Study of the Current–Voltage Characteristics of Membrane Systems Using Neural Networks

This article is dedicated to the construction of neural networks for the prediction of the current–voltage characteristic (CVC). CVC is the most important characteristic of the mass transfer process in electro-membrane systems (EMS). CVC is used to evaluate and select the optimal design and effective operating modes of EMS. Each calculation of the CVC at the given values of the input parameters, using developed analytical-numerical models, takes a lot of time, so the CVC is calculated in a limited range of parameter changes. The creation of neural networks allowed for the use of prediction to obtain the CVC for a wider range of input parameter values and much faster, saving computing resources. The regularities of the behavior of CVC for various values of input parameters were revealed. During this work, several different neural network architectures were developed and tested. The best predictive results on test samples are given by the neural network consisting of convolutional and LSTM (Long Short-Term Memory) layers.