Microscopic stress concentration of salt crystallization damage on the surface of cement-based materials: Simulation characterization and machine learning analysis

The salt crystallization pressure is an important effecting factor for the physical salt corrosion of cement-based materials. However, considering complex micropore structures and different material boundary conditions, the evolution law of stress concentration in cement paste caused by the crystallization pressure has not yet been clarified. This paper focused on the confined crystallization of NaCl within the micropores of hardened cement paste. Two-dimensional dual-pore FEA models considering the micropore structures and different material boundary conditions were established. The effects of micropore and crystal morphology, crystallization pressure, infilling degree, pore spacing, and boundary condition on the maximum tensile stress were numerically simulated and analyzed. In addition, a FEA-driven machine learning dataset was constructed, and maximum tensile stress machine learning models were established using RF, XGBoost, and BPNN methods. Further, by integrating SHAP, PDP, and MRA, the main controlling factors were quantified and nonlinear interactions were identified. Taken together, the findings show that the stress concentration induced by salt crystallization is not solely determined by the crystallization pressure, but is influenced by the pore geometry, crystal infilling degree, and the boundary condition. The crystallization pressure generates a significant local tensile stress concentration at the end of the micropore's long axis. When micropore is fully infilled, the maximum tensile stress in the cement paste caused by the salt crystallization pressure can reach 17.88 times the crystallization pressure. The maximum tensile stress increases with the increase of pore aspect ratio and crystallization pressure, and shows a significant nonlinear growth with the infilling degree; in contrast, the crystal aspect ratio and boundary length have a relatively weak influence. The results of machine learning interpretation further indicate that the type of boundary conditions and the degree of infilling are the main sensitive factors, and at the same time, high infilling degree shows a significant synergistic effect with high crystallization pressure. The BPNN method achieved the best prediction performance, with a R 2 of 0.992, RMSE of 0.302 MPa, and MAE of 0.148 MPa. The FEA-driven machine learning framework established provides a new analytical method for the quantitative characterization of micropore-scale salt crystallization stress concentration and the evaluation of physical salt corrosion damage in cement-based materials.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-19
DOI
https://doi.org/10.21203/rs.3.rs-11767243/v1
Primary Topic
Building materials and conservation
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Microscopic stress concentration of salt crystallization damage on the surface of cement-based materials: Simulation characterization and machine learning analysis

Ze Zhang, Zhenhua Huang, Jinbo Yang, Jinpeng ZHANG et al.
Research Square
Building materials and conservation
preprint

Microscopic stress concentration of salt crystallization damage on the surface of cement-based materials: Simulation characterization and machine learning analysis

Ze Zhang, Zhenhua Huang, Jinbo Yang, Jinpeng ZHANG, Siqi Li, Xuefeng Liu
preprint en

Abstract

The salt crystallization pressure is an important effecting factor for the physical salt corrosion of cement-based materials. However, considering complex micropore structures and different material boundary conditions, the evolution law of stress concentration in cement paste caused by the crystallization pressure has not yet been clarified. This paper focused on the confined crystallization of NaCl within the micropores of hardened cement paste. Two-dimensional dual-pore FEA models considering the micropore structures and different material boundary conditions were established. The effects of micropore and crystal morphology, crystallization pressure, infilling degree, pore spacing, and boundary condition on the maximum tensile stress were numerically simulated and analyzed. In addition, a FEA-driven machine learning dataset was constructed, and maximum tensile stress machine learning models were established using RF, XGBoost, and BPNN methods. Further, by integrating SHAP, PDP, and MRA, the main controlling factors were quantified and nonlinear interactions were identified. Taken together, the findings show that the stress concentration induced by salt crystallization is not solely determined by the crystallization pressure, but is influenced by the pore geometry, crystal infilling degree, and the boundary condition. The crystallization pressure generates a significant local tensile stress concentration at the end of the micropore's long axis. When micropore is fully infilled, the maximum tensile stress in the cement paste caused by the salt crystallization pressure can reach 17.88 times the crystallization pressure. The maximum tensile stress increases with the increase of pore aspect ratio and crystallization pressure, and shows a significant nonlinear growth with the infilling degree; in contrast, the crystal aspect ratio and boundary length have a relatively weak influence. The results of machine learning interpretation further indicate that the type of boundary conditions and the degree of infilling are the main sensitive factors, and at the same time, high infilling degree shows a significant synergistic effect with high crystallization pressure. The BPNN method achieved the best prediction performance, with a R 2 of 0.992, RMSE of 0.302 MPa, and MAE of 0.148 MPa. The FEA-driven machine learning framework established provides a new analytical method for the quantitative characterization of micropore-scale salt crystallization stress concentration and the evaluation of physical salt corrosion damage in cement-based materials.

Research Square
Chongqing University (CN), Zhengzhou University (CN), Shandong Agricultural University (CN), Shandong University of Science and Technology (CN)
Building materials and conservation
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon