Corrosion rate prediction for metals in biodiesel using artificial neural networks

Abstract

The objective of this research was to develop a direct artificial neural network with the ability to predict a corrosion rate of metals in different biodiesel. Experimental values were obtained by the electrochemical noise technique, EN, as well as, information reported in the literature. A backpropagation model was proposed with three layers; metal and biodiesel composition, blend biodiesel/diesel, total acid number (TAN), temperature and exposure time were considered as input variables in the model. The best fitting training data were acquired with 24:4:1, considering a Levenberg –Marquardt learning algorithm, a hyperbolic tangent and linear transfer functions in the hidden and output layer respectively. Experimental and simulated data were compared satisfactorily through the linear regression model with a correlation coefficient of 0.9885 and a mean square error, MSE, of 2.15 × 10−4 in the validation stage. Furthermore, the model agreed the requirements of the slope and the intercept statistical test with a 99% confidence. The obtained results indicated that the ANN model could be attractive as corrosion rate estimator.