Artificial neural network architecture for rheological property prediction of a novel hybrid nanolubricant for automotive spark-ignition engine

Abstract

In this research article, the rheological behavior of ZnO-TiO2/5W30 hybrid nanolubricant is evaluated experimentally to study the suitability of using it as an effective engine lubricant. According to the results, the hybrid nanolubricant (0.1 wt%) produced shows less viscosity than the base engine oil at the selected range’s lower temperature. It will also assist in reducing the cold-start phase issues of an automotive engine. An artificial neural network (ANN) approach is used to find the best model for predicting nanolubricant samples’ viscosities of various concentrations at different temperatures (range of 25–65 °C) and shear rates. A combination of 27 ANN models was selected and trained using the Levenberg–Marquardt backpropagation algorithm. As an optimum configuration, the ANN model designed for the hybrid nanolubricant consisted of two hidden layers with 3 and 2 neurons in the first and second layers, respectively. The tan-sigmoid function activated both the layers. The best neural network model’s regression coefficient value was 0.9987 and had a mean squared error lower than 0.0002. The resulting model could predict the rheological behavior of the hybrid nanolubricant presented in this study with high accuracy. Graphic abstract: [Figure not available: see fulltext.]