Abstract
Abstract A simple neural network with a three-node hidden layer has been used to identify and predict protein adsorption, desorption and fractionation profiles in a 25 × 1 cm ID ProductivTM CM ion-exchange column. To predict the effect of flow rate on the adsorption breakthrough curve, two sets of data obtained at the maximum and minimum of the full range of flow rates used were sufficient to train the neural network which was then able to predict the effects of flow rate changes within the training range on the adsorption breakthrough curve. This training method was also applied to explore the effects of flow rate on desorption and fractionation. It was found that the network training algorithm performed satisfactorily if the flow rate data for desorption and fractionation were scaled in the form of logarithm.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call