Aug 2026· Advancement of science· 0 citations· 45 references
Medicine
Abstract
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.
Results indicate that support vector regression (SVR) provides the most consistent overall performance across all regimes and offers a strong balance between accuracy and computational efficiency, and a Bayesian neural network (BNN) achieves competitive predictive performance while additionally enabling uncertainty estimation.
Muhammad Bilal Jan, Zengchao Wu, Mengyu Chai· Metals· 0 citations
An interpretable and uncertainty-aware machine-learning framework for estimating the shear capacity of FRCM-strengthened beams enables accurate, transparent, and uncertainty-aware assessment of shear capacity in FRCM-strengthened concrete beams.
Xiangsheng Liu, G. Figueredo, G. Gordon et al.· Journal of composites for co...· 0 citations
A novel hybrid method integrating physics-informed neural network (PINN) and Gaussian process regression (GPR) that enforces the B4 creep model as a physics-informed constraint by embedding its governing equations into the loss function, effectively incorporating physical knowledge into data-driven training.
Zhiren Tao, Jianxin Peng, Shijie Liao et al.· Journal of materials in civi...· 0 citations
A comprehensive data-driven framework integrating ensemble machine learning models with systematic hyperparameter sensitivity analysis and explainable artificial intelligence techniques is proposed, demonstrating that the XGB model significantly outperforms the other approaches, achieving superior accuracy and robust generalization.
Qaim Shah, Waheed Ali Khoso, Fawad Iqbal et al.· Discover Artificial Intellig...· 0 citations
Investigating the impact of as‐constructed air voids (AV
s
) on the long‐term evolution of asphalt pavement International Roughness Index (IRI) through field tests is severely constrained by prohibitive monitoring costs. Furthermore, the complex coupling effects of environmental aging and traffic‐induced secondary compaction make it exceptionally challenging to isolate the specific contribution of AV
s
to IRI progression. Based on the Long‐Term Pavement Performance (LTPP) database, this study utilizes mainstream machine learning (ML) models to develop a comprehensive predictive framework for long‐term IRI evolution. A rule‐based data cleaning procedure guided by pavement deterioration theory was first implemented to eliminate physically inconsistent anomalous observations. Evaluation results demonstrated that this cleaning strategy effectively enhanced the validation performance across all candidate models, with the multilayer perceptron (MLP) model exhibiting the highest predictive accuracy and generalization capability. The Shapley additive explanations (SHAP) framework was introduced based on the selected MLP model to conduct interpretability analysis, verifying that the data‐driven feature mappings were consistent with established engineering principles. On this basis, a synthetic input matrix was constructed to quantitatively evaluate the model‐estimated marginal influence of AV
s
on long‐term IRI progression. Under the controlled baseline scenario, the model‐predicted influence of AV
s
on the IRI exhibits a non‐monotonic U‐shaped trend, with an AV
s
content of approximately 6% yielding the lowest predicted IRI. Ultimately, this study bridges the gap between methodological interpretability and practical application, providing a model‐based exploratory framework for compaction quality control in pavement construction.
Peixuan Lin, Fan Gu· Applied Research· 0 citations
Highlights A PE-ECC database comprises 383 material-level records from 90 literature sources. The MoE model predicts four PE-ECC properties with R2 values of 0.950–0.971. SHAP, ALE and response maps distinguish strength trends from tensile deformation. Binder, W/B, S/B and fiber variables show property-specific relationships. Support-filtered screening yields database-supported candidates for laboratory validation. Abstract Featuring considerable tensile ductility and multiple cracking behavior, polyethylene fiber-reinforced engineered cementitious composites (PE-ECCs) are promising cement-based materials for engineering construction. However, establishing accurate design models for evaluating the mechanical properties of PE-ECC is a challenging task owing to the complex material components. This study presents an interpretable data-driven framework for predicting the mechanical properties of PE-ECC using mixture-of-experts (MoE) learning. A database comprising 383 deduplicated material-level records from 90 verified literature sources was compiled for modeling the compressive strength, ultimate tensile strain, ultimate tensile strength and first-cracking tensile strength of PE-ECC. An MoE prediction model was developed by integrating XGBoost, LightGBM, CatBoost, WDBPANN and TabPFN through out-of-fold stacking and learned gating. The model achieved coefficient of determination (R2) values of 0.971, 0.950, 0.970 and 0.954 for the four mechanical properties, respectively. Shapley additive explanations (SHAP), accumulated local effects (ALE) and response maps were used to examine the fitted nonlinear associations between the reported mixture variables and each target property. Based on these relationships, support-filtered virtual screening was conducted within the database-supported design space to identify candidate mixtures for subsequent experimental verification. The framework links target-specific prediction with mixture-response interpretation and confines screening to regions supported by reported PE-ECC mixtures.