Study the hydrotropic behaviour of butyl stearate using ANN tools

This study investigates the prediction of the thermophysical properties of butyl stearate in solutions with citric acid, urea, and nicotinamide using Artificial Neural Networks (ANNs). The ANN model uses hydrotropic concentration and temperature to predict these properties. The study focuses on binary mixtures at various temperatures (303, 313, 323, and 333 K). To achieve accurate predictions, researchers trained a committee of ANNs using experimental data. This iterative process optimizes the network architecture and avoids overfitting, a common problem in machine learning. The trained ANN can then predict thermophysical properties for intermediate hydrotropic concentrations without additional experiments. Besides, the study demonstrates the versatility of ANNs by implementing a successful model for multi-pass turning operations in MATLAB, showing superior accuracy compared to other methods. Visualizations like magnitude response curves, FFT spectrums, contour plots, and 3D surface plots helped to identify the optimal hydrotropic concentration with a remarkable 2% margin of error.