Predicting the Compressive Properties of Carbon Foam Using Artificial Neural Networks

Area of Science:

• Materials Science
• Computational Materials Science
• Machine Learning Applications

Background:

• Polyurethane-derived carbon foams are advanced materials with tunable properties.
• Predicting mechanical properties like compressive stress is crucial for material design and application.
• Artificial Neural Networks (ANNs) offer a powerful tool for complex property prediction in materials.

Purpose of the Study:

• To develop and evaluate an Artificial Neural Network (ANN) model for predicting the compressive properties of polyurethane-derived carbon foam.
• To investigate the influence of strain, pore density, and solvent type on compressive stress.
• To establish a reliable computational method for material characterization.

Main Methods:

• A feed-forward ANN with four hidden layers (100 neurons each) was designed.
• Input variables (strain, pore density, solvents) were preprocessed using one-hot encoding and normalization.
• Model performance was assessed using Mean Square Error (MSE), Mean Absolute Error (MAE), and Coefficient of Determination (R²), with the Adam optimizer and ReLU activation functions.

Main Results:

• The ANN model achieved a high Coefficient of Determination (R²) of 0.9797.
• The overall average Mean Square Error (MSE) was 36.34, and Mean Absolute Error (MAE) was 4.42.
• The model demonstrated excellent predictive capability for compressive stress based on the selected input parameters.

Conclusions:

• The developed ANN model accurately predicts the compressive properties of polyurethane-derived carbon foam.
• This approach provides an efficient and reliable method for material property prediction, reducing the need for extensive experimental testing.
• The findings highlight the potential of machine learning in accelerating materials discovery and development.