Improvement of Mixed-Mode I/II Fracture Toughness Modeling Prediction Performance by Using a Multi-Fidelity Surrogate...

Area of Science:

• Mechanical Engineering
• Materials Science
• Computational Science

Background:

• Artificial intelligence (AI) is increasingly used in mechanical engineering to solve complex problems, including fracture mechanics.
• Predicting crack behavior under mixed-mode loading using AI requires substantial experimental data, which is often a limitation due to the nature of destructive testing.

Purpose of the Study:

• To develop an AI model for fracture mechanics that overcomes the need for large datasets.
• To propose a data combination technique using limited experimental data and data from fracture criteria.

Main Methods:

• A multi-fidelity surrogate model was employed to combine scarce experimental data with data derived from established fracture criteria.
• The proposed data combination technique was validated by testing the mixed-mode I/II fracture toughness of Polymethyl methacrylate (PMMA) material.

Main Results:

• AI models trained with combined data (experimental + fracture criteria) demonstrated superior predictive performance compared to models trained solely on experimental data.
• The multi-fidelity surrogate model effectively enhanced the predictive accuracy of AI in fracture mechanics with limited data.

Conclusions:

• Combining limited experimental data with fracture criteria data via multi-fidelity surrogate modeling is a viable strategy to improve AI applications in fracture mechanics.
• This approach significantly reduces the data requirement for training AI models, making them more practical for complex engineering problems.