Predicting the Compressive Properties of Carbon Foam Using Artificial Neural Networks

This article focusses on predicting the compressive properties of polyurethane-derived carbon foam using an artificial neural network (ANN) approach. To train the model, strain, pore density (20, 40, and 60 ppi), and solvents (acetone, ethanol, and methanol) were used as inputs, while compressive stress was used as output. Categorical variables like acetone, ethanol, and methanol were converted to binary form before training the ANN model by using one-hot encoding mechanism. Both inputs and outputs were normalized to prevent features with larger numerical ranges from dominating the training process. A feed-forward ANN with four hidden layers, each containing 100 neurons, was constructed. The performance of the ANN model was tested using three metrics: mean square error (MSE), mean absolute error (MAE), and coefficient of determination (R²). The Adam optimizer was used to optimize the weights and biases of the ANN. The model was trained for 10,000 epochs with a batch size of 50. Rectified Linear Unit (ReLU) and linear functions were used as activation functions for the hidden layers and the output layer, respectively. From the results, overall average MSE, MAE, RMSE, and R² values of 36.34, 4.42, 6.00, and 0.9797, respectively, were obtained.