Separation and pre-concentration of trace quantities of iron ions using molecularly imprinted polymer and its application of artificial neural networks for predicting the extraction yield

In this study, a new modeling method based on three-layer artificial neural network (ANN) techniques has been employed to predict the extraction yield of iron from real samples by means of molecularly imprinted polymer. Input variables of the model were pH, absorption and desorption time, ligand amount and volume of solution while the output was extraction yield of iron ions. The mean squared error and correlation coefficient between the experimental data and the ANN predictions were determined as 0.0036 and 0.96428 for training, 0.0020 and 0.96232 for validation and 0.0004 and 0.9962 for testing data sets. The detection limit of the proposed method was 3.1 &mug. L-1. Dynamic linear range in the range of 200-1000 &mug. L-1 was obtained. The relative standard deviation was found to be below 8.8%. The method was applied to the recovery and determination of Fe in a few different real samples.