Lunar construction represents a strategic frontier in deep space exploration. Alkali-activated lunar regolith simulant (AALRS) is a highly promising in-situ construction material. Accurately characterizing, understanding and predicting its performance under extreme environments are critical to the safety of lunar structures during construction and service. However, the existing research shows significant discrepancies and controversies. These arise from a series of challenges, i.e., 1) the inherent compositional complexity and variability of lunar regolith simulant, 2) the coupled effects of multiple factors, 3) a general reliance on qualitative and localized understanding of microstructure, 4) the lack of an underlying foundation for understanding the mechanisms that influence macroscopic performance, and 5) the reliance on simple statistical analysis of experimental data without establishing quantitative association models between parameters and performance. To address these challenges, this study was to propose an interpretable, cross-scale and data-driven research paradigm of "algorithmic deconstruction-performance prediction-mechanism analysis".

AALRS pastes were prepared using lunar regolith simulant XJ-1, alkaline activators (i.e., sodium hydroxide or sodium silicate), and water at a specified alkali content of 0.075 and a water-to-binder ratio of 0.3. To simulate extreme environments, the samples were exposed to a vacuum (i.e, 20 ℃ and 100 Pa) and a thermal vacuum (i.e., 90 ℃ and 100 Pa) for 1, 3, 7, 14 d and 28 d, respectively. Macroscopically, the compressive strength was measured in a WAW-1000 testing machine at a loading rate of 480 N/s. The energy dispersive X-ray spectroscopy (EDS) mappings of key elements (i.e., Si, Al, Na, Ca, Mg, and Fe) were acquired in an accelerating voltage of 15 kV, a working distance of 10 mm, a dwell time of 100 μs/pixel, a process time of 6, and a channel count of 2048.

A quantitative analysis framework for EDS mappings driven by unsupervised clustering algorithms was developed. Implemented in Python, this framework automates the establishment of an interpretable analysis paradigm using libraries such as OpenCV, scikit-learn, and scikit-image. To address underfitting, overfitting, and poor performance of basic ML algorithms in complex hypothesis spaces, a two-level Stacking ensemble model was proposed. The first level integrates six heterogeneous base learners like AdaBoost, XGBoost, CatBoost, SVR, MLPNN, and RF. The predictions from these base learners served as input features for the second level by a 5-fold cross validation strategy. This second level employs a linear regression model as a meta-learner. To further enhance the predictive accuracy of the model, its configuration was optimized via applying six hyperparameter tuning algorithms (i.e., RS, BO, PSO, GWO, SA, and TPE) to efficiently search the hyperparameter space.

The results of compressive strength reveal the disparities governed by both the activation system and environment. The AALRS-SS consistently outperforms the AALRS-SH. The AALRS-SS under HV environment achieves a peak 28 d strength of 32.9 MPa, which is 3.65 times higher than its AALRS-SH counterpart under the same conditions. In addition, the HV environment proves crucial for the strength development, compared to the V environment. To understand the mechanisms behind these macroscopic differences, a quantitative EDS mapping analysis is employed. The results show that AALRS-SS generally contains a higher area fraction of reaction products and correspondingly lower proportions of unreacted XJ-1 and porosity. This indicates a higher overall degree of reaction and a denser microstructure in the AALRS-SS, providing a direct explanation for its superior strength. The HV environment facilitates a more homogeneous phase assemblage, compared to the greater heterogeneity observed in the V environment. The elemental analysis based on the superpixel clustering further reveals that calcium remains largely inert. The gels in the AALRS-SH exhibit a higher aluminum content, while those in the AALRS-SS are characterized at a higher silicon content. Based on the quantitative EDS analysis, the microstructural characteristics and reaction mechanisms of AALRS under different environments and activation systems are elucidated.

To capture the complex, non-linear relationships among material composition, environmental conditions, and mechanical performance, a two-level Stacking ensemble model is proposed. The model tuned with RS is identified as an optimal choice for its excellent balance of high performance and rapid computation after comparing multiple optimization algorithms on both accuracy and computational efficiency. This selected Stacking-RS model is interpreted based on the SHAP analysis to reveal the underlying strength development mechanisms. The analysis confirms the strong, positive, and monotonic effects of AC, CA and IT. In contrast, it identifies a critical negative threshold for MPD around 50 μm and reveals a non-monotonic influence of W/B. The effect of SM is complex, with significant interactions with other features. A graphical user interface (GUI) is developed based on the Stacking-RS model, thus facilitating the efficient design of AALRS. Furthermore, this study establishes a cross-scale, data-driven analytical framework that integrates algorithmic deconstruction, performance prediction, and mechanistic interpretation, wherein the microstructural analysis and macroscopic prediction modules mutually complement and validate each other to enable an intelligent understanding and accurate prediction of AALRS performance under extreme environments.

AALRS-SS under the HV environment exhibited the maximum strength. Its 28 d strength reached 32.9 MPa, which was 3.65 times greater than that of AALRS-SH (9.0 MPa). This was attributed to the SS system forming a higher proportion of gel products and a denser microstructure.

A quantitative EDS image analysis framework based on GMM and SLIC clustering algorithms was established. Compared with the V environment, the HV environment significantly increased the proportion of reaction products and compositional homogeneity, while reducing porosity. The reaction degree of AALRS-SS was significantly higher than that of AALRS-SH.

The reaction mechanisms of AALRS-SH and AALRS-SS under extreme environments were revealed. Under the V environment, AALRS-SH showed a significantly increased porosity with Na enrichment inside the pores. Under the HV environment, gels precipitated on the XJ-1 particle surfaces, forming a product layer that hindered subsequent dissolution, leading to a porous structure, and low strength. In contrast, AALRS-SS, due to the presence of silicate polymers, provided numerous nucleation sites through depolymerization under the HV environment, forming a relatively dense microstructure. Under the V environment, it provided a bonding force through self-assembly gelation of silicate polymers.

A Stacking ensemble model optimized by RS was identified as the best-performing model, showing excellent accuracy, generalization, and efficiency (i.e., R=0.9482, RMSE=5.58 MPa, MAE=4.07 MPa, and t=0.77 min).

The SHAP analysis (global, local, and dependence) showed that AC, CA, and IT were the most significant positive contributing features. The negative impact of MPD became more significant beyond 50 μm. The effect of W/B was non-monotonic. A GUI was developed based on the Stacking-RS model to enhance its practicality.

A data-driven, cross-scale intelligent analysis paradigm of "algorithmic deconstruction-performance prediction-mechanism analysis" was proposed. Its core lay in the macro-prediction module, which was used for strength prediction and identifying key factors, and the microstructure module, providing direct experimental evidence to elucidate the underlying strength development mechanism.

Since the late nineteenth century, railway construction has been a key driver of national development in China. The Yuehan Railway is one of the earliest modern railways in the country, playing an important role in advancing transportation infrastructure and regional economic growth. Over the past century, the concrete structures of the Yuehan Railway have been exposed to complex environmental actions, including high humidity, seasonal fluctuations in the Yangtze River water level, chemical ions, and sediment-induced abrasion. These factors have collectively accelerated surface deterioration and microstructural degradation. Recent decreases in river water levels have revealed the concrete remains of Xujia Peng Station, providing a rare opportunity to examine century-old concrete. This study conducts field sampling and laboratory analyses to characterize the composition and microstructure evolution of the relic concrete. Techniques including X-ray diffraction (XRD), thermogravimetric analysis (TG), scanning electron microscopy (SEM), backscattered electron imaging (BSE), and X-ray computed tomography (X-CT) are employed. The findings offer scientific support for the preservation of historic constructions and provide reference insights for durability design in cement materials.

Concrete samples were collected during the dry season from the exposed remains of the Xujia Peng Station section of the Yuehan Railway. Representative fragments were obtained from pile caps and bridge piers via manual chiseling, while intact cores were extracted from sleeper components. After drying and surface preparation, samples were processed for mechanical testing and microstructural analysis. Compressive strength and chloride ion penetration resistance were measured on φ 100 mm × h 100 mm and φ 100 mm × h 50 mm cylindrical specimens, respectively. Phase assemblages were quantified using (XRD) with Rietveld refinement, complemented by thermogravimetric analysis. Microstructural morphology, ITZ distribution, and elemental composition were examined via (SEM), backscattered electron imaging with image analysis (BSE/IA), and energy dispersive spectroscopy (EDS). Three-dimensional pore architecture was evaluated using high-resolution X-CT, enabling quantitative pore size distribution.

The sleeper concrete exhibited exceptional long-term stability, with an average compressive strength of 110.68 MPa and an electrical charge passed of 868 C, corresponding to a Q-IV chloride resistance grade. XRD and TG/DTG analyses revealed nearly complete hydration of clinker phases after nearly a century of service. C₃A and C₄AF were fully consumed, and only trace amounts of C₃S and C₂S remained. Due to prolonged exposure to air, pile cap and pier samples showed severe carbonation, with CaCO₃ mass fractions exceeding 20% and minimal portlandite. By contrast, sleeper samples exhibited limited carbonation. Quantitative BSE/IA demonstrated clear ITZ gradients in all samples. The ITZ thicknesses for pile cap, sleeper, and pier concretes were approximately 50, 30 μm, and 40 μm, respectively. ITZ regions consistently exhibited elevated porosity and reduced hydration product content near aggregate surfaces. The gradient of hydration product volume fraction stabilized beyond the ITZ boundary, indicating a distinct microstructural transition from porous interface zones to dense bulk matrix. SEM/EDS analyses revealed dense C-S-H gel, AFt/AFm products, localized carbonate phases, and well-preserved matrix integrity in the sleeper samples. While microcracks existed near aggregate boundaries, the overall matrix remained chemically stable. X-CT imaging showed a predominantly disconnected pore system consisting of fine capillary pores and a small number of large voids. The critical pore volume of the sleeper concrete was 0.1 mm³, above which macropores dominated total porosity but did not form continuous pathways.

This study provides a comprehensive microstructural and performance assessment of century-old Yuehan Railway concrete. Sleeper concrete retained high compressive strength (110.68 MPa) and low chloride permeability (868 C), demonstrating excellent durability after nearly 100 years of service. Pile cap and pier concretes underwent significant carbonation, while sleeper concrete exhibited limited carbonation due to environmental shielding. ITZ thicknesses were approximately 50 μm (pile cap), 30 μm (sleeper), and 40 μm (pier), with higher porosity and microstructural heterogeneity compared to bulk matrix. X-CT analysis identified a critical pore volume of 0.1 mm³, indicating that large isolated pores govern overall porosity without forming permeable networks. Overall, the Yuehan Railway concrete demonstrates outstanding long-term mechanical and durability properties, offering valuable insights for cementitious material design.

Sulfoaluminate cement-based grouting materials (SAGMs) have emerged as a high-performance solution in modern civil engineering, prized for their rapid setting, high early strength, and volumetric stability in applications ranging from rock reinforcement and concrete crack remediation to mine tunneling support. However, the intrinsic drawback of sulfoaluminate cement-its exceedingly rapid hydration kinetics and short setting times-severely restricts the operational window for grouting operations, challenging the feasibility of large-scale field applications. While chemical admixtures are conventionally employed to disperse cement particles, retard setting, and enhance fluidity to safeguard early-age workability, the distinct mineralogical composition of sulfoaluminate cement (characterized by ye'elimite and anhydrite) compared to ordinary Portland cement creates profound compatibility challenges. Conventional superplasticizers often fail to perform optimally, leading to issues such as flash setting or uncontrolled fluidity loss. Therefore, the rational design of molecularly tailored admixtures specifically compatible with the unique chemistry of SAGMs is of paramount importance. In this study, 3-mercaptopropionic acid (MPA) was utilized as a chain transfer agent to regulate molecular weight during the free radical polymerization of acrylic acid (AA), 2-acrylamido-2-methyl-1-propanesulfonic acid (AMPS), and acrylamide (AM), yielding a series of AA-AMPS-AM. The influence of MPA dosage on polymer architecture was systematically investigated, and the effects of molecular weight variation on the rheological properties, workability retention, and hydration mechanisms of SAGMs were evaluated, with a specific focus on dissolution-precipitation dynamics.

A suite of AA-AMPS-AM with tailored molecular weights was synthesized by modulating the MPA dosage during aqueous free radical polymerization. The successful synthesis and structural confirmation of these polymers were rigorously verified using Fourier-transform infrared spectroscopy (FT-IR), gel permeation chromatography (GPC), and proton nuclear magnetic resonance (¹H NMR). The performance of these polymers in SAGMs was quantitatively assessed according to Chinese National Standards: fluidity (GB/T 8077-2012), setting time (GB/T 1346-2011), and mechanical properties of hardened paste (GB/T 7897-2008). To elucidate the interaction mechanisms at the solid-liquid interface, adsorption behavior was probed using zeta potential measurements and total organic carbon (TOC) analysis. The modulation of the hydration process was investigated through the framework of dissolution-precipitation equilibrium, utilizing conductivity measurements and inductively coupled plasma (ICP) spectroscopy to track ionic concentration variations during mineral dissolution. The evolution of key hydration products was analyzed based on these ionic profiles and subsequent solid-phase characterization. Final hydration products were comprehensively identified using X-ray diffraction (XRD), thermogravimetric analysis (TG), laser particle size analysis, and scanning electron microscopy (SEM) for morphological assessment.

By maintaining a fixed monomer molar ratio of AA: AMPS: AM = 1:99:30 and adjusting the MPA dosage, water-reducing agents with systematically varied molecular weights were synthesized. Experimental results indicated that reducing the MPA dosage from 0.6% to 0.2% led to a controlled decrease in the molecular weight of AA-AMPS-AM. This molecular tuning significantly enhanced the fluidity of SAGMs; specifically, the incorporation of polymer-2, polymer-4, and polymer-6 increased the initial fluidity to 370, 365 mm, and 345 mm, respectively. Notably, the formulation containing polymer-2 exhibited exceptional fluidity retention, maintaining 100 mm of flowability even after 60 min. Adsorption analysis revealed that lower MPA dosages reduced the polymer's affinity for mineral surfaces, allowing a greater proportion to remain in the liquid phase to act as a steric dispersant. AA-AMPS-AM influenced the hydration process of the SAGMs, a lower MPA dosage strengthened the polymer's inhibitory effect on the dissolution of anhydrous calcium sulfate and calcium sulfoaluminate. This is attributed to AA-AMPS-AM effectively coating the mineral surfaces, creating a physical barrier that limits direct contact with the aqueous solution. Furthermore, AA-AMPS-AM reduced both the rate and total heat release of hydration, with the retarding effect becoming more pronounced at lower MPA contents. The polymers suppressed AFt formation and refined the particle size from 1600 nm to 200 nm, altering AFt morphology from long, slender rods to short, stubby rods. Conversely, lower molecular weight polymers (higher MPA dosage) promoted the formation of an aluminum-rich gel phase, facilitating particle aggregation on the external surface.

This study establishes a clear structure-performance relationship for polymer in SAGMs, demonstrating that molecular architecture-governed by MPA dosage-dictates their efficacy as water reducers and hydration modifiers. Reducing the MPA dosage yields polymers with lower molecular weights and narrower distributions, which optimizes slurry fluidity and effectively retards setting. Crucially, this reduction diminishes polymer adsorption onto cement grains, the resultant increase in free polymer concentration in the pore solution is pivotal for sustaining fluidity. Furthermore, the adsorbed polymer layer acts as a steric barrier that inhibits the dissolution of reactive minerals (anhydrite and ye'elimite), disrupting the dissolution-precipitation equilibrium. This leads to a delayed and modified crystallization of AFt, shifting its morphology from acicular to stubby prismatic crystals. Finally, polymerization conditions favoring lower MPA content promote the co-precipitation of Al(OH)₃ gel alongside the polymer, which encapsulates fine particles and creates a distinctly rough surface topography on the final product.

Due to the large amount of old mortar attached to the surface of recycled coarse aggregate (RA), its water absorption capacity is significantly higher than that of natural coarse aggregate. Consequently, the water saturation degree (S_W) of RA plays a crucial role in influencing the microstructure and resistance to Cl^- erosion of recycled aggregate concrete (RAC). However, the specific effects of different saturation states on RAC performance remain controversial. Some studies suggest that water released from surface-saturated RA weakens the newly formed interfacial transition zone (ITZ), reducing resistance to chloride penetration, whereas others argue that its desorption behavior acts as an internal curing agent that enhances RAC durability. Furthermore, the influence of S_W on the microstructure under different water control schemes-namely, constant mixing water supply (labelled as S1) and constant total water supply (labelled as S2)-has yet to be fully elucidated. Therefore, this study aims to systematically investigate the effects of varying S_W of RA on chloride diffusion and binding behavior in RAC under different control schemes, and to elucidate the underlying microscopic mechanisms.

The RA used in this study had a particle size range of 4.8-31.5 mm, an apparent density of 2650 kg/m³, a water absorption rate of 5.4%, a crushing index of 16.6%, and an old mortar content of 36.7%. The fine aggregate was river sand with a fineness modulus of 2.0, a moisture content of 1.0%, and a mud content of 0.8%. Ordinary Portland cement 42.5 R with a specific surface area of 345 m²/kg and a density of 3.12 g/cm³ was used. A polycarboxylate superplasticizer with a solid content of 40% and a water reduction rate of 25% was added to control the workability of fresh concrete. Under both S1 and S2, the effective water-to-cement ratio (W_e/C) and the total water-to-cement ratio of RAC were maintained at 0.55. The S_W of RA was adjusted to 0%, 50%, 80%, and 100% by controlling the soaking time.

After curing for 28 d, the rapid chloride migration coefficient (D_RCM) of RAC was determined using the rapid chloride migration method. Cubic specimens were then immersed in a 5% NaCl solution for one-dimensional natural chloride ingress tests. At exposure durations of 60, 120, 180 d, and 240 d, the concentration profiles of free Cl^- and total Cl^- were measured using potentiometric titration. Based on Fick's second law of diffusion, the effective and apparent chloride diffusion coefficients (D_a and D_e) were calculated. In addition, after 28 d of curing, the distribution of unhydrated cement content and the porosity in the new ITZ were analyzed using backscattered electron imaging (BSE), while the pore size distribution of the new mortar was determined by mercury intrusion porosimetry (MIP). Furthermore, scanning electron microscopy combined with energy-dispersive X-ray spectroscopy (SEM/EDS) and X-ray diffraction (XRD) were employed to characterize the microstructure and phase composition of chloride-binding products on the concrete surface layer after 240 d of exposure.

Under S1, as S_W of RA increases, the content of unhydrated cement content in the new ITZ decreases, while porosity increases. In contrast, under S2, the opposite trend is observed. MIP results further indicate that under S1, total porosity, effective porosity, and the proportion of macropores (>100 nm) in the new mortar all increase significantly with increasing S_W, whereas under S2, these parameters decrease. Under S1, the water absorption capacity of RA decreases with increasing S_W; moreover, RA with high S_W may release water into the fresh mixture, resulting in an increase in the actual W_e/C ratio of the cement matrix. Conversely, under S2, higher S_W corresponds to a lower effective W_e/C ratio during the early hydration stage, thereby suppressing both hydration and pore development.

Since higher porosity and improved pore connectivity lead to increased chloride diffusivity, under S1, the values of D_RCM, D_e, and D_a all increase with increasing S_W. In contrast, under S2, the chloride diffusion coefficients decrease as S_W increases. SEM/EDS and XRD analyses confirm the occurrence of chloride binding in RAC, revealing the presence of chloride-bearing C-S-H gel and Friedel's salt after chloride ingress. During the early exposure stage (60-120 d), RAC under S1 exhibits a stronger chloride-binding capacity due to a higher initial degree of hydration. Under S2, however, a larger amount of residual unhydrated cement continues to hydrate during prolonged immersion, resulting in superior long-term chloride binding capacity. Moreover, under S1, the chloride binding capacity decreases with increasing S_W, whereas under S2, the opposite trend is observed. The decay rate of the chloride diffusion coefficient increases with S_W under S1 but decreases under S2. This may be attributed to the fact that, compared with S2, higher S_W under S1 leads to a larger early-age chloride diffusion coefficient, which promotes the formation of more corrosion products and significantly refines the pore structure, thereby accelerating the time-dependent reduction in chloride diffusivity.

Under S1, increasing S_W reduces the water absorption capacity of RA, and highly saturated RA may release water into the fresh mixture. This results in an increased W_e/C ratio in the new ITZ and mortar, leading to higher porosity and a greater proportion of macropores, and consequently higher chloride diffusion coefficients. In addition, the extra moisture promotes the hydration reaction, reducing the content of unhydrated cement particles in the new mortar and the new ITZ. In contrast, under S2, higher S_W corresponds to a lower effective W_e/C ratio during early hydration, which significantly suppresses hydration, pore development and chloride diffusion. SEM/EDS and XRD analyses confirm that chloride binding in RAC involves both physical adsorption and chemical binding. During the exposure period of 60-120 d, RAC under S1 exhibits a stronger chloride-binding capacity due to a higher initial degree of hydration. Under S2, however, the continued hydration of residual unhydrated cement during long-term immersion provides superior chloride binding capacity after 180 d. Furthermore, under S1, increasing S_W is detrimental to chloride binding capacity, whereas under S2, it is beneficial. The decay rate of the chloride diffusion coefficient increases with S_W under S1 but decreases under S2. This behavior is attributed to the larger initial diffusion coefficient under S1, which promotes the formation of more corrosion products and leads to progressive refinement of the pore structure over time.

Despite progress on applying data-driven methods to chemically activated aluminosilicate cementitious materials, several critical challenges remain. Firstly, the inherent complexity and variability of raw materials, combined with the absence of standardized design specifications, lead to high-dimensional and redundant datasets that compromise the model performance. To address this issue, it is necessary to establish standardized open databases covering precursor composition, reactivity, activator parameters, and curing conditions, as well as the knowledge-embedded hybrid models that incorporate geopolymerization mechanisms and thermodynamic constraints. Secondly, data scarcity is a pervasive challenge in this field, with most datasets characterized by small sample sizes, severely limiting model generalizability. This issue can be alleviated through multi-source data fusion, data augmentation techniques, and transfer learning for small-sample scenarios. Lastly, the existing inverse design relies primarily on a two-step prediction-then-optimization framework, whereas deep generative models for direct material design remain largely underexplored. Integrating these advanced models with reaction kinetics and microstructure evolution can be essential to realize end-to-end inverse design.

The evolution of artificial intelligence technology offers a significant potential for advancing chemico-activated aluminosilicate cementitious materials technology. This promises to serve as a robust reference for the deep integration of artificial intelligence into future material development. Future efforts should focus on developing standardized workflows, improving model interpretability and generalizability, as well as leveraging multi-scale simulations and deep generative models to accelerate material innovation. Data-driven methods will enable more accurate prediction and design of chemico-activated aluminosilicate materials due to these advancements, thereby contributing to the sustainable transformation of the construction industry.

This review summarizes the latest research progress on the performance changes and modulation of early-age concrete under vehicle-bridge coupled vibration, focusing on the analysis of vibration characteristics, concrete performance changes, and the underlying mechanisms, and discussing the existing modulation measures. Future research should develop lab-scale simulation methods that more closely reflect actual vibration characteristics, considering the coupling effects of temperature, humidity, load, and corrosion, to comprehensively explore the evolution of concrete performance under practical conditions. In addition, in-situ characterization systems for vibration loading should be also developed, using techniques such as 3D CT, NMR, DIC, and acoustic emission to capture the evolution of the concrete microstructure and crack dynamics, thus providing a basis for the design and life prediction of vibration-resistant concrete.

X-ray computed tomography (CT) is an important non-destructive technique for analyzing the microstructure of cement-based materials, which can characterize aggregate distribution, pore structure and interfacial transition zone. The accuracy of subsequent 3D reconstruction and quantitative analysis is highly dependent on the quality of image segmentation. However, concrete CT images generally have some problems such as low contrast between aggregate and matrix, blurred boundary and strong noise, which seriously limit the effect of conventional segmentation methods such as threshold method.

Deep learning performs well in complex image segmentation, but its training relies on large-scale and high-quality labeled data sets, while CT image labeling is difficult and time-consuming. In practice, there are often only a few labeled slices, leading to insufficient generalization ability and unstable performance of the model, especially in low contrast and fine structure areas. To solve the contradiction between model complexity and data availability, it is necessary to develop efficient learning strategies that can make full use of unlabeled data.

This paper was to propose a semi-supervised deep learning framework for automatic segmentation and three-dimensional reconstruction of concrete CT images. This method could effectively utilize a small number of labeled slices and a large number of unlabeled slices based on U-Net segmentation model and self-training mechanism. Through pseudo-label generation and confidence filtering, the boundary recognition and structural consistency under low contrast conditions could be improved.

The proposed framework consisted of three stages, i.e., initial supervised training, pseudo-label generation, and iterative semi-supervised optimization.

Firstly, U-Net was used as a backbone model, and its encoder-decoder structure was combined with jump connection, thus extracting high-level semantic and fine spatial features at the same time. The initial segmentation ability was obtained with 42 manually labeled CT slices, combined with Dice loss and binary cross entropy loss for training.

Secondly, the trained model was applied to 2000 unlabeled CT slices to generate pixel-level prediction and probability maps. Through the confidence threshold, the high-reliability region was retained as a pseudo-label, and the low-confidence region (usually near the fuzzy boundary) was excluded. After morphological operation, the isolated noise was removed, and the boundary continuity was improved.

Finally, the filtered pseudo-labels were merged with the original labeled data to form an augmented training set. The model was iteratively retrained on this mixed set, with labeled data providing stable supervision, and pseudo-labeled data supplementing structural information. Also, the learning rate attenuation strategy was introduced to ensure the stability and convergence of training. The Dice coefficient and the intersection-union ratio (IoU) were used to evaluate the segmentation performance.

The performance of manual annotation quality and initial supervision model is evaluated. The marking mask can be close to the aggregate boundary in different scales, indicating great marking consistency and reliability. The U-Net model trained with only 42 labeled slices can quickly learn the discriminant characteristics of aggregate and matrix, and obtain a better verification performance (i.e., the optimum verification Dice = 0.860 6). However, in low contrast and complex structure areas, the performance still fluctuates, indicating that the generalization ability is limited when the training data is extremely limited.

The semi-supervised learning strategy significantly improves the performance of the model. The pseudo-label visualization shows that the high confidence region is highly consistent with the actual aggregate distribution, and the uncertainty mainly appears at the aggregate-matrix interface. With iterative training, the indicators further improve. The Dice, IoU and accuracy of the final model (S-40) reach 0.971 2, 0.944 0 and 0.982 0, respectively, which are 11.7 %, 22.7 % and 7.4 % higher than the fully supervised baseline, respectively, and the loss is reduced by 82.9 %. These results show that high-confidence pseudo-labels can effectively enhance feature learning without compromising training stability.

Three-dimensional reconstruction demonstrates that the proposed semi-supervised framework effectively preserves the spatial continuity and geometric integrity of aggregate particles. Compared with the fully supervised baseline, the reconstructed aggregates exhibit markedly fewer adhesion artifacts, better-defined particle boundaries, and more complete representations of fine aggregates and complex interfaces. These improvements enable reliable quantitative characterization of aggregate morphology, connectivity, and spatial organization, demonstrating that the proposed segmentation framework provides high-fidelity input for three-dimensional reconstruction.

Quantitative analysis reveals that aggregate volumes follow a log-normal distribution and exhibit a pronounced surface area-volume scaling relationship (i.e., ), reflecting the geometric self-similarity of the reconstructed particles. Large particles contribute most of the total aggregate volume, whereas smaller particles occupy the interstitial spaces to form a dense skeleton-filling structure. The reconstructed aggregates are uniformly distributed throughout the specimen without evident segregation, and the extracted morphological descriptors yield a mean sphericity of 0.757, indicating moderate particle roundness while preserving angular characteristics. Collectively, these findings validate the reliability of the proposed reconstruction framework and provide a robust basis for quantitative three-dimensional microstructural characterization and aggregate gradation analysis.

This paper proposed a semi-supervised deep learning framework for concrete CT image segmentation and 3D reconstruction. This method effectively overcame the difficulties caused by limited labeled data and low imaging contrast via combining pseudo-label learning with U-Net architecture. The results showed that the model converged stably, the segmentation accuracy improved significantly, and an effective balance between manual labeling and high-precision microstructure analysis could be achieved. The framework could provide an extensible solution for the intelligent characterization of cement-based materials and have application prospects in the field of digital engineering.

Conventional development of low-carbon clinker mainly relies on empirical or mechanistic understanding of high-temperature and hydration reactions among clinker phases. This strategy, which combines experimental trial-and-error with mechanistic analysis, is generally time-consuming. Therefore, conventional development approaches are no longer sufficient to meet the demand for the rapid development of novel low-carbon clinkers. The development of high-throughput (HT) methods and machine learning (ML) has provided a new paradigm for the design of cement-based materials. However, available clinker data for ML-based clinker design are usually collected from scattered sources, and they differ substantially in preparation procedures, testing conditions, and compositional definitions, making them difficult to use directly. To address these issues, this study used clinker phases as compositional units and constructed a clinker dataset through a standardized HT preparation procedure. On this basis, the predictive performance of different ML models for the 28 d compressive strength of clinker was compared. Furthermore, the optimized surrogate model was coupled with a genetic algorithm to screen for low-carbon clinkers under constraints on 28 d compressive strength and CO₂ emissions. This approach enabled evaluation of the applicability of the proposed data-driven framework for the design of novel low-carbon clinkers.

In this study, 11 clinker phases were selected as the basis for compositional design, and each clinker composition prepared by dry-mixing these phases with gypsum in specified proportions was regarded as one specimen. A clinker dataset was constructed using an HT experimental procedure. The dataset contained information on phase composition, water-to-binder ratio (W/B), 3 and 28 d compressive strengths for a total of 264 specimens. Among them, 228 specimens were used for training and validation, while 36 specimens were reserved as a test set. The test set was further divided into in-system and out-of-system specimens to evaluate model stability. For model construction, five nonlinear ML algorithms were employed. The normalized phase composition after gypsum incorporation was used as the input. The 28 d compressive strength was used as the output. Model hyperparameters were determined by Bayesian optimization, and repeated constrained holdout validation was adopted to reduce the influence of random splitting on model evaluation. Model performance was evaluated using mean squared error, root mean squared error (RMSE), mean absolute error, and mean absolute percentage error. After the surrogate model had been determined, it was coupled with a genetic algorithm to search the phase composition space for low-carbon clinkers with a predicted 28 d compressive strength higher than 42.5 MPa and CO₂ emissions lower than 671 or 592 kg CO₂/t. These CO₂-emission thresholds were reduced by 15% and 25%, respectively, relative to Portland clinker.

Dataset analysis showed that the specimens in the dataset exhibited pronounced compositional dispersion and system-level heterogeneity. This heterogeneity was one of the main sources of model instability. The W/B gradually increased from the C3S/C2S-rich region toward regions enriched in aluminate/ferrite phases and C4A3$/C5S2$. The 3 d and 28 d compressive strengths of the training/validation set covered a range of 0-70 MPa, whereas the test set was mainly concentrated in the medium-to-high strength range. Some specimens exhibited strength regression from 3 d to 28 d. These complex strength-development patterns increased the difficulty of model learning but also provided useful information for identifying nonlinear relationships between composition and compressive strength. Validation results showed that, under the conditions of limited data and high compositional heterogeneity, controlled model capacity and moderate regularization were more beneficial for improving prediction stability. The optimized hyperparameters of different models all reflected a tendency to suppress overfitting. Among the models, the artificial neural network (ANN) adopted a relatively small architecture with regularization and achieved the best performance on the validation sets. The test-set results further demonstrated that the ANN had better cross-system generalization capability. Compared with the other models, the ANN achieved an overall RMSE of 3.9 MPa on the test set. Subsequently, the ANN was selected as the surrogate model and coupled with a genetic algorithm for the inverse design of low-carbon clinkers. When the CO₂-emission reduction constraint was tightened to 25%, the genetic algorithm further favored low-carbon phases and generated candidate clinkers dominated by aluminate phases or the C4A3$ phase, with CO2 emissions of approximately 555 kg CO₂/t. Experimental validation showed that the measured 28 d compressive strengths of the candidate clinker compositions were close to the model predictions. The prediction errors for all specimens were lower than 10% of the measured strengths.

This study constructs a clinker dataset based on a HT preparation procedure and establishes an ML prediction and constrained-optimization framework for screening low-carbon clinkers. The results show that the dataset is characterized by broad compositional coverage, local sparsity, and pronounced system heterogeneity. Therefore, the model needs to balance nonlinear representation capability with regularization constraints. Comparisons on the validation sets and the test set show that the low-capacity ANN with moderate regularization achieves the best performance in terms of prediction accuracy. Its RMSE is 3.9 MPa on the overall test set, and the RMSE values for the in-system and out-of-system subsets are 3.1 MPa and 4.2 MPa, respectively, indicating that it can serve as the surrogate model for inverse design of low-carbon clinker. XGBoost can be regarded as a competitive potential surrogate model, but it tends to produce relatively large prediction errors for some specimens. By coupling the ANN surrogate model with a genetic algorithm, four candidate low-carbon clinkers satisfying the 28 d compressive strength and CO₂ emission constraints are screened. Experimental validation confirms good agreement between the predicted and measured compressive strengths. These results demonstrate that the proposed dataset construction and surrogate model screening strategy can be used to explore of novel low-carbon clinkers.

The clinker calcination process involves high-temperature solid/liquid reactions, and the controllable preparation of multiple clinker phase compositions, which further affect subsequent synergistic hydration reactions. The relationship between composition and 28 d compressive strength is not simply linear, but rather a complex mapping governed by the coupling of multiple processes. Conventional point-by-point experimental screening makes it difficult to cover a sufficiently broad compositional space within a limited timeframe. This restricts the exploration efficiency of novel low-carbon clinker. Machine learning models can be used to learn the composition and property relationships between clinker phase and strength, and to support the screening of candidate clinker compositions. However, under small data conditions, these models are susceptible to insufficient training distribution coverage, resulting in a unstable cross-system predictive performance. Therefore, this study was to develop a self-iterative data augmentation framework to address the above-mentioned issues. This framework was to expand effective supervised information to optimize the model's predictive performance in sparsely sampled compositional regions. Subsequently, model interpretability analysis was conducted to identify the key clinker phases affecting 28 d compressive strength prediction and their response patterns. This could provide a testable model basis for the optimization of low-carbon clinker compositions.

This study defined each simulated clinker as one sample. The dataset contained 264 simulated clinker samples, with the compressive strengths ranging from 0 to 70.3 MPa. The model inputs consisted of 12 features, including 11 clinker phases and gypsum, while the output was the 28 d compressive strength. In the self-iterative data augmentation strategy, an artificial neural network (ANN) model was first trained using the experimental data. Virtual data were then generated under compositional constraints, and their 28 d compressive strengths were predicted by the model from the previous iteration to assign pseudo-labels. The virtual and experimental samples were subsequently combined for training, and the surrogate model was updated iteratively. Two sampling patterns were adopted for virtual sample generation, i.e., 1) Pattern A: dense sampling within the experimental samples' compositional systems, and 2) Pattern B: generating virtual samples at a ratio of 1 ∶ 1 from the in- and out-of-system regions of the experimental samples' compositional space. Before training, 36 experimental samples were set aside as a test set, which remained isolated throughout training, iteration and hyperparameter selection. The validation set consisted exclusively of real experimental samples, whereas virtual samples were included only in the training set. Model hyperparameters were determined based on the Bayesian optimization, with the validation-set mean squared error (MSE) as the selection criterion. Finally, the model predictive performance was evaluated by MSE, the coefficient of determination (R²), mean absolute error (MAE), and mean absolute percentage error (MAPE). After the final surrogate model was determined, multilevel model interpretation was further performed via combining constrained permutation feature importance, accumulated local effects and SHapley additive explanations (SHAP).

The self-iterative augmentation strategy effectively fills sparsely sampled regions, while maintaining feasible-domain constraints. This provides richer training information for subsequent accuracy evaluation and interpretability analysis of the surrogate model. Meanwhile, the data augmentation process does not generate a large number of near-duplicate samples, and the training set retains the necessary heterogeneity. The inclusion of virtual samples reduces the model's sensitivity to randomness in data partitioning. Moreover, the overall test set error and the in-/out-of-system errors decrease progressively over successive iterations, indicating an improved cross-system generalization performance. Pattern B in Iteration II achieves a favorable balance between predictive accuracy and extrapolation stability. Its overall test set MSE is 9.8, with in-system and out-of-system MSE values of 6.9 and 11.3, respectively. The R², MAE, and MAPE are 0.79, 2.36 and 5.11%, respectively. These results indicate that, under the current small data conditions, data augmentation can substantially improve the model's ability to identify performance trends. Subsequent interpretability analysis of the optimal surrogate model shows that different clinker phases can be classified as positive or negative driving phases. The analysis further indicates that when C₃A content exceeds approximately 10%, its positive driving effect on strength decreases rapidly. Although CA and C₁₂A₇ act as positive driving phases, their benefits diminish when their contents exceed 40%, and even shift to strongly negative responses.

The self-iterative data augmentation framework proposed could expand a supervised training information for low-carbon clinker systems characterized by small datasets and sparse compositional spaces. The evaluation bias introduced by pseudo-labels could be also suppressed under a rigorous evaluation strategy. The results showed that data augmentation significantly improved the prediction accuracy and cross-system generalization capability of the ANN surrogate model for 28 d compressive strength. In Iteration II, Pattern B achieved the optimum overall performance under the current data conditions. Although further iterations reduced some in-system errors, an error rebound was observed for out-of-system samples. This indicated that diminishing marginal information could gain from pseudo-labels constrained the model's generalization performance. The combined use of constrained PFI, ALE, and SHAP revealed the decision-making basis of the surrogate model in terms of global sensitivity, average response trends, and sample-level contributions. The results indicated that C₄A₃$ and CA were the main positive driving clinker phases for compressive strength enhancement, whereas C₃A, C₅S₂$ and C₂S were the main negative factors. Moreover, C₁₂A₇ exhibited a pronounced threshold effect. These findings could provide a testable reference for subsequent composition screening, constraint-boundary definition and inverse design of low-carbon clinker.

Low-density polyethylene (LDPE) is limited in its high-end applications due to insufficient mechanical and thermal properties. Using silicate minerals as fillers is an effective strategy, as they possess high modulus, thermal stability, and abundance. Anorthite (An), as a framework silicate mineral, is a highly promising candidate material. However, its hydrophilic surface is incompatible with the hydrophobic LDPE, thus hindering the performance. Silane coupling agent (i.e., KH570) can improve interfacial bonding, but its role in the An/LDPE system remains unclear. In this study, the effect of KH570-modified An on the LDPE performance was investigated, offering some insights into the development of high-performance LDPE composites and the efficient utilization of An resources.

The silane coupling agent KH570 was used to modify An. KH570 was hydrolyzed in an ethanol-water solution (7∶3 v/v) and reacted with An at 80 ℃ for 6 h. After centrifugation, washing, and drying, modified anorthite (K-An) was obtained. K-An/LDPE composites (in a mass ratio of 95 ∶ 5) were fabricated via melt blending with a twin-screw extruder (130-180 ℃) followed by hot pressing. The modification mechanism was characterized by X-ray diffraction (XRD), Fourier transform infrared spectroscopy (FT-IR), and X-ray photoelectron spectroscopy (XPS). The morphology and dispersion were characterized by scanning electron microscopy (SEM). The surface wettability was determined by contact angle measurement. The mechanical properties were determined by tensile testing (ISO 527-1), and the thermal stability was determined by thermogravimetric analysis/derivative thermogravimetry (TG/DTG).

KH570 effectively modifies the chemical properties of the surface of An without altering its crystal structure or morphology. The FT-IR and XPS spectra confirm the formation of covalent bonds, enhancing the hydrophobicity of the An surface and improving its dispersibility in LDPE. The mesoporous structure remains stable, and the distribution of small pores is improved, indicating a regulatory effect on the pore characteristics. The K-An/LDPE composite material exhibits excellent tensile strength, elongation, and thermal stability due to the rigidity of An, intense interfacial bonding, and thermal barrier effects, highlighting its potential as an LDPE reinforcing filler.

KH570 could react with -OH groups on the surface of An to form stable bonds, reducing surface polarity, while preserving its crystal morphology, thereby lowering the interfacial energy difference with LDPE and enhancing compatibility. As a result, K-An/LDPE composites modified with KH570 showed enhanced tensile strength, elongation at break, tensile modulus and thermal residue, which could be increased by 38.01%, 104.39%, 48.32%, and 15.05%, respectively. This was attributed to the rigidity and thermal stability of the An framework, intense interfacial bonding, and a uniform thermal barrier network.