Artificial intelligence (AI) has emerged as an increasingly prevalent approach for the constitutive modeling of geomaterials. However, it encounters challenges in generalization when data are scarce, which severely restricts its applications, especially in complex engineering. This research develops a novel data-driven constitutive modeling framework that effectively leverages multi-fidelity data and incorporates two core technologies: (1) encoding the generalized plasticity theory in neural network architectures to develop models with strong generalization capabilities using large amounts of low-fidelity data; (2) implementing the infusion of scarce, high-fidelity experimental data based on transfer learning to enhance prediction accuracy. A comprehensive sensitivity analysis demonstrates the impact of each component in this data-driven model on capturing material strength and deformation features, which guides the integration strategy of high-fidelity physical experimental data. Validation across various stress paths verifies the model’s accuracy and robustness. The engineering feasibility of the framework has been demonstrated through finite element method (FEM) simulations of a concrete face rockfill dam (CFRD), highlighting its ability to address complex geotechnical engineering problems and accurately reproduce stress–strain distributions. This study offers a feasible approach to developing high-fidelity, data-driven constitutive models, even in the context of sparse data.

In this paper, a cementitious capillary crystalline waterproofing admixture (CCCW) comprising different crystalline components was developed to prepare highly impermeable mortar. Sodium silicate served as the crystallizing agent, while tetrasodium ethylenediaminetetraacetate (EDTA4Na) and glycine functioned as the complexation catalysts. Calcium carbonate can promote crystal growth and cement hydration by providing nucleation sites. The above four crystalline components facilitate the production of hydrated calcium silicate (C-S-H), ettringite (AFt), and calcium carbonate. Meanwhile, sodium acetate can plug pores through a process of selfcrystallization. Following the addition of the self-prepared CCCW, a substantial quantity of needle-and-rod crystals were generated and the porosity and average pore size exhibited a decline. The pore refinement was observed in both early-age and long-age samples, and the harmful pores were refined to harmless or less harmful pores. The utilization of CCCW was found to convert calcium hydroxide within the mortar, but its effect on pH value was relatively minor. The matrix demonstrated an impermeability strength of 1.1 MPa, with a corresponding impermeable strength ratio reaching 275%. Experimental data revealed a 15% reduction in water absorption compared to control samples. In mechanical properties, the compressive strength at 28 days rose from 43.3 MPa to 49.8 MPa, representing an approximately 15% increase. It also ensured excellent bonding strength between the old and new mortar. The bond strength experiments demonstrated that the loss of flexural strength decreased from 42.4% to 26.32%. Furthermore, the application of CCCW exerted a comprehensive protective effect, helping to resist mortar damage caused by a variety of erosion factors, including ions, liquids, and gases. In freezethaw erosion, the mass loss rate decreased from 6.77% to 2.05%. Additionally, the damage pattern shifted from deep cracking to skin peeling. The chloride ion corrosion testing revealed a decrease in the non-steady-state migration coefficient from 8.65 x 10_12 m2/s to 6.28 x 10_12 m2/s. The depth of carbonation for the samples reduced from 7.71 mm to 3.63 mm.

Applications in underwater environments impose stringent requirements on the toughness and workability of cement-based materials. In this study, acrylamide in-situ polymerization was introduced to modify conventional underwater cementitious materials without the use of superplasticizers, and its effects on material performance were systematically investigated. Results indicate that before polymerization, the modified sample exhibited a 79.41% increase in flowability compared with the control group, demonstrating excellent grouting capacity. Once polymerization was triggered, rapid solidification occurred within 6 min. This makes it possible to control the setting time of cement. Furthermore, compared with the control group, the sample incorporating acrylamide and cellulose achieved a 99.79% increase in underwater flexural strength and a 318.62% increase in ultimate strain. The underwater bond strength with old mortar reached 5.25 MPa, highlighting the material's potential for underwater repair applications. This work provides theoretical insights and experimental evidence for optimizing underwater cement-based materials.

Shape-stabilized phase change materials (SSPCMs) are promising for thermally functional mortars, but slight leakage during mixing and phase transition can degrade the particle–matrix interface and reduce durability. This study developed a surfactant-free interfacial self-assembly route to build a hydrophobic silica shell on diatomite/n-hexadecane SSPCM particles. The shell increased the particle contact angle from 112.1° to 140.2° and reduced leakage to 0.12% after 4 h at 60 °C. At equal PCM dosage, the shell-modified mortar showed lower water absorption and chloride diffusion and higher compressive strength than mortar with unmodified SSPCM. After 500 thermal cycles between 0 and 40 °C, the shell-modified mortar retained a melting enthalpy of 8.85 J/g with only 1.8% latent-heat loss. Microscopy and nanoindentation indicated a denser interfacial region, showing that a simple mineral shell can improve thermal-storage mortar reliability for building applications.

The superabsorbent polymer (SAP) offers a promising avenue for self-healing in cementitious materials, yet its efficacy is often limited and accompanied by significant mechanical strength loss. To overcome these limitations, an organic-inorganic hybrid SAP (SAP NS) was designed to couple rapid physical crack sealing with sustained chemical precipitation while keeping compressive strength loss low. SAP NS was synthesized from a poly(acrylamide-sodium acrylate) polymer modified with nano-silica and sodium alginate (SA), and simultaneously loaded with sodium carbonate. The structure and swelling behavior of SAP NS were characterized, and its effects on cement paste hydration, compressive strength, and self-healing performance were evaluated. The results demonstrate that SAP NS was successfully synthesized and exhibited excellent water absorption and mineralization capacity. Incorporation of 0.5% SAP NS into cement paste promoted cement hydration, resulting in a denser microstructure and a low 28-day compressive strength reduction of only 2.7%. Notably, it achieved over 90% healing ratio for cracks ranging from 100 to 300 mu m in width, nearly 90% recovery of impermeability, and over 85% recovery of compressive strength. Microstructural analysis revealed the underlying selfhealing mechanism, confirming that SAP NS facilitates the crystallization of calcite, resulting in dense self-healing products. These findings confirm the viability of SAP NS in achieving efficient self-healing while effectively addressing the critical issue of mechanical strength loss.

Accurate deformation analysis is crucial for the safety assessment and risk management of high earth-rock dams. While conventional surrogate-assisted optimization improves prediction accuracy, the neglect of intrinsic physical parameter correlations often leads to non-unique solutions, limited accuracy gains, and numerical divergence. This study proposes a physics-constrained digital twin (DT) framework that enables high-fidelity virtual-physical synchronization. The key innovation is a physical constraint mechanism utilizing a beta-variational autoencoder (beta-VAE) to extract parameter correlations from global experimental datasets as prior knowledge. By integrating this mechanism with multi-objective optimization and an elite archiving strategy, the framework ensures stable and physically consistent model evolution. Validated on the 303 m high LHK dam, the results demonstrate a 28 % improvement in prediction accuracy and a transition to near real-time computational performance. This framework provides a more reliable and physically consistent modeling approach for intelligent dam operation and lifecycle risk assessment.

The key of granular photoelastic experiments lies in the efficient and accurate measurements of contact forces from photoelastic images. Although some studies have attempted to address this task through deep learning, their performance on experimental images remains limited due to the scarcity of well-annotated experimental data. To overcome this limitation, this study proposes a style transfer-based force measurement model (SFMM) for reconstructing contact forces from experimental photoelastic images. The proposed framework consists of three stages: (1) particle detection, (2) style transfer using generative AI to construct a realistic experimental image dataset, and (3) prediction of contact position, force magnitude and impact angle. Experimental results demonstrate that the model trained on generated images through style transfer can be effectively applied to experimental images, enabling reliable measurements of contact forces under limited well-annotated experimental data. Compared with conventional force resolving method, the proposed method achieves a nearly threefold increase in processing speed and an 11.49% improvement in contact force reconstruction accuracy. Furthermore, the stress-force-fabric (SFF) relationship is validated using the proposed model, and proven to be consistently independent of density, particle size distribution, and loading path. This work presents the first application of deep learning to inter-particle contact force measurement in experimental photoelastic images, establishing a methodological foundation for experimental investigations of granular materials.

Polyhedral oligomeric silsesquioxane (POSS), as an organic-inorganic hybrid material with tunable functionality, is innovatively applied in this study to regulate cement hydration and microstructural evolution. The hydration kinetics, hydration products, and compressive strength of Portland cement incorporating four POSS, including octamethyl-silsesquioxane (POSS-A), octavinyl-silsesquioxane (POSS-B), octaphenyl-silsesquioxane (POSS-C), and poly methylsilsesquioxane (PMSQ), were systematically studied. Through isothermal calorimetry, XRD, and TGA analysis, the results reveal that POSS influences hydration via synergistic physical filling and chemical interactions. Specifically, POSS-A and POSS-B enhance early hydration in the first three days, promoting the formation of calcium hydroxide (CH), and facilitate secondary hydration by consuming CH at later hydration stages. In contrast, POSS-C and PMSQ retard hydration by prolonging the induction period. Mechanical tests indicate that optimal POSS-A or POSS-B incorporation (0.2% similar to 0.6%) enhances 28-day compressive strength, whereas excessive content slightly reduces performance. All POSS types exhibit excellent colloidal stability in alkaline environments, ensuring uniform dispersion and improving the fluidity of mortar. This work demonstrates the potential of POSS for designing high-performance cement-based materials through the tailored control of hydration.

Recent studies indicate that cyclic wetting-drying processes, driven by factors such as rainfall, evaporation, and reservoir level fluctuations, can significantly degrade the mechanical properties of rockfill materials. In this study, to unveil the microscopic mechanisms controlling this mechanical deterioration, a discrete-element method (DEM) simulation strategy based on the deterioration mechanism of rockfill particles is proposed, incorporating a deterioration coefficient derived from single-particle crushing tests to model the effects of cyclic wetting-drying on the mechanical behaviors. The finite-difference method and DEM coupling strategy is employed to more accurately replicate laboratory triaxial compression tests of rockfill materials. Then, triaxial compression tests are conducted to investigate the effect of cyclic wetting-drying on the macroscopic and microscopic responses of rockfill materials. DEM results demonstrate that the cyclic wetting-drying process significantly deteriorate the shear strength and secant modulus of rockfill materials, shifting the deformation behavior from strong dilation to weak dilation or even contraction. Moreover, cyclic wetting-drying reduces strong contact forces within the granular assembly, making crushed particles in the weak-contact subnetwork more susceptible to contact sliding. Finally, through mesoscopic anisotropy analysis, we conclude the origin of the reduction of shear strength under cyclic wetting-drying conditions. Our study illustrates the effects of cyclic wetting-drying on the responses of rockfill materials, offering novel insights for better modeling and prediction of their complex mechanical behaviors.

Shifts in temperature and humidity greatly affect the deformation and stress in concrete, especially during early age. While substantial research has explored the effects of temperature, there is limited study on the role of humidity. To accurately investigate the deformation and stress characteristics of concrete under service conditions influenced by humidity changes, this paper develops a modified chemo-thermo-hygro-mechanical (CTHM) coupled model. The model establishes a relationship between hydration, temperature, humidity, and deformation. It not only effectively describes the influence of temperature variations on humidity and their reciprocal effects but also a comprehensive description of deformation and stress characteristics driven by temperature and humidity changes. The suggested model is capable of more precisely calculating the humidity variations during the early stages of concrete and demonstrate the significant contribution of humidity shrinkage to the overall deformation. The model is validated through uniaxial diffusion drying, self-desiccation and autogenous shrinkage experiments under various water-to-cement (w/c) ratios and curing environments, demonstrating good adaptability and accurate simulation of humidity and deformation evolution processes. This model is utilized to replicate the concrete block placement process and simulate the mesoscale uniaxial diffusion drying in concrete. The results show a robust coupling between the internal temperature and relative humidity (RH) in concrete, accompanied by significant humidity shrinkage occurring within the material. The model facilitates safety evaluations and lifespan forecasts for hydraulic concrete structures.

To address the limitation that single-scale modification cannot comprehensively enhance performance of the cement-based materials, this study introduces a novel multi-scale toughening strategy by combining in-situ polymerized sodium acrylate (SA) with polyethylene (PE) fibers. This methodology concurrently reinforces the cement matrix and the fiber-matrix interface, thereby surpassing the constraints of traditional single-scale enhancement techniques. Mechanical testing revealed that an optimal formulation comprising 3 wt% SA and 0.75 vol% PE fibers resulted in a 114% increase in flexural strength relative to the neat paste, while preserving compressive strength. Microscopic investigations demonstrated that the in-situ polymerized SA established a dense polymeric network, which not only occupied microvoids within the cement matrix but also chemically grafted onto the hydrophobic fiber surfaces, enhancing physicochemical compatibility and interfacial bonding. Single-fiber pullout experiments further substantiated these findings, showing improvements of 46.5% in frictional bonding and 34.1% in chemical bonding energy, thereby confirming the efficacy of this synergistic system. By leveraging the combined mechanisms of fiber bridging, matrix densification, and interfacial strengthening, this research delineates a multiscale toughening mechanism encompassing the matrix, interface, and fiber scales.

This paper studied self-assembled cellulose nanofibers (SCNF) with in-situ polymerization potential, aiming to enhance the fiber-matrix interface bonding of engineered geopolymer composites (EGC). The freeze-thawing method was used to introduce cellulose oligomers into the alkaline solution, ensuring the simultaneous occurrence of SCNF self-assembly and geopolymerization during curing. This process facilitated the formation of interpenetrating organic-inorganic networks within EGC. Microstructural characterization confirmed that SCNF improved the fiber-matrix interface through physical bridging and chemical bonding. Single fiber pullout test showed that SCNF increased the frictional bond by 18.6 % and the chemical bonding energy by 657 %, while reducing the slip-hardening coefficient beta by 68.6 %. The results revealed a two-fold effect of SCNF: the interface bonding was strengthened but the slip-hardening behavior of PVA fibers was suppressed. Additionally, compressive and uniaxial tensile tests were conducted to evaluate the modification effect of SCNF on the mechanical properties of EGC. The optimal SCNF content was determined to be 0.5 wt%, which improved the compressive strength, tensile strength, strain capacity, and energy absorption of the EGC by 46.3 %, 44.2 %, 32.6 %, and 93.1 %, respectively. These findings provide direct and quantitative support for developing highperformance geopolymer composites.

Cracking in early-age concrete is a common and complex problem, which is hard to study by physical experiments and numerical methods. This paper establishes a new hydration-thermo-hygro-mechanical phase-field coupling model for the fracture process in early-age concrete. The established model not only considers the influence of crack on hydration and heat conduction, but also takes into account their impact on moisture diffusion. Unlike previous multi-physics coupling models, an additional source term is introduced to reflect the changes in convection boundaries caused by crack generation, and the crack evolution is automatically tracked using a phase-field variable. After incorporating the additional source term, the relative humidity distribution near the crack tip agrees more closely with ambient relative humidity, while the model's average relative humidity decreases compared with the case without the source term. This implies that ambient relative humidity can rapidly penetrate the concrete along crack surfaces, which will aggravate cracking and increase the number of through-cracks. Mesoscale analysis reveals that the relative humidity becomes highly localized and preferentially distributed along the major cracks. This phenomenon has not been observed in previous studies. Overall, the proposed multi-physics coupling model provides new theoretical insights into the prevention and control of early-age concrete cracking.

Sulfate-induced cracking shortens the service life of concrete structures. Numerical modeling is a valuable tool for investigating the degradation process. Most previous models can assess the damage extent, but struggle to predict cracking induced by erosion. This study proposes a coupled chemical-transport-mechanical phase-field model to effectively simulate the cracking process of sulfate-eroded concrete. The diffusion-reaction process is modeled based on transport law and reaction kinetics. A simplified kinetic equation is employed to describe the calcium leaching phenomenon. By employing the phase-field model, discrete erosion cracks are converted into regularized cracks, enabling easy coupling of the cracking process with the diffusion-reaction process. The cracking driving force in the phase-field model is calculated by the expansion strain, which is derived by solving the diffusion-reaction model. A new piecewise function is used to describe the influence of cracks and pores on ion transport, achieving bidirectional coupling between the cracking and transport processes. By solving the phase-field equations, complex erosion cracks can be automatically predicted. The calculation results align well with experimental data and can reproduce the transverse cracks observed in the erosion-expansion experiment. Compared to other models, the proposed model achieves more accurate results with a larger residual error. Furthermore, the deterioration of concrete column corners under various factors is simulated, and the significance of different factors and their interactions is analyzed, providing new insights for enhancing the durability of concrete structures in sulfate environments.

Rubber is extensively utilized in modified concrete and cement mortar, which enhances the materials' resistance to erosion but results in a reduction of their mechanical properties. Numerous researchers have endeavored to mitigate the decrease in mechanical characteristics caused by the low binding force between rubber particles and cement matrix by incorporating ultrafine rubber powder. While micrometer-sized rubber particles are commonly employed, the efficacy of nanorubber powder remains to be comprehensively investigated. In this study, nanonitrile rubber (NNR) powder was incorporated into Portland cement pastes, and detection techniques were used to analyze the properties of NNR modified cement pastes, including calorimetry, XRD, FTIR, SEM, and MIP. Experimental findings manifest that the addition of NNR significantly alters the pore structure and microscopic morphology of cement paste, increasing the Ca(OH)2 concentration while decreasing the CaCO3 content. Furthermore, this study also contrasts the effects of different mass fractions of NNR on macroscopic performance of the cement pastes, including mechanical properties and durability. The results reveal that the incorporation of NNR brings about a decrease in the mechanical properties of the samples, but an improvement in their durability. Specifically, the water absorption rate of the samples decreased by 80.8 %, the carbonation depth was reduced by 59.3 %, the chloride ion diffusion coefficient decreased by 66.8 %, and the strength failure rate before and after the freeze-thaw test was reduced by 49.9 %.

Cracks in early-age concrete result from the competition between the continuously increasing stress levels and the evolution of material properties. Accurately predicting the temperature and the hydration degree evolution in early-age concrete is crucial for accurately predicting crack propagation. Therefore, this paper employs a new hydration model to more accurately reflect the temperature and the hydration degree evolution during the early hydration stage. Subsequently, a length-scale insensitive hydration-thermal-mechanical-phase-field coupling model is established. The cracks in early-age concrete are automatically tracked through a phase-field variable. Based on the established multi-field coupling model, crack propagation in early-age ordinary concrete and roller-compacted concrete is simulated. The simulation results agree well with the experiments. The simulation results indicate that the improved hydration model leads to a faster hydration rate and the time for crack emergence and penetration is advanced. In addition, the proposed model will obtain a more severe crack distribution for large-scale structures. The proposed model can be effectively applied to mass concrete structures.

The discrete element method (DEM) is proving to be a reliable tool for studying the behavior of granular materials and has been increasingly used in recent years. The accuracy of a DEM model depends heavily on the accuracy of the particle property parameters chosen which is of vital importance for studying the mechanical properties of granular materials. However, the existing DEM parameter calibration methods are limited in terms of applicability, and the trial-and-error method remains the most common way for DEM parameter calibration. This paper presents a novel calibration method for DEM parameters using the multi-objective tree-structured parzen estimator algorithm based on prior physical information (MOTPE-PPI). The MOTPE-PPI does not rely on the training datasets and may optimize with every single test, significantly reducing the computational efforts for DEM simulation. Moreover, MOTPE-PPI is suitable for a variety of contact models and damping parameters in DEM simulation, showing robust applicability and practical feasibility. Taking an example, the DEM parameters of sandy gravel material collected from Dashixia rockfill dam in China are calibrated using MOTPE-PPI in the paper. The prior physical information is obtained through a series of triaxial loading–unloading tests, single-particle crushing tests, and literature research. Seven parameters in the rolling resistance linear contact model and breakage model are considered, and the optimization process takes only 25 iterations. Through quantitative comparison with existing parameter calibration methods, the high efficiency and wide applicability of the DEM parameter calibration method proposed in this study. The calibrated DEM parameters are used to investigate the hysteretic behavior and deformation characteristics of the granular material, revealing that the accumulation of plastic strain and resilient modulus is related to confining pressure, stress level, and the number of cycles.

Landslides pose significant threats to human life, property, and critical infrastructure. Due to complex landslide dynamics and the lack of high-quality monitoring data, the timely and accurate prediction of landslide displacement remains challenging. This study proposes a deep learning (DL) model to predict landslide displacement by integrating multiscale interferometric synthetic aperture radar (InSAR) and global navigation satellite system (GNSS) monitoring data. The InSAR technique provides high-density surface measurements of the landslide, while the GNSS technique captures high-precision and real-time landslide dynamics. We first extract temporal features from high-frequency GNSS data and build a heterogeneous spatial network to represent the topological relationship between the multiscale datasets. Two alternative strategies are developed to fuse GNSS and InSAR data by updating node features (attribute augmented model) or reshaping graph edge relationships (graph topological model). Then, a DL model composed of graph convolutional networks (GCNs) and gated recurrent units (GRUs) leverages the integrated global information to deliver high-precision landslide displacement prediction. Experimental results show that the proposed model achieves an MAE of 0.014 and the mean absolute percentage error (MAPE) of 7.8%, outperforming single-source models and accurately simulating displacement fluctuation signals of the landslide. The graph topological model excels with stable, strongly correlated monitoring data, while the attribute augmented model remains robust under weaker or fluctuating monitoring correlations. These findings underscore the necessity and feasibility of multisource monitoring data in landslide displacement prediction, providing a robust and scalable framework for landslide disaster prevention.

Shape optimization is one of the most critical phases in arch dam design and construction, aiming to reduce concrete volume and improve the dam stress distribution. Currently, the mainstream methods are still manual and empirical, thus lacking of efficiency and generalizability. Surrogate-assisted optimization demonstrates to be useful for enhancing structure design efficiency, yet it requires a significant amount of computationally expensive training data to ensure accurate outcomes. To accelerate the procedure, we propose a novel Paretoguided Active Learning (PgAL) framework. In the preprocessing step, the optimization mathematical model is established based on domain knowledge, and we introduce an automatic modeling technique to reduce the time cost of Finite Element (FE) simulation. Subsequently, the Gaussian Process-based PgAL is developed to accelerate the NSGA-II with the guidance of the prior information of the Pareto front. A planned ultra-high arch dam was selected as a case study, the proposed PgAL improves significantly over the traditional surrogate-assisted optimization methods, saving 70 % of the time cost to achieve similar accuracy. After optimization, the volume of dam and the volume of tensile stress region are reduced by 13.79 % and 26.57 %, respectively, achieving a good balance between economy and safety. This research provides an advanced manner for arch dam shape optimization, significantly enhancing the dam design, and may serve as a valuable reference for other similar shape optimization problems.