A Multi-layer Perceptron-based Approach for Prediction of the Crude Oil Pyrolysis Process

In this research, thermogravimetric analysis was used to examine pyrolysis behavior of six crude oils. Thermogravimetric analysis experiments were performed in nitrogen atmosphere at the heating rates of 1, 5, and 10 degrees C/min up to 800 degrees C. Then a multi-layer perceptron neural network was developed to predict residual crude oil as a function of temperature in the pyrolysis process based on a trial and error technique. The results showed that the proposed neural network can predict the residual crude oil function of temperature with an acceptable accuracy, approximately 3.5% average relative error.