Related Experiment Videos
Trustworthy Multimodal Attention Framework for Creep Rupture Life Prediction Under Data-Scarce Conditions: A Case
Haopeng Lv1, Dayong Wu1, Ziyuan Rao2
1Hebei Short Process Steelmaking Technology Innovation Center, School of Materials Science and Engineering, Hebei University of Science and Technology, Shijiazhuang, People's Republic of China.
Advanced Science (Weinheim, Baden-Wurttemberg, Germany)
|August 10, 2026
Summary
This study introduces an AI framework to predict the creep rupture life of IN718 by combining processing data and microstructural images. The model demonstrates high accuracy and interpretability, offering trustworthy materials design insights.
Area of Science:
- Materials Science
- Artificial Intelligence
- Mechanical Engineering
Background:
- Predicting creep rupture life is crucial for IN718 alloy performance.
- Traditional methods often require extensive experimental data.
- Data-driven approaches offer potential for accelerated materials design.
Purpose of the Study:
- Develop an attention-based multimodal deep learning framework.
- Fuse processing parameters and microstructural micrographs for creep life prediction.
- Establish a trustworthy AI paradigm for materials design under data limitations.
Main Methods:
- Implemented an attention-based multimodal deep learning architecture.
- Utilized processing parameters and microstructural micrographs as input features.
- Performed composition-stratified, sample-level data splitting for validation.
- Incorporated uncertainty quantification for model confidence assessment.
Main Results:
- Achieved a mean test-set R² of 0.917 ± 0.014.
- Reported mean RMSE of 0.14% and MAPE of 6.0%.
- Interpretability analysis confirmed alignment with metallurgical understanding, highlighting δ-phase characteristics.
Conclusions:
- The developed framework accurately predicts creep rupture life of IN718.
- The AI model exhibits physical interpretability and self-assessment capabilities.
- This study provides a validated approach for trustworthy AI in materials design, especially with limited data.