Multi-nozzle electrospinning state prediction based on image processing and recurrent neural network

Due to the complex multi-physical field coupling during the multi-nozzle electrospinning process, it is necessary to establish a fast and effective control strategy to maintain a stable multi-jet ejection state. In this study, a multi-jet image processing and jet state prediction system based on OpenCV and PyTorch was introduced to monitor the current electrospinning state in real time and predict the electrospinning state in the next two seconds. The multi-jet ejection process was recorded by a CMOS camera, and the time series images were processed to distinguish different jet states according to the jet area data. Then, the jet area data under different electrospinning jet states were counted as the training data of the neural network, and the prediction accuracy of the best network model after training was greater than 98%. This work has built a significant foundation for the advanced regulation of electrospinning jet ejection and to improve the quality of electrospun membranes.