DeepOFormer: Deep Operator Learning with Domain-informed Features for Fatigue Life Prediction

2025-03-28

AI Summary

Accurate corrosion fatigue life prediction for aluminum alloys under small-sample conditions remains challenging due to poor generalizability and susceptibility to overfitting. Method: This work reformulates the S–N curve modeling task as a deep operator learning problem, integrating three physics-informed, domain-aware features—Stüssi, Weibull, and PM—to ensure both physical interpretability and data adaptability. A Transformer-based encoder architecture is employed, coupled with a mean L2 relative error loss function to enhance regression accuracy. Contribution/Results: Evaluated on 54 aluminum alloy corrosion S–N curves, the model achieves an R² of 0.9515, MAE of 0.2080, and MRE of 0.5077—substantially outperforming baselines including DeepONet, TabTransformer, and XGBoost. This study pioneers a physics-embedded operator learning paradigm, offering a novel framework for small-sample fatigue modeling with strong theoretical grounding and empirical efficacy.

Technology Category

• Machine Learning: Feature Construction/Reformulation
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Fatigue life characterizes the duration a material can function before failure under specific environmental conditions, and is traditionally assessed using stress-life (S-N) curves. While machine learning and deep learning offer promising results for fatigue life prediction, they face the overfitting challenge because of the small size of fatigue experimental data in specific materials. To address this challenge, we propose, DeepOFormer, by formulating S-N curve prediction as an operator learning problem. DeepOFormer improves the deep operator learning framework with a transformer-based encoder and a mean L2 relative error loss function. We also consider Stussi, Weibull, and Pascual and Meeker (PM) features as domain-informed features. These features are motivated by empirical fatigue models. To evaluate the performance of our DeepOFormer, we compare it with different deep learning models and XGBoost on a dataset with 54 S-N curves of aluminum alloys. With seven different aluminum alloys selected for testing, our DeepOFormer achieves an R2 of 0.9515, a mean absolute error of 0.2080, and a mean relative error of 0.5077, significantly outperforming state-of-the-art deep/machine learning methods including DeepONet, TabTransformer, and XGBoost, etc. The results highlight that our Deep0Former integrating with domain-informed features substantially improves prediction accuracy and generalization capabilities for fatigue life prediction in aluminum alloys.

Problem

• Predicts fatigue life of materials using operator learning
• Addresses overfitting in small fatigue experimental datasets
• Improves accuracy with domain-informed features and transformers

Innovation

• Transformer-based encoder for operator learning
• Mean L2 relative error loss function
• Domain-informed features from empirical models