Artificial neural networks for the accurate prediction of the thickness of fiber membranes prepared by airflow‐assisted electrostatic spinning

Abstract

The practical application of fiber membranes prepared by electrostatic spinning requires a uniform thickness. However, the parameters involved in the spinning process are highly complex, particularly when it comes to precise control over membrane thickness. Hence, achieving a consistent and homogeneous thickness for fiber membranes has posed significant challenges thus far. To accurately predict membrane thickness during the spinning process and ensure consistent outcomes, this study employed polyvinylidene fluoride as the raw material and optimized electrostatic spinning preparation using an innovative airflow‐assisted strategy. The as‐proposed preparation strategy effectively enhances the volatilization of solvents within the jet, also promotes jet stretching through assistance from airflow, resulting in a more stable and homogeneous fiber diameter. Additionally, COMSOL simulations were utilized to provide rational explanations for these findings. By collecting multiple process parameters, including positive voltage, liquid supply speed, winding speed, and solution concentration, this study has successfully developed an artificial neural network (ANN) that exhibits reliable predictive capabilities for membrane thickness. Notably, the model demonstrates a remarkable goodness‐of‐fit with an R2 value of 0.9981. Through the integration of ANN with electrostatic spinning technology, significant advancements have been achieved in accurately predicting and controlling the uniformity of electrostatically spun polymer fiber membranes.