Artificial Neural Network Prediction of pH in Nylon 66 Salt Solutions: A Machine Learning Approach

ABSTRACT This study investigates the application of artificial neural networks (ANNs) to predict the pH of Nylon 66 salt solutions. The ANN was optimized by systematically evaluating key parameters—including the number of hidden layers and neurons, activation function type, and optimization algorithm—using a dataset of pH values measured across varying temperatures and concentrations. The results showed that when the number of hidden layers was 24, and the number of neurons in each hidden layer was 253, the mean squared error (MSE) between the predicted pH by ANN and the experimental data reached 10 −4 . The ReLU activation function and the lbfgs optimization algorithm were identified as the most effective for the prediction task. The optimized ANN demonstrated superior predictive accuracy with a determination coefficient ( R 2 ) exceeding 0.99, outperforming the traditional first‐order ionization theory. This research provides a robust method for controlling the Nylon 66 synthesis process and highlights the potential of ANN in complex chemical systems.