Nanoscratch-Informed Micro-to-Macro Surrogate Modeling of Fracture Properties in Nano-Silica Modified Portland Cement Paste
4 affiliations
- 1School of Urban Construction, Hangzhou Polytechnic, Hangzhou 311402, China
- 2Zhejiang Scientific Research Institute of Transport, Hangzhou 311402, China
- 3College of Architectural Engineering, Shanxi Vocational University of Engineering Science and Technology, Jinzhong 030619, China
- 4Intelligent Transportation System Research Center, Southeast University, Nanjing 210096, China
Abstract
Macroscopic fracture testing of nano-modified cement paste requires the labor-intensive fabrication of notched beams. This study establishes a micro-to-macro surrogate modeling framework mapping localized nanomechanical scratch signatures to macroscopic fracture resistance. A database was compiled for pure Portland cement paste (w/c = 0.30, 28-day curing) across seven nano-silica (NS) dosages (0.0–3.0%). Across 21 independent physical beam specimens (Nmacro = 21) and 470 validated phase-segmented nanoscratch observations (Nmicro = 470), localized descriptors were acquired across unhydrated clinker, hydration products, and the clinker–hydrate interfacial transition zone (micro-ITZ). Within a leak-free pipeline, an XGBoost surrogate model was trained using specimen-grouped nested cross-validation (Nested GroupKFold), achieving cross-validated R2 values of 0.878 ± 0.024 (95% CI [0.834, 0.919]) for macro-fracture toughness (KICmacro) and 0.865 ± 0.028 (95% CI [0.815, 0.912]) for fracture energy (Gf). Leave-one-dosage-out validation demonstrated that when provided with the target mix’s in situ nanoscratch fingerprint, the model successfully tracked the non-monotonic fracture evolution, including the peak at 2.0% NS and the agglomeration degradation at 3.0% NS. Bootstrap Tree-SHAP attribution identified interfacial micro-toughness (KIC,ITZmicro) as the primary predictive contributor (42.5% [95% CI: 37.8%, 47.1%]), revealing a model-suggested transition region around 0.80 MPa ⋅ m1/2 above which interfacial crack resistance associates with substantial macroscopic toughening. This framework provides an efficient surrogate approach for evaluating macro-fracture properties from micromechanical indicators.
Related
- Electron-Density-Informed Molecular Property Prediction Using Surrogate Functionals
- Hierarchical Surrogate Modeling and Optimization of the SO2 Recovery–Water Loss Trade-Off in Basic Aluminum Sulfate Regeneration
- Physics-Informed Deep Learning Surrogate for Mechanical-Response Prediction and Failure Assessment of High-Performance Polymer Composites
- The Role of Carbon Fibers in FFF 3D Printed Composites: A Comprehensive Review on the Effect on Mechanical Properties, Fracture Toughness, and Tribological Performance
- NanoDMA and Nanoscratch Study of (Visco)‐Elastic and Plastic Behaviors of a Thin Film of a Molecular Spin Crossover Complex
BibTeX
@article{zhou2026nanoscratchinformed,
title = {{Nanoscratch-Informed Micro-to-Macro Surrogate Modeling of Fracture Properties in Nano-Silica Modified Portland Cement Paste}},
author = {Xiangong Zhou and Manhui Liu and Xiancheng Zhou and Weikang Kong},
journal = {Nanomaterials},
year = {2026},
month = nov,
doi = {10.3390/nano16211440},
url = {https://www.mdpi.com/2079-4991/16/21/1440},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Nanoscratch-Informed Micro-to-Macro Surrogate Modeling of Fracture Properties in Nano-Silica Modified Portland Cement Paste
AU - Xiangong Zhou
AU - Manhui Liu
AU - Xiancheng Zhou
AU - Weikang Kong
PY - 2026
DA - 2026/11/06
JO - Nanomaterials
DO - 10.3390/nano16211440
UR - https://www.mdpi.com/2079-4991/16/21/1440
AB - Macroscopic fracture testing of nano-modified cement paste requires the labor-intensive fabrication of notched beams. This study establishes a micro-to-macro surrogate modeling framework mapping localized nanomechanical scratch signatures to macroscopic fracture resistance. A database was compiled for pure Portland cement paste (w/c = 0.30, 28-day curing) across seven nano-silica (NS) dosages (0.0–3.0%). Across 21 independent physical beam specimens (Nmacro = 21) and 470 validated phase-segmented nanoscratch observations (Nmicro = 470), localized descriptors were acquired across unhydrated clinker, hydration products, and the clinker–hydrate interfacial transition zone (micro-ITZ). Within a leak-free pipeline, an XGBoost surrogate model was trained using specimen-grouped nested cross-validation (Nested GroupKFold), achieving cross-validated R2 values of 0.878 ± 0.024 (95% CI [0.834, 0.919]) for macro-fracture toughness (KICmacro) and 0.865 ± 0.028 (95% CI [0.815, 0.912]) for fracture energy (Gf). Leave-one-dosage-out validation demonstrated that when provided with the target mix’s in situ nanoscratch fingerprint, the model successfully tracked the non-monotonic fracture evolution, including the peak at 2.0% NS and the agglomeration degradation at 3.0% NS. Bootstrap Tree-SHAP attribution identified interfacial micro-toughness (KIC,ITZmicro) as the primary predictive contributor (42.5% [95% CI: 37.8%, 47.1%]), revealing a model-suggested transition region around 0.80 MPa ⋅ m1/2 above which interfacial crack resistance associates with substantial macroscopic toughening. This framework provides an efficient surrogate approach for evaluating macro-fracture properties from micromechanical indicators.
SN - 2079-4991
ER -