In-situ autocalibrated electrospinning process via U-Net based real-time monitoring

Electrospinning stands out for its ability to produce ultrafine fibers with nanoscale or microscale diameters, though it demands precise adjustment of process parameters to ensure stable fiber production. It is challenging for operators to directly observe the subtle and rapid changes in the jet formed at the needle tip. To address this issue, an autocalibration system was developed that employs artificial intelligence (AI)-based semantic segmentation to analyze process images, enabling the determination of the optimal voltage parameter. The U-Net model, selected for its effectiveness with limited datasets and its simple architecture for rapid processing, segmented images of the initially spun jet into three categories: needle tip, Taylor cone, and ejected jet. An algorithm capable of measuring the Taylor cone volume and jet angle was developed, facilitating the in-situ analysis of jet formation states. Additionally, it was integrated with real-time voltage calibration, thereby adeptly responding to defective jet formations, and adaptively controlling the voltage parameter. Its effectiveness was validated across varied conditions by adjusting the flow rate, solution concentration, the needle-to-collector distance, and the type of polymer. This AI-based method consistently identified optimal voltage settings for each condition, ensuring a uniform electric field and process stability over time.