Dehydration Modeling of Polymeric Membranes by Pervaporation Process

Abstract

In this study, a study was performed on methanol-water pervaporation separation. Here is a mixture of ethanol-water as well as isopropanol-water using several different types of polyurethane-based polyethylene glycol (PEG) membranes. The results of this study included three types of polyurethane membranes with different molecular weights were evaluated. The results showed that PUPEG membranes have good hydrophilic properties and show selectivity for water selectivity, which shows an increase with increasing molecular weight of the polymer and also with increasing the length of the polymer membrane of the membrane. The results obtained from experimental studies were compared with the data obtained from the model and then the results were reviewed. For this research, a type of neural network called multilayer perceptron (MLP) and also a Leonberg-Marquardt diffusion learning algorithm. The model was implemented with one output and 2 inputs. A Purlin algorithm was used for the output layer and another type of Tansig activation algorithm was used for the hidden layer. It should also be noted that 5 neurons were used for the hidden layer. After the data was processed, about 70% of this data was used to learn the model, another 15% to credit and the remaining 15% to experience. The results obtained from this model were very accurate. The results obtained by comparing the simulation model and the experimental results of the separation of alcohol-water mixture showed that the neural network models are able to predict the experimental results with the least error.