SA-PINN: A self-attention enhanced physics-informed neural network for multiaxial fatigue life prediction with small samples

Key Points

• Fatigue life prediction improves with self-attention mechanisms, enhancing accuracy overall.
• The model achieved a notable accuracy increase of 20% for fatigue life predictions under multiaxial stress conditions.
• Analysis employs physics-informed neural networks to leverage underlying physical principles and small datasets.
• These findings suggest enhanced predictive capabilities, calling for wider application in engineering contexts.