Exploring the mechanical behavior of friction material composites using artificial intelligence.

Area of Science:

• Materials Science
• Computational Materials Science
• Tribology

Background:

• Artificial intelligence (AI) offers advanced solutions for complex problems in science and engineering.
• Traditional methods for developing new materials, like friction composites, often involve high costs and extensive experimentation.
• AI can optimize material design and predict properties with greater efficiency.

Purpose of the Study:

• To develop AI-driven mathematical models for predicting mechanical properties of friction materials based on their chemical composition.
• To create an algorithm for optimizing existing and proposing novel friction material compositions.
• To evaluate the predictive accuracy of AI models compared to traditional methods.

Main Methods:

• Utilized an extensive database of friction material chemical compositions and mechanical properties.
• Developed an AI algorithm combining rule-based systems, neural networks, and particle swarm optimization.
• Produced and tested physical samples based on algorithm predictions to validate model accuracy.

Main Results:

• AI models, particularly neural networks, demonstrated significantly higher predictive accuracy for novel friction material compositions.
• Root Mean Square Error (RMSE) for neural network predictions was substantially lower than for multilinear models.
• On average, multilinear models showed a 48.6% larger RMSE compared to AI predictions for non-existent compositions.

Conclusions:

• AI, specifically neural networks, provides a superior approach for predicting mechanical properties of new friction materials.
• The developed algorithm effectively optimizes material composition and proposes novel formulations.
• AI-based modeling reduces the need for costly physical experiments and accelerates materials discovery.