Explainable Machine Learning Framework for Optimizing Electrospinning Parameters in ZnO-PVP Nanofiber Synthesis

The electrospinning technique is crucial for synthesizing zinc oxide (ZnO) nanofibers within a polyvinylpyrrolidone (PVP) polymer matrix, where precise control over fiber diameter is essential for optimizing functional properties. This study investigates the application of Machine Learning (ML) methods, including Multiple Linear Regression (MLR), Support Vector Regression (SVR), and XGBoost, to model and optimize key electrospinning parameters: PVP concentration, flow rate, needle tip-to-collector distance, and applied voltage, aiming to produce uniform nanofibers with minimal diameters. A systematic experimental dataset was generated to capture fiber diameter variations under different parameter settings. Among the models evaluated, SVR demonstrated the best predictive performance, achieving an R² score of 0.97 with accuracy of 0.96, indicating its effectiveness in capturing the complex nonlinear relationships inherent to the electrospinning process. Feature importance analysis using F-scores, SHAP values, and permutation importance consistently identified PVP concentration as the most influential factor, followed by flow rate and voltage, while distance showed minimal impact. The proposed ML-based predictive modeling framework offers a data-driven approach to fine-tune electrospinning conditions, enabling greater precision and efficiency in the fabrication of ZnO-PVP nanofibers.