Preparation and optimization of chitosan/polyethylene oxide nanofiber diameter using artificial neural networks

Chitosan/polyethylene oxide (PEO) solution makes electrospun nanofibers with decreased beads and diameters in comparison with lonely chitosan (CS). The aim of this work was to find an artificial neural network (ANN) model for predicting the chitosan/PEO blend electrospun nanofiber diameter. Chitosan/PEO concentration ratio, distance between nozzle tip and collector, applied voltage, and flow rate were considered as input variables, and chitosan/PEO blend electrospun nanofiber diameter was considered as output variable. Scanning electron microscopy images indicated that electrospun nanofiber diameter was approximately 50-185 nm. For increasing validity, k-fold cross validation method was applied to dataset. The ANN technique was used for training and testing via fivefold of dataset. The best results of prediction were obtained via network with three hidden layers including 10, 15, and 5 nodes in each layer, respectively. The mean square error (MSE) and correlation coefficient between the observed and predicted thickness of the nanofibers in the chosen model were about 0.0707 and 0.9630, respectively, indicating the ANN technique validity in the prediction procedure. For the analysis of interactions between the involved electrospinning parameters and nanofiber diameter, 3D graphs in various levels were plotted. In conclusion, the results indicated that using the prediction process via ANN could be relevant in the decision to produce nanofibers with desired shape and diameter via electrospinning.