Trustworthy Multimodal Attention Framework for Creep Rupture Life Prediction Under Data‐Scarce Conditions: A Case Study on IN718
5 affiliations
- 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
- 2National Engineering Research Center of Light Alloy Net Forming Shanghai Jiao Tong University Shanghai People's Republic of China
- 3State Key Laboratory for Advanced Metals and Materials University of Science and Technology Beijing Beijing People's Republic of China
- 4Hebei Lron and Steel Laboratory North China University of Science and Technology Tangshan China
- 5Yanzhao Lron and Steel Laboratory Tangshan People's Republic of China
Abstract
An attention‐based multimodal deep learning framework is developed to fuse processing parameters with microstructural micrographs for predicting creep rupture life of IN718 under a fixed creep testing condition. Under a predefined composition‐stratified, sample‐level split, the framework achieved a mean test‐set R2 of 0.917 ± 0.014 across 50 random‐seed training repetitions. The corresponding mean RMSE and MAPE were 0.14% and 6.0%, respectively. Interpretability analyses suggest that the predictions are consistent with established metallurgical understanding, particularly the important role of δ‐phase characteristics. Furthermore, uncertainty quantification endows the model with self‐assessment capabilities, allowing it to reliably quantify the confidence of its predictions. This study establishes a methodological framework that unifies predictive accuracy, physical interpretability, and model confidence, providing a validated paradigm for developing trustworthy AI models for materials design under data‐limited conditions.
Related
- Graph attention framework for site-selectivity prediction in undirected C(sp3)–H borylation
- Weighted Views Representation with Multi-Head Coulomb Attention for Molecular Property Prediction
- Prediction of Singlet-Triplet Gaps Using Self-Attention–Based Regression with Model Integration into a Molecular Editor
- Hierarchical Multimodal Large Language Model-Based Interpretation and Aggregation of PDP and ICE Explanations for Battery State-of-Charge Prediction in Electric Vehicles
- Task-adaptive multimodal molecular representations for structure-sensitive property prediction
BibTeX
@article{lv2026trustworthy,
title = {{Trustworthy Multimodal Attention Framework for Creep Rupture Life Prediction Under Data‐Scarce Conditions: A Case Study on IN718}},
author = {Haopeng Lv and Dayong Wu and Ziyuan Rao and Qian Wang and Haikun Ma and Jie Kang and Wang Li and Huicong Dong and Yandong Wang and Zhinan Yang and Ru Su},
journal = {Advanced Science},
year = {2026},
month = aug,
doi = {10.1002/advs.77026},
url = {https://doi.org/10.1002/advs.77026},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Trustworthy Multimodal Attention Framework for Creep Rupture Life Prediction Under Data‐Scarce Conditions: A Case Study on IN718
AU - Haopeng Lv
AU - Dayong Wu
AU - Ziyuan Rao
AU - Qian Wang
AU - Haikun Ma
AU - Jie Kang
AU - Wang Li
AU - Huicong Dong
AU - Yandong Wang
AU - Zhinan Yang
AU - Ru Su
PY - 2026
DA - 2026/08/10
JO - Advanced Science
DO - 10.1002/advs.77026
UR - https://doi.org/10.1002/advs.77026
AB - An attention‐based multimodal deep learning framework is developed to fuse processing parameters with microstructural micrographs for predicting creep rupture life of IN718 under a fixed creep testing condition. Under a predefined composition‐stratified, sample‐level split, the framework achieved a mean test‐set R2 of 0.917 ± 0.014 across 50 random‐seed training repetitions. The corresponding mean RMSE and MAPE were 0.14% and 6.0%, respectively. Interpretability analyses suggest that the predictions are consistent with established metallurgical understanding, particularly the important role of δ‐phase characteristics. Furthermore, uncertainty quantification endows the model with self‐assessment capabilities, allowing it to reliably quantify the confidence of its predictions. This study establishes a methodological framework that unifies predictive accuracy, physical interpretability, and model confidence, providing a validated paradigm for developing trustworthy AI models for materials design under data‐limited conditions.
SN - 2198-3844
ER -