding author upon
in predicting shrinkage behavior, achieving an RMSE of reasonable request.
1.04% for RD and 2.00% for TD. XGBT was most effective References
in forecasting shrinkage stability, with RMSE of 0.16 for RD
and 0.17 for TD. SHAP analysis identified applied voltage 1. Cheng F, Song D, Li H, Ravi SK, Tan SC. Recent progress in
and TPU concentration as the most influential factors biomedical scaffold fabricated via electrospinning: Design,
governing shrinkage stability, whereas collector speed and fabrication and tissue engineering application. Adv Funct
distance between electrodes had relatively minor effects. Mater. 2025;35(1):2406950.
Using the Monte Carlo-based parameter screening, doi: 10.1002/adfm.202406950
the framework successfully identified sets of process 2. Cho Y, Beak JW, Sagong M, Ahn S, Nam JS, Kim ID.
parameters that met predefined shrinkage targets while Electrospinning and nanofiber technology: Fundamentals,
minimizing variability. innovations, and applications. Adv Mater. 2025;37:2500162.
In conclusion, this methodology offers a robust and doi: 10.1002/adma.202570190
generalizable strategy for high-variability, small-sample 3. Wang C, Su Y, Xie J. Advances in electrospun nanofibers:
manufacturing scenarios, with significant potential for Versatile materials and diverse biomedical applications. Acc

Volume 2 Issue 3 (2025) 75 doi: 10.36922/IJAMD025260022