Prediction of polyvinyl alcohol (PVOH) properties synthesized at various conditions by artificial neural networks technique

Abstract

In this research samples of PVOH were synthesized at various reaction conditions (temperature, time, and amount of catalyst). First at 25˚C and 45˚C and constant catalyst weight samples of PVOH were prepared with different degree of hydrolysis at various times. For investigation of the effects of temperature, at times 20 and 40 min and constant weight of catalyst PVOH was prepared at various temperatures. Increasing the time and temperature of the hydrolysis reaction caused increasing degree of hydrolysis and reducing the molecular weight of the samples. Considering the variation of reaction condition, the effects of each parameter on molecular weight, degree of hydrolysis and conversion were investigated individually and also collective. Also, by an artificial neural network method, using experimental results (temperature, time and catalyst amount as input and conversion, degree of hydrolysis and molecular weight as output) a network by Levenberg-Marquardt (LM) back propagation with tan-sigmoid transfer function was established. Finally, the established model presented a good prediction capability and enabled us to predict the output in terms of arbitrary in puts. PVOH is an important polymer and prediction its properties during production significantly improves the quality of the products. Neural network technique is used to model the chemical processes to predict the behavior of the process. In this research we investigated the effects of various processing parameters on the properties of PVOH.