ARTIFICIAL NEURAL NETWORK MODELING OF ELECTROSPINNING PROCESS

It has been found that morphology of electrospun fibers, such as diameter, is dependant on processingparameters. In this investigation, Artificial Neural Networks have been used to model and optimize theelectrospinning parameters for Bombyx mori silk and polyacrylonitrile (PAN) nanofibers. In order to do that,two sets of data were extracted from the literature available. Each was divided into two groups: Train dataand test. Data considered as test were not used for training of the networks whereas others as train data wereemployed. Concentration of polymer and applied voltage in one and concentration of polymer, appliedvoltage and tip to collector distance in the other were inputs and the output was average electrospun fiberdiameter. Number of hidden layers together with size of neurons in each of them was selected to provide thebest results for the system. Root Mean Square Errors (RMSE) were calculated for the evaluation of themodels. ANN model was compared with conventional Regression method. The amounts of errors of ANNwere significantly lower than those of Regression. The results show that the use of ANN in order to predictthe average electrospun fiber diameter has been successful.

Keywords:

Nanofibers , Electrospinning , Artificial Neural Networks , Multiple Regression