Prediction of density and viscosity of distilled tire pyrolysis oil blends with diesel by artificial neural networks

Abstract

Density and viscosity are key physical properties of fuels with regard to utilization in the compression ignition engine. In this study, Diesel and Distilled Tyre Pyrolysis oil (DTPO) were blended at various ratios by mass. Experimental data was then obtained for density and viscosity of the blends at temperature range of 20 – 80°C. The data was then used to train an Analytical Neural Network model with blend concentration and temperature being inputs while viscosity and density were outputs. Both the density and viscosity increased with increase in the concentration of DTPO in the blend, and both decreased with rise in temperature. Levenberg–Marquardt learning algorithm for was used with logsig and purelin transfer function for the hidden and output layers respectively. Different numbers of neuron were tried for the hidden layer. It was found that a network with seven neurons in the hidden layer was able to make accurate predictions. The correlation coefficients (R) for Training, testing and validation of the model were 0.99846, 0.99882 and 0.99915 respectively, while that for the whole network was 0.99858. The Mean Square error between the predicted and desired values was found to be 0.002 by this model