Forecasting the Viscosity of Bio-Nano Oils Using the Best Fit Algorithm for Artificial Neural Network

This article explores the scope of predicting the viscosity of titanium dioxide (TiO2) bio-nano oil using an artificial neural network (ANN) for effective load-bearing applications. Jatropha oil bio-nanolubricant samples were prepared at nanoparticle concentrations ranging from 0.3% to 1.8%, and their viscosity was measured using a Stabinger viscometer across a shear rate range of 0–200 s−1. These data were used to train an ANN model for accurately predicting the viscosity at any operating condition of the jatropha bio-nanolubricant using licensed matlab software (version R2024b). Initially, a feed-forward perceptron model was developed with two input variables, e.g., temperature and particle concentration, and one output variable, e.g., viscosity. Later, shear rate was also incorporated as an additional input, which further reduced the root-mean-square error (RMSE), which is a measure of the average magnitude of the error between the predicted values by the ANN and the actual experimental values. It was observed that the ANN model with only ten hidden neurons achieved an average RMSE of 17.254 using the initial input variables. However, when shear rate was included in the feature set, the RMSE was found to reduce by almost 96.13%, thereby improving the accuracy of the model. The proposed ANN model was also compared with machine learning (ML) methods, and a higher coefficient of determination (R2 = 0.94) was achieved ANN, showcasing its superior predictive capability.