Experimental investigation and prediction of pour point using artificial intelligence

Predicting the pour point temperature is one way to prevent the formation of wax deposition. In this paper, two types of neural networks, MLP and RBF, have been studied and modeled to predict pour point. In the MLP network, the optimal number of input parameters as well as the best activation function have been investigated. The constructed network is then compared to the RBF neural network. The results of the constructed networks were evaluated and validated by the data set obtained in the laboratory. Comparing the two networks, it was concluded that the multilayer perceptron neural network (MLP) has a higher prediction accuracy than the radial base neural network (RBF). By comparing the number of input parameters in the multilayer perceptron network, it was concluded that Using two parameters of wax content and cloud point temperature in predicting pour point temperature has a better performance than using ۷ parameters as input, which reduces the costs associated with performing tests such as the SARA test. The results also indicate that using two parameters, appropriate predictions of the pour point temperature are achieved, which leads to a decrease in the testing costs such as SARA analysis.

Keywords:

Pour point , Multilayer perceptron neural , network , Radial base neural network , Wax ,