Artificial neural networks for predicting potentiodynamic tests of brass 70-30. (2023)

• Record Type: Journal Article
• Authors: Monzamodeth, R.S.; Flores-Alvarez, J.F.; Reyes-Chaparro, I.; Castillo, F.; Campillo, B.; Flores, O.
• Is Part Of: Materials today. Volume 80:Part 2(2023)
• Page Start: 1514
• Page End: 1518
• Publication Date: 2023
• Subjects: Artificial neural network -- Brass 70–30 -- Potentiodynamic polarization curves -- Molar concentration -- Artificial intelligence; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2023.01.287
• Languages: English
• ISSNs: 2214-7853

Abstract

Highlights: Electrochemical tests were performed employing brass 70–30 at three different molar concentrations. A predictive model of potentiodynamic curves was developed using a back-propagation artificial neural network. Results obtained of the comparison between the experimental and predicted curves were satisfactory, achieving a Pearson's coefficient of 0.90. The artificial neural network was tested, simulating potentiodynamic curves at 0.7 M. It is feasible to consider the wide use of artificial intelligence to predict potentiodynamic curves of brass 70–30. Abstract: In the present work an artificial neural network (ANN) was developed for the prediction of brass (70–30) potentiodynamic polarization curves. Several samples at three different molar concentrations were assessed; a silver reference electrode was used in ammonium hydroxide (NH4 OH) electrolyte, and the potential sweep was performed at 800 mV between the cathode and the anode. The developed neural network was trained using 30 potentiodynamic polarization curves and its hidden layers were developed employing a hyperbolic tangent activation function. Furthermore, the error was adjusted by the Levenberg-Marquardt gradient descent method. The results obtained from the comparison between the experimental and simulated curves were satisfactory, achieving a Pearson's coefficient of 0.90. Therefore, the predictive model developed is capable of being used as a data mining tool, auxiliary in the polarization tests under similar experimental conditions.